Exponential Industry

This weeks digest:

Choosin' Texas for Manufacturing Energy

David Rogers — Sun, 12 Apr 2026 13:03:13 GMT

Shop Talk

Capturing this week’s zeitgeist

The Texas triangle of Houston, Austin, and Dallas have secured a manufacturing renaissance by doubling down on the gritty “substrate” of the physical world /SightBringer on X/. These cities have positioned themselves as the masters of hard power, offering the energy /Manufacturing Dive/, logistics, and buildable land necessary to turn AI abstractions into physical output. By prioritizing functional scale and industrial competence over aesthetic vibes, Texas has built a resilient machine capable of powering, cooling, and housing the future of global industry.

Key Drivers of Texas Expansion

Texas CHIPS Act: Governor Abbott signed legislation in 2023 to establish the Texas Semiconductor Innovation Fund /Texas Gov/ to attract and support these massive investments.
Workforce Development: Projects like Texas A&M’s $200M semiconductor project are being developed to support the expanding talent needs through 2028. /YouTube/
AI Demand: The surge in AI and data center needs is the primary driver, with companies aiming for vertical integration and increased localized production. /Manufacturing Dive/

Up Next: Hannover Messe 2026, April 20 - 24

Hannover Messe coming up next week and it’s a great place to see the entire exponential ecosystem!

The top 10 smart manufacturing technology vendors /IoT Analytics/

I will be focusing on the leaders and startups of scalable Data, AI, and Intelligence:

Databricks at Hannover Messe /Databricks/
Litmus at Hannover Messe 2026: The Industrial Data Foundation for Enterprise AI /Litmus/
HighByte at Hannover Messe 2026: Ecosystem in Action /HighByte/
AVEVA at Hannover Messe 2026: Harness your industrial intelligence /AVEVA/
Meet Zeta Alpha at Hannover Messe 2026 /Zeta Alpha/. See how Agentic AI is transforming how engineers and R&D teams discover, organize, and act on technical knowledge.
Connect With Azumuta at Hannover Messe 2026 /Azumuta/

Reply to this email if you would like to catch up in Hanover!

New in 2026: We are gifting a 1-year paid subscription to the first person to identify the content link AND cultural reference to the title/subject line of this week’s digest. Comment on this post with your guess before next week’s digest is released to be eligible!

Last week’s answer: From Industrial Versailles to Man on the Moon: Earthset From the Lunar Far Side /NASA/

Assembly Line

This week’s Industry 5.0 breakthroughs and frontier technologies of the built world.

This Is How AI Will Automate Every Industry

Sorcery got rare access inside Applied Intuition’s commercial garage with Co-Founders CEO Qasar Younis and CTO Peter Ludwig, where they’re building AI systems that power cars, trucks, mining equipment, defense systems, and more. In a never-before-seen tour to the public, Applied Intuition showcases where/how real machines across industries are being rebuilt with AI.

Qasar and Peter break down how one platform can run everything from a car to a boat to a mining site, and why the future of AI is no longer just software, but systems that move the real world.

A New Chapter for Pronto: We Have Officially Been Acquired by Atoms /Pronto/. As of today, Pronto is proud to become the core technology engine of Atoms Mining, one of three divisions within the company dedicated to digitizing the physical world through purpose-built, intelligent machines.

Inside a Warehouse Automation Project: How Sumitomo Drive Technologies Is Transforming Logistics and Reshoring Operations

Ocado to launch AI-powered Ocado IQ software at MODEX /AutomatedWarehouse/
Auditors for Dexory have issued a rare warning over gaps in its financial records after the British robotics firm Dexory reported surging revenues alongside deepening losses /City AM/. The accounts come just days after Dexory secured £8.5m from the British Business Bank as part of a wider Series C round, with backing from investors including Eurazeo, Atomico and Lakestar.

NOTE: MODEX 2026 is on display April 13th-16th in Atlanta, GA as the largest manufacturing and supply chain event in 2026.

X-Force Threat Intelligence Index 2026

For the fifth consecutive year, the manufacturing sector remained the most attacked industry, representing 27.7% of all incidents within the top industries. This ongoing targeting reflects its critical role in global supply chains and the high value of operational and intellectual property data.

Share of incidents by industry in 2025 and 2024. Source: IBM X-Force.

Attackers leveraged several methods to breach manufacturing systems, with exploitation of public-facing applications (32%) emerging as the most common vector. Valid accounts (domain) (16%) and external remote services (11%) were also prominent, reflecting attackers’ reliance on exploiting misconfigured or insufficiently secured access points.

Information to Atoms

Below are simulations and printed-on-silicon images from a scanning electron microscope (SEM) of a complex metal one (M1) layer printed in a single exposure with Substrate’s X-ray lithography tool, enabled by AlphaEvolve-optimized simulations. The printed patterns are equivalent in resolution to “high numerical aperture” (High-NA) lithography, which is equivalent to the requirements of the 2 nm semiconductor node.

Beyond optimizing for success, the real power of this framework is its ability to predict not just that something will fail, but why. When we understand where and why printing fidelity breaks down, we can compensate by adjusting other parameters, simulate, and then validate with real-world printing.

We are now using AlphaEvolve in other areas, from material discovery to automated circuit layouts. We are also exploring ways to combine AlphaEvolve with other models and techniques to speed up research. One promising area is integrating AlphaEvolve with models focused on scientific knowledge, such as Gemini 3 Deep Think, in a semi-autonomous research loop.

New Product Introduction

Highlighting new and innovative facilities, processes, products, and services

Intel Foundry Achieves Breakthrough with World’s Thinnest GaN Chiplet Technology

✍️ Author: Han Wui Then

Researchers at Intel Foundry have demonstrated a first-of-its-kind GaN chiplet technology built on 300 mm GaN-on-silicon wafers, marking a significant leap forward in semiconductor design. Presented at the 2025 IEEE International Electron Devices Meeting (IEDM), this work tackles one of the most pressing challenges in modern computing: how to deliver more power, speed, and efficiency in an increasingly compact space. To meet the demand of graphics processors, servers, and wireless networks for ever-greater performance, the Intel Foundry team developed an ultra-thin GaN chiplet — its base silicon measuring just 19 μm thick, roughly one-fifth the width of a human hair — along with the industry’s first fully monolithic on-die digital control circuits, all built using a single integrated manufacturing process.

Thinning a semiconductor wafer sounds straightforward, but doing it on a fully processed 300 mm GaN-on-silicon wafer — one that already contains all the transistors and metal wiring layers — without damaging those structures is a formidable engineering challenge. The Intel Foundry team accomplished this using a technique called stealth dicing before grinding (SDBG) /DISCO/, which uses a precisely controlled laser to create microscopic fracture lines inside the wafer before a mechanical grinding step reduces its thickness. The result is a GaN chiplet with an underlying silicon base of just 19 μm.

Real-Time Injection Molding Simulation, Powered by AI

Introducing Cadmould AI Solver, the world’s first neural physics engine for plastic injection molding. This technology is set to revolutionize design for manufacturing by combining the accuracy of physics engines with the speed of neural networks to deliver simulation results in seconds.

This interactive research preview allows you to test-drive our AI model live. Instead of waiting for results, you receive instant feedback on filling pattern, pressure, and temperature in seconds. It is capable of modeling complex physical interactions, including multi-gate scenarios, with accuracy that rivals traditional solvers. Please note that this preview is a focused showcase of the filling phase only.

Business Transactions

This week’s top funding events, acquisitions, and partnerships across industrial value chains.

🚀🕴️ Inside the SpaceX founder factory—and the race to solve the next generation of impossible problems /Fast Company/. Fast Company built a database of 400+ ex-SpaceX employees turned company builders, the largest ever. What they learned, how they’re applying it, and the future they’re building.

🇺🇸💿 SiFive Raises $400 Million to Accelerate High-Performance RISC-V Data Center Solutions; Company Valuation Now Stands at $3.65 Billion

SiFive, the gold standard for RISC-V processor IP, announced it has raised $400 million in an oversubscribed Series G financing to accelerate its high-performance data center roadmap. This equity funding round was led by Atreides Management with Apollo Global Management, NVIDIA, Point72 Turion, and T. Rowe Price Investment Management, Inc., alongside repeat investors Prosperity7 Ventures, and Sutter Hill Ventures. The financing values the company at $3.65 billion and will accelerate SiFive’s RISC-V CPU and AI IP solutions for the data center.

CPUs are critical in agentic AI systems because they excel at orchestrating complex system level coordination tasks that GPUs and accelerators aren’t designed to handle efficiently. As AI evolves toward more complex agentic models, efficient CPU performance is critical to expanding compute capacity within existing power envelopes. SiFive enables this transition by replacing complex, power-hungry legacy architectures with modern RISC-V CPUs that are inherently lower power. RISC-V integrates scalar, vector, and matrix compute into a single, standards-based interface that helps customers scale rapidly, significantly accelerating hardware development to match the speed of AI innovation.

Read more at Business Wire

🇺🇸⚗️ Membrane Technology & Research Secures $27 Million Investment to Accelerate Global Deployment of Membrane-Based Solutions

Membrane Technology & Research, Inc. (MTR), a global leader in membrane-based separation technologies, today announced it has secured a $27 million equity investment led by Climate Investment, with participation from Hartree Partners. Combined with MTR’s commercially-proven and scalable technology platform, this investment strengthens the company’s position to accelerate global deployment in industrial separations and carbon capture applications.

Read more at Business Wire

Constraint Capital #4 - Apr 2026

David Rogers — Mon, 06 Apr 2026 13:03:35 GMT

In the history of industrial revolutions, we didn’t just invent the engine; we built the factory. March 2026 has officially marked that same transition for the digital age. We have moved past the era of singular crafted AI models and entered the era of the AI Factories.

Demand for intelligence has effectively outstripped our ability to house it. In this fourth edition of Constraint Capital, we set the stage of the dual seismic events of this month: NVIDIA’s GTC 2026 /YouTube/ and the joint SpaceX/Tesla TERAFAB announcement /SpaceX on X/.

As NVIDIA CEO Jensen Huang recently declared, “I love constraints... because in a world of constraint, you have no choice but to choose the best.” /Tom’s Hardware/

If 2025 was about the ‘what’ of AI, 2026 is about the ‘where’ and ‘with what.’ We are mapping a future where the bottlenecks are no longer just in the code, but in the very atoms required to stand up a factory.

Priority 1: The “Triple Wall” (Land, Power, and Shell)

Huang identifies these as the ultimate gatekeepers. When land, power, and shell are constrained, an AI firm cannot afford to experiment with arbitrary hardware.

The Flight to Quality: Because you cannot easily expand a 1-gigawatt footprint, you must populate it with the highest-performing silicon to maximize revenue-per-watt. As Huang notes, “If I choose poorly, my revenues are affected... they can’t choose poorly.”
The TERAFAB Pivot: Musk’s move to orbital compute is the ultimate admission that the terrestrial “Triple Wall” has become impenetrable. 80% of Terafab’s output is heading for space specifically to bypass the land and power constraints of Earth.

Priority 2: The “Atoms” (Wafers, CoWoS, and HBM)

The physical production of chips remains the most immediate bottleneck. NVIDIA has moved from a just-in-time to a “total-security” model.

Secured Supply: Huang has confirmed NVIDIA has “all the memories (HBM), all the wafers, and all the CoWoS (packaging).” By securing the entire DRAM and TSMC production lines for years in advance, NVIDIA has turned a component shortage into a competitive moat.

Money is the most obvious challenge for Elon Musk’s chip venture. To build 1 TW of AI silicon per year, Elon Musk’s TeraFab will need to process the equivalent of 22.4 million Rubin Ultra GPU wafers per year, 2.716 million Vera CPU wafers per year, and 15.824 million HBM4E wafers per year, according to estimates from premier semiconductor analysis firm Bernstein. To do so, TeraFab will need from 142 to 358 fabs, the report claims.

From “Analyzing Elon Musk’s TeraFab — A step towards Tesla and SpaceX’s partial vertical integration, or an unattainable dream?” /Tom’s Hardware/

The 2% Problem: Musk’s TERAFAB is a response to this exact constraint. If NVIDIA has secured the atoms, Tesla/SpaceX have no choice but to build their own recursive loop fab to bypass the global supply ceiling.

Priority 3: The “Unseen” Essentials (Copper, MLCCs, and Cables)

The most overlooked constraints are the passive components that hold the factory together.

Everything from Copper to Capacitors: Huang highlighted that NVIDIA has secured everything down to multilayer ceramic capacitors (MLCCs), connectors, and cables.
The Infrastructure Burden: The transition to GTC 2026’s 800V DC power architectures and liquid-to-chip cooling requires a massive influx of specialized copper and thermal components that are currently in a state of fantastic scarcity for those who own the supply.

Among the most important talks at GTC was Jeff Dean and Bill Daly exploring “the critical intersections of hardware innovation, systems scaling, and algorithmic advancement needed to propel AI into the 2026–2030 era of agentic systems, ultra-low-latency reasoning, and energy-efficient scaling.” Rewatch below.

While the TERAFAB announcement sparked speculation about non-conventional AI compute…

So what could this unconventional compute be? Let’s dive in…

[Overcapacity] From Industrial Versailles to Man on the Moon

David Rogers — Sun, 05 Apr 2026 16:02:56 GMT

Overcapacity

Thank you for being a paid subscriber! This week’s bonus board is available until 2026-05-03.

Click to access ⤵️

This weeks digest:

From Industrial Versailles to Man on the Moon

David Rogers — Sun, 05 Apr 2026 13:02:07 GMT

Shop Talk

Capturing this week’s zeitgeist

In 1956, the GM Tech Center was hailed as an “Industrial Versailles” /JSTOR/ a sprawling monument to the belief that the future could be manufactured through sheer precision and architectural grandness. But as we’ve learned over the last 70 years /Assembly/, the most iconic leaps often happen in the frantic, unpolished spaces between the prototype and the finish line.

In this week’s issue, we look at the man on the moon mindset that bridges Detroit’s mid-century dominance with today’s silicon-and-intelligence frontiers. Apple’s own “man on the moon” project the high-stakes was a six-month sprint to replace the iPhone’s plastic screen with glass just weeks before launch.

Flash forward to today, and the legacy of these high-pressure tech centers is alive and well this morning. The four-person crew of Artemis 2 streaks past the halfway point between Earth and the Moon, we see the ultimate result of turning impossible prototypes into operational reality which is what Exponential Industry is all about.

Quote of the Week

And what confirms that the site is very competitive is that in the current market environment and all the headwinds we are facing from what’s happening in the Middle East, still, after a smooth startup of our plants, the plants are running at a very high utilization rate, while our competitors, many of them had to reduce their run rate, some even had to declare force majeure. Given that we are having very good cost positions and we are having a mixed-feed cracker that allows us to shift feedstocks between naphtha and butanes, we can run full.
Stephan Kothrade, CTO BASF /Yicai/

Failed AI tractor company, Monarch Tractor, lays off all employees, abandons Bay Area headquarters /SFGATE/

Last week’s answer: Agency to Build Thinking Factories

Assembly Line

This week’s Industry 5.0 breakthroughs and frontier technologies of the built world.

🚀 Artemis II Launches Astronauts to the Moon

Artemis II’s SLS (Space Launch System) rocket lifted off from Launch Pad 39B at NASA’s Kennedy Space Center in Florida, at 6:35 p.m. EDT (2235 UTC) on April 1, 2026, sending NASA astronauts Reid Wiseman, Victor Glover, Christina Koch, and CSA (Canadian Space Agency) astronaut Jeremy Hansen on an approximately 10-day test flight around the Moon and back aboard their Orion spacecraft.

🚀 Liftoff! NASA Launches Astronauts on Historic Artemis Moon Mission /NASA/
🚀💰 SpaceX Has Filed Confidentially for IPO Ahead of AI Rivals /Bloomberg/ seeking a valuation in the IPO of more than $1.75 trillion while raising as much as $75 billion.

🚗 Reimagine marketing at Volkswagen Group with generative AI

✍️ Author: Liam Byrne, Kim Robins, Philipp Hinderberger, Sebastian Angersbach, Philip Trempler, and Weiran Zhang

The first step in Volkswagen’s generative AI journey was deceptively simple: could foundation models (FMs) generate photorealistic images of their vehicles? Initial experiments with base diffusion models revealed two critical gaps. First, while these models could produce impressive automotive imagery, they lacked decades of Volkswagen design language. The tiniest features matter: the exact texture of a grille mesh, the precise geometry of headlight housings, the specific wheel spoke patterns for each model line. The models would generate a Volkswagen, but with generic wheels and grille patterns that didn’t match an actual model year. Second, base models had no knowledge of unreleased vehicles. They couldn’t generate images of next year’s models still under development, severely limiting their utility for forward-looking marketing campaigns.

The solution required fine-tuning foundation models on Volkswagen’s proprietary visual assets. Working with SolidMeta, the team used DreamBooth fine-tuning techniques with training data collected from digital twins in NVIDIA Omniverse. The following diagram illustrates this process for the Volkswagen Tiguan. DreamBooth training works in two parts: first, the model learns from VW Tiguan images paired with a unique identifier token ‘VW Tiguan’ that teaches it this specific vehicle. Second, the model trains on generic car images to preserve its general capabilities and help prevent overfitting to the training set.

Autonomous CNC Programming

Tanmay Aggarwal, CEO of Lambda Function, in a deep conversation about autonomous CNC programming, real-time shop floor integration, and why memory and causal inference matter for the next wave of physical AI.

🎨🦾 A New Way to Spray Paint Color

✍️ Author: Sandesh Manik, a mechatronics engineer and the founder of a stealth startup focused on 3D modeling for hardware.

I wanted my solution to be inexpensive and simple enough to build as a DIY project and functional enough for an artist to use, without breaking their flow. So I began prototyping a system that combines base colors while they are still in pressurized form from off-the-shelf cans.
Even though the system worked, the solenoid valves I used constantly clogged up. Designed for water purifiers, the valves didn’t prevent paint from entering the mechanism, where the paint would harden. Moreover, when the valves were turned off, they could stop backflow only if the inlet remained pressurized. So disconnecting a paint can from the system would cause instant leaking. Other off-the-shelf valves I tried couldn’t cycle fast enough and were too expensive.

Performant rotary pinch valve design

So I created my own mechanism: a high-speed, electronically controlled, rotary pinch valve. It has a stepper motor that rotates a lever with a rolling bearing to constrict fluid flow inside a flexible tube. This concept isn’t new—there’s something like them in every peristaltic pump. But I added a spring to firmly hold the lever in the closed position against any back pressure when the motor isn’t powered, making it a normally closed valve that isolates the attached can. Additionally, the valve is fast enough to be open for as little as 30 milliseconds.

🚗🧠🔧 Can AI Predict Loose Bolts in Automotive Assemblies?

Standardized test methods, such as DIN 65151 and DIN 25201, are effective in terms of performance evaluation for single bolt and nut types. However, the rigid conditions in those tests are not necessarily applicable to bolted joints in automobile chassis. The conditions are different, and the external loads acting on the joints have variable frequency.

Our approach consisted of three steps. In the first step, we collected external loading values from mechanical testing of engine suspension joints. In the second step, we performed Taguchi method experiments with actual joint conditions to obtain training and test data for the neural network. In the third step, we created a neural network with the Levenberg-Marquardt (LM) and Bayesian Regularization (BR) algorithms to create a relationship between the controlled factors and the loosening rate. We then collected the mean squared error (MSE) values to evaluate the prediction errors of the neural network.

Read more at Assembly and Machines

New Product Introduction

Highlighting new and innovative facilities, processes, products, and services

🏭📊🧠 HighByte Brings AI to Industrial Data Pipelines with Version 4.4: Federated Namespaces and AI-Driven Pipelines

✍️ Author: Jeffrey Schroeder

HighByte Intelligence Hub version 4.4 is built for exactly that environment. This release introduces Central Data for composing federated namespaces and pipelines with remote data sources, new connectivity options including native Databricks Zerobus support and the CESMII i3X Server interface, and a Pipeline AI Agent that makes pipeline configuration more accessible across teams and skill levels. Taken together, they extend what Industrial DataOps teams can do and who can do it.

Read more at Business Wire and HighByte

🏭📊 Unified Namespace, Revisited: Toward a Composable Industrial Architecture /Solace/
🏭📊🧠 HP Indigo boosts manufacturing yield (from 60% to 92%) and traceability with Databricks and Unity Catalog /Databricks/

📶📐🧠 Introducing Atlas RF Studio: Toward a Foundation Model for Electromagnetics

✍️ Author: Christopher Bryant, Tommaso Dreossi, Hao Liu, Nathan Mirman, Noah Kessler, Usman Farman, Daniel Pedraza, Trevor Beaton, Akshay Kumar, Ben Gelb, Mike Frei, Arun Natarajan, Harish Krishnaswamy

Atlas RF Studio is powered by two models that mirror the two fundamental problems of electromagnetic engineering. Both develop an understanding of how geometry and electromagnetic behavior relate, but they apply this understanding in opposite directions.
We call the forward model Heaviside-0, after Oliver Heaviside, who reformulated Maxwell’s equations into the compact vector form that made modern electromagnetic analysis possible. Heaviside was the great characterizer. He gave engineers the mathematical language to describe what fields do. Our forward model carries that legacy. Given a circuit layout, it predicts S-parameters in milliseconds rather than the significant compute time required by a numerical solver. This is what makes rapid design iteration possible.
We call the inverse model Marconi-0, after Guglielmo Marconi, who invented and built the first practical radio systems through relentless experimentation, turning desired wireless behavior into physical reality. Our inverse model does the same thing. Given a target S-parameter response, it generates a physical geometry that produces it. This is the creative half, the one that proposes designs a human might never think to try.
These two models work in tandem. Marconi-0 proposes a collection of designs that it thinks will produce your target behavior, and Heaviside-0 quickly checks whether the proposed designs exhibit that behavior. Based on that feedback, Marconi-0 can then iteratively propose new designs that are more likely to produce the desired behavior, and Heaviside-0 can then characterize them again to see if they are any better.

Business Transactions

This week’s top funding events, acquisitions, and partnerships across industrial value chains.

🇺🇸 Saronic Closes $1.75B Series D at $9.25B Valuation to Accelerate a New Era of Maritime Autonomy

Saronic Technologies announced it has closed $1.75 billion in Series D funding to advance its mission of ensuring maritime superiority for the U.S. and its allies by delivering autonomous platforms at scale across defense and commercial sectors. The round was led by Kleiner Perkins and values the company at $9.25 billion.

An important component to Saronic’s efforts is the development of Port Alpha — its next generation shipyard — alongside the rapid expansion of its existing production facilities in Louisiana and Texas. This capital will allow Saronic to accelerate the growth and maturation of its shipbuilding enterprise, one that is defined by innovation, scalability, and strategic impact.

Read more at PR Newswire

🇺🇸 iBase-t Announces Strategic Growth Investment from TA to Accelerate AI-Powered Digital Manufacturing and MRO Transformation in Aerospace & Defense /Business Wire/

🇺🇸🛰️ Starcloud Raises $170M Series A at $1.1bn Valuation Led by Benchmark and EQT Ventures

Starcloud, the company building data centers in space, announced it has raised a $170 million Series A, at a $1.1 billion valuation. Achieving unicorn status just 17 months after its Y Combinator demo day, Starcloud is now the fastest unicorn in Y Combinator history. The round is also more than double the size of the next largest YC Series A, and brings the company’s total capital raised to $200 million.

Starcloud has demonstrated a pace of execution previously unseen in the aerospace or tech industries. With just $3 million in pre-seed funding, the company designed, built, and launched its first satellite, Starcloud-1, in a record 21 months. Launched in November 2025, the mission achieved several historic industry firsts:

First NVIDIA H100 in orbit: Successfully deployed the most powerful GPU in space, delivering a 100x increase in AI compute.
First orbital AI training: Successfully trained an AI model in space.
First orbital inference on Gemini: Successfully ran a version of Gemini in orbit.
First orbital fine-tuning: Demonstrated high-powered inference and model fine-tuning in space.

Read more at Business Wire

🇪🇺⚡🛰️ TerraSpark advances space-based solar power with €5M funding /Tech.eu/. TerraSpark is developing space-based solar power to deliver continuous energy independent of weather or time of day, starting with wireless transmission on Earth before scaling to orbital systems.

Founders Factory & Rio Tinto invest in reduced-impact mining through AI & Breakthrough Technologies

The six investments are:

Durin, automated drilling rigs to accelerate mineral discovery and reduce exploration costs (Los Angeles, US)
Rock Zero, zero-waste, cost-efficient process to extract lithium from hard rock deposits (Cambridge, US)
Terra AI, AI for mineral exploration to reduce drilling costs and improve accuracy (Palo Alto, US)
Ekion, electrokinetic in-situ recovery (EK-ISR) technology to extract metals from stranded deposits with minimal waste and emissions (Perth, Australia)
Rainstick, agriculture & nature restoration using electrical fields to mimic the natural effects of lightning, boosting seed yields faster and more sustainably (Cairns, Australia)
Thunderstone, reduced-impact mining using underground electric stimulation to extract critical metals with minimal waste and cost (US)

What Could Bezos Buy?

David Rogers — Wed, 01 Apr 2026 13:03:49 GMT

Back on February 27th, the Financial Times broke the news on Jeff Bezos’ Project Prometheus /FT/ and its associated ~~$30~~ $100 billion /WSJ/ "manufacturing transformation vehicle." It appears Jeff Bezos and Vikram Bajaj are targeting a specific profile of business that blends traditional heavy industry with high-complexity engineering.

Exponential Industry first started tracking the Project in the November 23rd digest.

The Project strategy seems to not just to invest in AI, but to buy legacy companies that are currently being devalued or disrupted by AI, then use proprietary World Model /NVIDIA/ technology to radically improve their margins and speed.

Last week, following the Wall Street Journal update, the gentleman at TBPN proposed five options: Lear, BorgWarner, Hexcel, Goodyear, and Rockwell Automation /TBPN/ with a caveat:

This is all mostly a wild guess at this point, the fact that Bezos could potentially link to this Project Prometheus means there’s a chance this whole project will be much more aligned to the traditional software singularity path, and there’s always a chance we get an Elon-style megacorp with Blue Origin involved at some point. Either way, it’s a big move from an experienced operator with lots of opportunity ahead. Will be fun to follow along.

Now that speculation is abound, I figured its time to outline the 'target profile' and raise several companies to share with you and see how well they fit the criteria.

My speculation began back on March 4th.

[Overcapacity] Agency to Build Thinking Factories

David Rogers — Sun, 29 Mar 2026 16:02:52 GMT

Thank you for being a paid subscriber! This week’s bonus board is available until 2026-04-26.

Click to access ⤵️

Connect with Allie Systems on LinkedIn for an invite

This weeks digest:

Agency to Build Thinking Factories

David Rogers — Sun, 29 Mar 2026 13:03:31 GMT

Shop Talk

Capturing this week’s zeitgeist

I am excited to continue my collaboration with Allie Systems with a virtual round table on "Fact vs. Fiction: What It Really Takes to Build a Factory That Thinks”, this Tuesday, March 31 at 11 AM CST. The session is designed to get past the hype cycle and share what's working on the floor. We will dig into the data foundation problem, change management, use case mapping, and what "autonomous operations" actually looks like in practice.

Quote of the Week

Build samples. Test them. Analyze deeply why things fail. Build tests that catch those failures. Correlate your module-level tests with system-level FATP tests. Go back to the process, tune it. Build more samples. Repeat.
This is how you observe the “reinforcement learning” of manufacturing. Not by reading about it — by running the loop yourself, staring at failure pareto charts, arguing with the process engineer about whether the epoxy dispense volume is drifting, and watching the yield numbers tick up over weeks as you close the gaps.
Read more at “The Dark Matter of Hardware Engineering” /Dark Matter/

🧪📊🧠 53 Years of Trade Secrets, Locked in Software That Will Never See an AI /Dan Thompson on X/

Last week’s answer: The Automation High Point Story ends as a new era takes over.

This is the written version of a talk I gave at the OSU HAMMER Engineering Research Center in February 2026…

Assembly Line

This week’s Industry 5.0 breakthroughs and frontier technologies of the built world.

👕♻️ How Renasens Cracked Textile Recycling And Raised $11.5 Million

✍️ Author: Brooke Roberts-Islam

The engineer’s technology uses supercritical carbon dioxide (a fluid) as the solvent (dissolving solution) rather than water or harsh chemicals, thereby reducing water and energy use and the overall footprint of the recycling process. While the use of supercritical CO₂ is not novel, the application for textile recycling is.

🎛️🖥️ What Happens When Manufacturers Build Their Own MES Applications in Two Days

✍️ Author: Jenna Gabriel

Production Lab is a two-day, hands-on building event where MachineMetrics customers work side by side with our engineering team to design and build custom applications on the MachineMetrics platform. The tools they used: Max AI for data exploration and prompt development, Lovable and Cursor for AI-assisted application building, and Carbide, MachineMetrics’ custom application builder, for deploying production-grade applications directly into the platform. Every team shipped a working solution tied to real, live machine data by the end of day two.

The execution gap is real. It lives between what your ERP plans and what your shop floor actually does, and it costs manufacturers margin, capacity, and hours every single shift. Closing it requires software that fits how your operation actually runs, not a generic template you adapt yourself.

📐🧠 Engineering intent to compiled into engineering artifacts

In our setup, a textual description of a production environment and the objects to be placed is enough to generate USDA files, the native scene description format used by NVIDIA Omniverse to describe 3D scenes, assets, transformations, materials, hierarchies and layouts.
The workflow looks like this:
An engineer describes what they want to visualize.
Claude (Opus 4.6) compiles this description into machine-readable USDA files.
Omniverse loads the scene and the factory layout appears.
What sounds simple is actually a big shift. In the future, engineers will no longer interact directly with engineering tools.

Read more at Dr. Dirk Alexander Molitor on LinkedIn

📐🧠 Continuous Design Reviews with AI, in addition to manual human reviews is the future. /Dalus on X/
📐 The Myth of CAD as Code /emm0sh on X/
📐🧠 Synopsys Brings Agentic Engineering Into Focus as AI-Driven Design Tests Productivity Limits /EETimes/

Spectech Newsletter

Fungible Factories

20 days ago · 24 likes · 1 comment · Ben Reinhardt

New Product Introduction

Highlighting new and innovative facilities, processes, products, and services

🏆 The most innovative manufacturing companies for 2026

The most consequential changes are unfolding across materials and components that form the foundation of industrial supply. Hertha Metals and Electra are disrupting the carbon-intensive economic model of steel and iron production with electrified, cost-effective methods that can be integrated into existing infrastructure. Abstrax is stabilizing climate-volatile agricultural inputs with precision-engineered flavor systems that enable beverage makers to maintain product consistency despite climate variability. Likewise, TerraSlate is disrupting coated and laminated materials with a recyclable waterproof paper that can withstand industrial use, without relying on plastic.
At the factory level, Acerta AI and Sight Machine are embedding diagnostics and analysis directly into live operations so defects can be traced and corrected in minutes instead of days. Freeform’s high-throughput metal printing architecture and Cellares’s automated cell therapy facilities showcase that machine learning and robotics can maintain tight tolerances while dramatically increasing throughput.

💿 Arm expands compute platform to silicon products in historic company first

Arm Holdings plc (NASDAQ: ARM) today announced the next evolution of the Arm compute platform, extending into production silicon products for the first time in the company’s history. This begins with the launch of the Arm AGI CPU, an Arm-designed CPU for AI data centers, built to address a rising class of agentic AI workloads.

For more than three decades, the industry has innovated on the Arm compute platform to deliver scalable, power-efficient computing across hundreds of billions of devices. As AI transforms global computing infrastructure, partners across the ecosystem are asking for ways to deploy Arm technology at scale. In response, Arm is expanding its platform strategy beyond IP and Compute Subsystems (CSS) to include Arm-designed silicon products – giving partners the broadest set of options to build on Arm and enabling faster innovation across the AI ecosystem.

🛢️🧠 Aker BP, Armada to deploy offshore modular data center for AI-driven drilling operations

Aker BP and Armada have entered into an agreement to deploy Armada’s Galleon modular data center for use in an offshore drilling environment on the Norwegian Continental Shelf. Under the agreement, Armada will deliver a Galleon designed to operate in offshore conditions, enabling local processing and analysis of drilling and operational data directly on the rig.

By deploying compute infrastructure directly offshore, the Galleon strengthens operational resilience and standardizes adoption and use of critical vendor applications. Running AI models locally allows energy companies to predict and prevent equipment failures before they occur, reduce unplanned downtime, and maintain continuity during connectivity disruptions. Local compute infrastructure also enhances cybersecurity by minimizing reliance on external networks.

Read more at PR Newswire

🛢️🧠 Onshore Well Exploration & Intervention Demo /Cognite/
🛢️🧠 Cognite and NVIDIA Operationalize NV-Tesseract to Transform Industrial Forecasting /Business Wire/

“Despite heightened market volatility amid the ongoing Middle East turmoil, bankers and analysts say capital markets remain open, particularly for oil ⁠and gas issuers” /Reuters/ as HMH Holding seeks $948 million valuation in US IPO.

Business Transactions

This week’s top funding events, acquisitions, and partnerships across industrial value chains.

🇺🇸🦾🧠 Physical Intelligence in Talks to Raise $1 Billion at $11 Billion Valuation

✍️ Authors: Natasha Mascarenhas, Rebecca Torrence, and Shirin Ghaffary

Physical Intelligence, a two-year-old robotics startup founded by AI academics and former Google DeepMind researchers, is discussing a new funding round of about $1 billion that would bring the company’s valuation to more than $11 billion including dollars raised, according to people familiar with the matter.
The ultimate goal, the company has said, is building models that enable robots to perform any function requested by the user through “vision-language models,” in the same way that large-language models complete digital tasks on command. In recent weeks, the company has released new research about improving memory in robotic software models as well as performance on precise tasks that require careful and detailed manipulation of objects.

🇺🇸🖥️🎛️ DOSS raises $55M to Build the Operations Cloud for the Real World

DOSS raised a $55M Series B, co-led by Madrona and Premji Invest, with Greyhound, Commerce Ventures, and Intuit Ventures as new investors, and continued investment from Theory, General Catalyst, Contrary Capital, Mintaka, and Pathlight VC, and 47th Street Partners.

This raise enables them to: invest deeply in core platform and unified data model, so companies can go live faster with intelligent systems that adapt as fast as their business grows, develop agentic systems that remove the human bottleneck from configuring operations, further define a new category around adaptable operations, not static systems

🏴󠁧󠁢󠁷󠁬󠁳󠁿⚗️ Biomanufacturing Pioneer Cauldron Raises $13.25M

Cauldron Ferm, a leader in advanced biomanufacturing, announced a $13.25 million Series A2 priced round. The round was led by Main Sequence Ventures, with participation from Horizons Ventures, SOSV, and NGS Super, bringing Cauldron’s total funding to $26 million.

Cauldron’s develops a continuous “hyper-fermentation” platform – a proprietary technology designed to deliver step-change improvements in bioprocess efficiency. Biomanufacturing uses living cells as programmable factories, where bio-engineered microbes convert simple inputs like sugars into valuable products. Cauldron’s key advantage is keeping these microbes in a highly productive steady state and running processes continuously for long periods. This increases output while significantly lowering production costs, helping bioproducts compete with conventional industrial alternatives.

Read more at Business Wire

🇺🇸 Helix Earth Secures $12 Million in Oversubscribed Seed 2 Funding to Revolutionize Energy Efficiency and Humidity Management in Commercial HVAC Systems

Helix Earth, a Houston-based hardware company built on proprietary liquid-gas chemistry with roots in NASA aerospace technology, announced today the successful close of its $12 million oversubscribed Seed 2 funding. The oversubscribed financing round was led by Veriten, an energy research, strategy, and investing firm, with participation from Rua Ventures, Carnrite Ventures, Skywriter LLC, Textbook Ventures, and others.

Helix Earth is developing a one-of-a-kind, cost-effective commercial air conditioner add-on that bifurcates the energy required for temperature management from moisture control. The result is healthier indoor air, lower energy bills, reduced building maintenance, and more comfortable spaces for building owners and occupants.

Read more at Business Wire

Looking for additional content? Paid subscribers get more stories delivered each week as part of Overcapacity!

[Overcapacity] The Automation High Point Story

David Rogers — Sun, 22 Mar 2026 16:01:52 GMT

Thank you for being a paid subscriber! This week’s bonus board is available until 2026-04-19.

Click to access ⤵️

This weeks digest:

The Automation High Point Story

David Rogers — Sun, 22 Mar 2026 13:01:59 GMT

Shop Talk

Capturing this week’s zeitgeist

According to a research tool by Andrej Karpathy /Karpathy's Blog/ industrial engineering technologists face a moderate 6/10 AI exposure because their data-driven CAD and cost-analysis tasks are perceived as highly automatable, yet protected by the physical buffer of on-site floor inspections. In contrast, manual manufacturing roles like assembly and welding maintain low 2/10 to 3/10 exposure scores due to the persistent difficulty of replicating human dexterity in unstructured environments. Broadly, the analysis reveals a paradox where higher-paying, education-intensive roles are significantly more vulnerable to digital AI than the foundational labor powering the factory floor.

What do you find interesting about this analysis?

US Job Market Visualizer with Digital AI Exposure /GitHub/

Last week’s answer: The Catalina Wine Mixer is a scene in the movie Step Brothers and inspired an annual event by the Catalina Island Company.

Assembly Line

This week’s Industry 5.0 breakthroughs and frontier technologies of the built world.

🏭🦾 Inside Rockwell Automation’s Factory Where Automation Is Built

✍️ Author: David Petersen

The assembly line, fabricated in large part by Calvary Robotics, incorporates the iTRAK modular conveyance system, along with plenty of Rockwell controllers, drives, and I/O, to progress the product along the assembly stages.
The assembly stations themselves are nearly all operated robotically. Depending on the needs, the robots are gantry, SCARA, articulated, or delta styles, some floor-mounted and others mounted above the line. Specialized grippers pick each piece as it builds the contactors according to the order specifications.

📐 Engineering journey of a lightweight drone rotor blade with tape‑sandwich injection moulding

Developing a next generation drone rotor blade calls for new material concepts, accurate process control and smart manufacturing strategies. The objective behind this project was to design a blade that combines very low weight with strong mechanical robustness in demanding flight conditions.

Together with specialists in composites, simulation and moulding technology, a fully automated tape sandwich injection moulding workflow was created. Continuous carbon fibre thermoplastic tapes form the outer structure and are positioned in the mould with high accuracy. The centre of the component is then formed by injecting a short fibre reinforced compound that expands slightly during processing.

🦾 Rapid Operator AI - Hardware, Vision, and Physical AI in One Solution /Vention on YouTube/
Arburg Reveals Production Plans, New Technologies and Results /Plastics Technology/

🧠 Unlocking Industrial Knowledge with AI Agents

We sit down with Jakub from Zeta Alpha to uncover how companies like Festo are leveraging advanced search, RAG, and agentic AI to unlock decades of expertise buried in documents and data silos. We share practical examples, discuss integration challenges, and debate the future of specialized versus generic AI in the industrial landscape.

📑 Factory data could unlock £129bn in UK output report from FourJaw /The Engineer/

🚛🧠 Inside Kodiak AI and the Future of Driverless Freight with Founder and CEO Don Burnette

🚛🧠🤝 Nissan, Uber, Wayve unveil robotaxi tie-up /Reuters/
🚛🧠🤝 Uber and Rivian Partner to Deploy up to 50,000 Fully Autonomous Robotaxis /Rivian/

Should Robot Generalists Get Off Their High Horse?

A conversation between @AnneliesGamble and @Ken_Goldberg, professor of Robotics and Automation at UC Berkeley

The assumption has been: just collect data on everything out there and throw it all together, and it will suddenly start to become general. That has not panned out, and that’s not what customers are looking for.
Even if generalist systems improve, most customers aren’t looking for a robot that might eventually do many things. Many want a robot that can at least do one thing really well, over and over again. “Most successes in robotics require 99.99% uptime,” @samanfarid, CEO of Formic, told me. “Most people underestimate how hard that bar is to hit.” Not to mention, real deployments are also incredibly complex.

📐💿 Design Conductor: An agent autonomously builds a 1.5 GHz Linux-capable RISC-V CPU

Design Conductor (DC) is an autonomous agent which applies the capabilities of frontier models to build semiconductors end-to-end -- that is, from concept to verified, tape-out ready GDSII (layout CAD file). In 12 hours and fully autonomously, DC was able to build several micro-architecture variations of a complete RISC-V CPU (which we dub VerCore) that meet timing at 1.48 GHz (rv32i-zmmul; using the ASAP7 PDK), starting from a 219-word requirements document. The VerCore achieves a CoreMark score of 3261. For historical context, this is roughly equivalent to an Intel Celeron SU2300 from mid-2011 (which ran at 1.2 GHz). To our knowledge, this is the first time an autonomous agent has built a complete, working CPU from spec to GDSII. This report is organized as follows. We first review DC’s design and its key components. We then describe the methodology that DC followed to build VerCore -- including RTL implementation, testbench implementation, frontend debugging, optimization to achieve timing closure, and interacting with backend tools. We review the key characteristics of the resulting VerCore. Finally, we highlight how frontier models could improve to better enable this application, and our lessons learned as to how chips will be built in the future enabled by the capabilities of systems like DC.

New Product Introduction

Highlighting new and innovative facilities, processes, products, and services

🦾🧠 NVIDIA and Global Robotics Leaders Take Physical AI to the Real World

NVIDIA is partnering with the global robotics ecosystem — including leading robot brain developers, industrial robot giants and humanoid pioneers — to power production-scale physical AI. NVIDIA also unveiled new NVIDIA Isaac™ simulation frameworks and new NVIDIA Cosmos™ and NVIDIA Isaac GR00T open models for the industry to develop, train and deploy the next generation of intelligent robots.

In addition, during his GTC keynote, Huang previewed GR00T N2, a next-generation robot foundation model based on DreamZero research. Built on a new world action model architecture, the model helps robots succeed at new tasks in new environments more than twice as often as leading vision language action models. Slated to be available by the end of the year, GR00T N2 currently ranks No. 1 on MolmoSpaces and RoboArena for generalist robot policies.

Business Transactions

This week’s top funding events, acquisitions, and partnerships across industrial value chains.

💰 Partech’s €300M Impact Fund targets Europe’s next generation of industrial and climate tech leaders /Tech.eu/

🇳🇿🐄 Halter AI Cow Collar Startup reaches $2 Billion Valuation

Billionaire Peter Thiel’s venture capital firm, Founders Fund, is set to lead the financing, said the people, who asked not to be identified discussing private information. The talks are ongoing and details could change.
Halter’s technology is designed to create a virtual fence for cattle and enable farmers to monitor the animals’ locations and health indicators through an app. Its collars, which are solar-powered, connect to farmers’ phones to allow them to manage pastures remotely — for example, a rancher can herd their cows using vibrations and audio cues from the collars.

🇬🇧 Cocoon Secures $15M Series A to Fix A Critical Supply-Demand Gap in the Concrete Industry

Cocoon Carbon, a company unlocking a new, scalable supply of key materials used to make concrete, announced the closing of a $15 million Series A round. The investment was co-led by 2150 and Brick & Mortar Ventures, investors in the built environment and construction sectors. The round included participation from TVC (The Venture Collective) and continued support from existing investors Wireframe Ventures, Celsius Industries, Gigascale Capital, and SOSV.

Cocoon unlocks a new source of SCMs by converting an underutilized and robust supply of steel slag, a byproduct of electric arc furnaces (EAFs), into a low-cost, high-performance cement replacement. EAFs re-melt scrap using electricity, compared to coal powered blast furnaces that turn iron ore into pig iron. With tens of millions of tons of steel slag produced annually in the U.S. and Europe, and EAF steelmaking projected to double by 2050, Cocoon expands the domestic supply base concrete producers rely on to keep costs stable. Cocoon’s new product matches the performance of traditional SCMs while reducing the embodied CO2 of concrete by up to 40%.

Read more at Business Wire

🇺🇸 Rubi Raises $7.5M and Secures Over $60M in Offtake Term Sheets, Advancing Commercialization of Its Breakthrough CO2-to-Materials Platform /PR Newswire/
🇩🇪 Level Nine raises €4M to unlock local feedstocks for chemicals /Tech.eu/ by developing advanced catalysts to enable sustainable chemical production, converting biomass and waste into renewable materials while improving efficiency and supporting industrial-scale adoption.
🇺🇸 V-Glass Secures $3M Investment to Initiate Move from Lab Scale Production into a Pilot Plant /Business Wire/. V-Glass has developed next generation platform technology enabling the low-cost production of vacuum insulated glass (VIG) having 2-3x the thermal performance of ENERGY STAR® windows, lasting twice as long, and costing no more to manufacture.

🇺🇸 Tenkara Raises $7M led by True Ventures to Build Ops Agents for US Manufacturers

Tenkara, the company building ops agents for US manufacturers, announced $7 million in funding led by True Ventures. The round includes participation from HF0, WndrCo, Articulate Capital, Night Capital, SF1, Transpose, early employees at Flexport, and others.

Tenkara is building the next-generation tooling for supply chains. The company’s agents take on the back-office burden so that lean teams can operate like organizations 10 times their size. In its first 18 months, Tenkara signed multiple seven-figure contracts with U.S. manufacturers.

Read more at PR Newswire

🇩🇪 Fibionic secures €3 million seed funding for bionic fiber placement

✍️ Author: Grace Stubbins

Fibionic (Götzens, Austria), a deep-tech startup founded in 2021, has announced the successful closing of a €3 million seed financing round led by co-lead investors Redstone with Euregio+, Caesar and Leap435, and supported by various angel investors. This financing will enable the lightweight engineering company to further industrialize its nature-inspired technology for precise fiber placement and drive market entry into new industries.
Fibionic addresses material and automation inefficiencies via fibionic fiber placement (FFP). The patented, sustainable process uses bionic principles to align fibers specifically along actual load paths, inspired by structures found in nature, such as the wings of a dragonfly. FFP achieves cycle times of up to 1 minute per component, offering optimization potential for composites manufacturing — high-performance, stable components with minimal material use.

[Overcapacity] Catalyna Wine Mixture

David Rogers — Sun, 15 Mar 2026 16:02:58 GMT

Thank you for being a paid subscriber! This week’s bonus board is available until 2026-04-12.

Editor’s note: Overcapacity will be sent at noon Eastern US time each Sunday moving forward.

Click to access ⤵️

This week’s digest:

Catalyna Wine Mixture

David Rogers — Sun, 15 Mar 2026 13:02:43 GMT

Shop Talk

Capturing this week’s zeitgeist

The era of the AI assistant has replaced by a paradigm where “Everything is ~~Computer~~ Code”, a shift led by tools like Claude Code (Coding) and Genie Code (Data Work) that move us from simply writing syntax to delegating entire autonomous lifecycles.

Who will make “Foundry Code”?

This product will autonomously refactor assembly workflows and optimize thermal cycles in real-time, treating a physical production floor with the same agility and version control we currently apply to coding and data work. Some features could include:

Software-Defined Hardware: Acts as a real-time compiler that translates high-level operational goals into optimized machine instructions for PLCs and CNCs.
The “Edit Button” for Reality: Synchronizes digital twins with physical lines to autonomously refactor and deploy updated G-code paths without stopping production.
Agentic Maintenance: Proactive background agents diagnose mechanical anomalies, source parts via ERP, and schedule repairs before failures occur.
Batch-of-One Orchestration: Dynamically rewrites machine recipes for every individual unit to compensate for raw material variances in real-time.

Quote of the Week

Software and automation are becoming increasingly central as machine capabilities improve. Chen outlined three areas where artificial intelligence is being applied across Creality’s platforms: model creation and visualisation, fault detection during printing, and optimisation of print parameters such as supports and toolpaths. “AI can help us to do better 3D modeling,” he said, adding that it is also used for “AI-based detection of faults, of failures”.
Jack Chen, Founder and CEO, Creality /3DPI/

EOS CEO Marie Langer on Next Frontier for Industrial 3D Printing /3DPI/

Last week’s answer: Appetite to Touch Grass and Build Donuts

Assembly Line

This week’s Industry 5.0 breakthroughs and frontier technologies of the built world.

🦾 Ed Mehr on Transforming Manufacturing at Machina Labs

Discover how Machina Labs is transforming manufacturing with flexible, robotic-powered factories that can switch products instantly. In this episode, Ed Mehr shares insights on software-defined factories, advanced automation, and strategic partnerships shaping the future of hardware production.

Machina Labs Funding: $124M Series C to Build the Factory of the Future /LA Times/
Isembard Raises $50m Series A to Open 25 AI-Powered Factories Serving Aerospace and Defence /Isembard/

🏭📊 This open-source bet is paying off as United Manufacturing Hub takes on industrial giants

✍️ Author: Cate Lawrence

United Manufacturing Hub Systems (UMH) is building an open-source data platform designed to solve exactly that challenge. The company now counts manufacturing, food and beverage, and top-20 automotive suppliers among its customers, supporting deployments across more than 150 sites globally.
The platform supports two interaction modes. Developers can configure deployments through YAML files — something Hebborn says AI tools can generate quickly when connecting hundreds of machines. For non-technical users, UMH also offers a visual drag-and-drop interface, which becomes important when deployments scale across dozens of sites.
UMH’s next phase is about scaling through team expansion. Geographically, the team is not aggressively pushing international expansion yet as the DACH region already has a huge concentration of global manufacturing companies.

🍷🔋♻️ Wine-making waste helps recycle cobalt and nickel from batteries

✍️ Author: Prachi Patel

Cobalt and nickel, bound together in lithium-ion battery cathodes, are difficult to separate during battery recycling. Researchers have now found a way to tease the elements apart electrochemically with the help of tartaric acid, which is found in grapes and is a wine-industry by-product.
Electrolytic methods offer a sustainable alternative, especially if powered by renewable electricity. One such technique, electrowinning, involves reducing metal ions in an electrolyte solution so they deposit as solid metal on an electrode. It’s commonly used in metal mining. Researchers are now putting it to use for recovering critical metals from electronic waste.

🦾🤖🦿🇨🇳 Inside the AI labs training China’s humanoid robots

China is rapidly expanding a network of state-funded robot training farms where human instructors manually pilot machines to generate the massive datasets needed to develop embodied intelligence. Supported by national policy and billions in investment, these facilities aim to overcome the global shortage of robot-specific data by recording and labeling thousands of physical repetitions of everyday tasks. Although this push provides an early market for Chinese robot manufacturers, significant technical hurdles remain, most notably the difficulty of transferring training data between rapidly evolving and non-standardized hardware models.

🌬️ Adaptive human–robot collaboration in wind turbine manufacturing using digital twins

✍️ Authors: Ali Ahmad Malik, Tariq Masood

The push for higher wind turbine-rated capacity has spurred the development of larger generators, extended blade lengths, and taller towers. Wind turbines with capacities of up to 16 megawatts are available in the market, reflecting an almost 60% surge in design capacity over the past five years. However, due to frequent design changes and the diverse range of tasks involved, conventional automation methods are less practical, leading to a labor-intensive process. Handling and assembling these large components pose challenges to human capabilities. To address these challenges, this study proposes integrating collaborative robots (cobots) to develop a hybrid approach to automating wind turbine manufacturing. Employing cobots can reduce manufacturing costs, increase production speed, and improve working conditions. The article details the development of a mobile robotic assistant designed to collaborate with human operators during wind turbine assembly, based on a case study from a leading global wind turbine manufacturer. Besides highlighting the areas attractive for collaborative automation, the article’s key contributions are to introduce a framework based on digital twin technology for the design, commissioning, and operation of robots. It also presents a human-robot interface using smartwatches to enable fluid interaction between humans and robots on production floors. The developed system can be scaled to other large-size component manufacturing involving intensive manual effort.

New Product Introduction

Highlighting new and innovative facilities, processes, products, and services

☢️ Building the Future of Energy | Aalo’s Sodium Test Loop Unveiled

Sodium is the coolant of choice for Aalo’s mass-manufactured reactors. It transfers heat far more efficiently than water and operates at atmospheric pressure, which means you don’t need the massive pressure vessels that conventional reactors require. The tradeoff: sodium reacts violently with water and ignites on contact with air. There’s no room for error; but sodium can be managed.

Getting a sodium loop running is a major technical milestone. It validates the pumps, the heating systems, the instrumentation, and the material interfaces that will eventually go into a full-scale reactor. Very few organizations in the world operate hardware like this. See inside the skunkworks lab where our engineering team walks through prep, heats sodium in a glovebox, and fires up the Sodium Test Loop for the first time. You’ll want to see what happens.

🧠 Phi-4-reasoning-vision and the lessons of training a multimodal reasoning model

Phi-4-reasoning-vision-15B is a compact and smart open‑weight multimodal reasoning model that balances reasoning power, efficiency, and training data needs. It is a broadly capable model that allows for natural interaction for a wide array of vision-language tasks and excels at math and science reasoning and understanding user-interfaces.
One of our goals was to train a model that performs well across general vision-language tasks, while excelling at mathematical and scientific reasoning and computer-use scenarios. How to structure datasets for generalizable reasoning remains an open question—particularly because the relationship between data scale and reasoning performance can lead to starkly different design decisions, such as training a single model on a large dataset versus multiple specialized models with targeted post-training.

📐🧠 Introducing Wonder 3D: New text and image to 3D AI models in Autodesk Flow Studio

Flow Studio is designed to help you explore more ideas without locking you into a single result.

Business Transactions

This week’s top funding events, acquisitions, and partnerships across industrial value chains.

🇺🇸🦾 Rivian CEO’s AI-Powered Robotics Startup, Mind Robotics, Raises $500 Million

Mind Robotics, an AI startup founded by Rivian CEO RJ Scaringe, has raised $500 million at a $2 billion valuation to develop robots for factory automation. The company is collaborating with Rivian to train its systems using real-world factory data, focusing on practical manufacturing tasks like component assembly and parts handling. Scaringe aims to have a large number of these robots deployed by the end of the year, emphasizing functional utility in manufacturing over flashy robotic stunts.

🇺🇸🖥️🕸️ Nexthop AI Accelerates Into Hypergrowth With Oversubscribed $500M Series B Funding, Catapulting the Company’s Valuation to $4.2 Billion

Nexthop AI, the leading pioneer of highly efficient AI Networking, announced successful closure of an oversubscribed $500 Million Series B funding round, catapulting the company’s valuation to $4.2 Billion. This round was spearheaded by Lightspeed Venture Partners, with Andreessen Horowitz joining as a major investor, and participation from Altimeter and all existing investors.

Nexthop AI has quickly emerged as a frontrunner in the most advanced AI & Cloud Networking segment. The company delivers both off-the-shelf and highly customized switching solutions built on open source operating systems such as SONiC and FBOSS.

Read more at Business Wire

🇺🇸🖥️🕸️ Eridu Emerges from Stealth with Over $200M in Funding To Break Through the Network Wall and Unlock Faster AI /Business Wire/
🇺🇸🖥️🕸️ Xscape Photonics Announces $37 Million in New Funding, Launches Eight-Wavelength Laser for AI Data Center Networks /Business Wire/
🇺🇸🖥️🕸️ Arycs Establishes Itself as Independent Company with $24M to Enable AI to Scale Everywhere /Business Wire/

🇺🇸🦾 Uber co-founder Kalanick launches Atoms in specialized robotics push

Atoms is focused on specialized industrial robotics designed to automate tasks ‌in the mining, transport and food sectors. Kalanick is betting that task-specific machines are the key to improving industrial productivity. He is expanding and renaming City Storage Systems, the startup he started building after leaving the ride-hailing giant.

[Overcapacity] Appetite to Touch Grass and Build Donuts

David Rogers — Mon, 09 Mar 2026 01:01:21 GMT

Thank you for being a paid subscriber! This week’s bonus board is available until 2026-04-05.

Click to access ⤵️

This weeks digest:

Appetite to Touch Grass and Build Donuts

David Rogers — Sun, 08 Mar 2026 13:01:08 GMT

Shop Talk

Capturing this week’s zeitgeist

In the coming decade, the most dominant enterprises won’t just be those with the largest balance sheets, but those fueled by a workforce with an insatiable appetite for building. The modern builder’s toolkit has evolved with AI and during this exponential rate of progress, more physical world activities like manufacturing will be soon be accessible to AI. Here’s a few already upon us:

Personalized Business Systems: “100% of my code is written by Claude Code”, Boris Cherny of Anthropic /YouTube/. Your business process and systems can be fully written by AI agents while also incorporating AI agents to respond to market signals, shop floor data, and supply chain interruptions in real-time. The only limitation is how you want to design your processes.
Orchestrating Autonomous Labs: Self-optimizing hardware that can iterate on physical prototypes while the team sleeps.
Hyper-Personalized Go-to-Market: The building doesn’t end at the shipping dock. AI allows builders to create messaging tailored so it feels like a 1-to-1 conversation with every prospect and customer.

🗣️ Shift Turnover

Once upon a time, Ford built airplanes and television sets, and General Motors built buses and locomotives. In general, car companies seemed to be a lot more ambitious then. All of that’s gone out the window. What are your thoughts on why that is and whether it could change?
I think it’s about to change more than people expect. As you redefine a software-defined product, with different levels of electrification, it turns out the electric architecture we’re building for our vehicles isn’t so different than electric architecture for a drone, isn’t so different than the electric architecture for VTOL (vertical takeoff and landing aircraft). It also turns out that as we start building batteries and getting closer to tech, we can build batteries, just like we built ventilators in COVID or bombers in World War II. And some of these adjacencies are very attractive businesses, number one.
Ford CEO, Jim Farley /Car and Driver/

"As a founder in CAD, you must answer one simple question: 'What will be possible to design and manufacture with my solution or what important workflows will be possible that are impossible today?'" /István Csanády on X/

Last week’s answer: Cards Against Humanity to Give Tariff Refunds to Buyers Who ‘Overpaid’ /Bloomberg/ and AI-driven factories /Schneider Electric/. Congrats to on winning!

Industrial Sector Insights

Assembly Line

This week’s Industry 5.0 breakthroughs and frontier technologies of the built world.

🔋 Donut Solid-State Battery: High Temperature Performance Test

Donut Lab has commissioned the internationally renowned VTT Technical Research Centre of Finland to conduct independent measurements of the battery’s performance and key characteristics. Unless otherwise stated, all tests presented here have been conducted by VTT. Full reports will be published alongside detailed video documentation that walks through the procedures, showcases the test setups, and explains the results.

Read more at Donut Labs and Two Bit da Vinci

EV Rivals Are Closing In—Can QuantumScape and Volkswagen Bring Solid-State Batteries to Market First? /Inc./ QuantumScape just announced a production milestone for its solid-state batteries.

⚗️ Using Individual Atoms to Achieve Fossil-free Chemistry

Methanol is a key starting material for chemical products. Researchers from ETH Zurich can now produce this precursor from CO2 and hydrogen with high efficiency by using isolated metal atoms as catalysts.

“Our new catalyst has a single atom architecture, in which isolated active metal atoms are anchored on the surface of a specially developed support material,” Pérez-Ramírez explains. In conventional catalysts, on the other hand, metals are usually present as aggregates, usually small particles. Although these particles are tiny, they often contain between a hundred and several thousand metal atoms.

🦾🧠 Bringing Robotics AI to Embedded Platforms: Dataset Recording, VLA Fine‑Tuning, and On‑Device Optimizations

Bringing VLA models to embedded platforms is not a matter of model compression, but a complex systems engineering problem requiring architectural decomposition, latency-aware scheduling, and hardware-aligned execution. Addressing these challenges is essential to translate recent advances in multimodal foundation models into practical and deployable embedded robotic systems.

This guide presents NXP’s hands‑on best practices for recording reliable robotic datasets, fine‑tuning VLA policies (ACT and SmolVLA), and highlights the real-time performance that NXP i.MX95 achieves after optimization.

New Product Introduction

Highlighting new and innovative facilities, processes, products, and services

🎪📶 Qualcomm Brings On‑Premises Industrial AI and Connectivity to a Siemens Factory Model at MWC Barcelona

At MWC Barcelona, Qualcomm Technologies, Inc. and Siemens AG showcased a technology implementation to accelerate the autonomous factory production. Qualcomm Technologies will be demonstrating on‑premise intelligence with AI integrated into a Siemens autonomous factory model. Together, the companies showcase the integration of private Industrial 5G connectivity with edge AI-based use-cases to enable more dynamic, efficient and intelligent autonomous production.

The demonstration includes:

An Automated Guided Vehicle (AGV) connected over a Siemens private Industrial 5G Network powered by Qualcomm technology, transporting materials through the factory model.
Assembly performed by a robotic arm communicating with the AGV and production cells.
An AI agent is running locally on Siemens IPC hardware equipped with the Qualcomm Cloud AI 100 Card, analyzing system state in real time to determine the root cause and recommend corrective actions.
Additional AI‑driven tasks powered by Qualcomm Cloud AI 100 Accelerator Card such as real‑time process quality inspection and on‑premises industrial intelligence.

🎪🦾 Tesollo and Techman Robot unveil robot for high-mix, low-volume production

At the 2026 Smart Factory & Automation Industry Exhibition, also known as Automation World or AW, Techman Robot Korea and Tesollo Inc. presented an articulated, multi-jointed grasping system they said is optimized for high-mix, low-volume production.

🎪🚜🧠 Driving Excellence Without Drivers: Vale Brucutu’s Shingo Prize Journey

Vale Brucutu Mine Autonomous Haulage System has been awarded the prestigious Shingo Prize— the world’s highest recognition for operational excellence. Located in Minas Gerais, Brazil, Vale Brucutu Mine Autonomous Haulage, one of Vale’s flagship operations, is home to one of the largest autonomous haulage fleets in the world. Take a glimpse into Brucutu’s journey of innovation, team empowerment, and commitment to continuous improvement that led to earning this renowned distinction from the Shingo Institute.

Yimin coal mine: what 100 AI-driven electric trucks taught the mining sector /Mining Technology/
Caterpillar Transforms the Construction Worksite with Advanced Technology, Services and Solutions at CONEXPO-CON/AGG 2026 /Caterpillar/
HD Construction Equipment Unveils Next-Generation Models at CONEXPO 2026, Accelerating Expansion into Advanced Markets /PR Newswire/

J. Mintzmyer of Value Investor's Edge discusses shipping stocks with the current situation going on with global shipping /YouTube/. Traders Warn $100 Oil Is Imminent If Iran War Keeps Raging /Bloomberg/

Business Transactions

This week’s top funding events, acquisitions, and partnerships across industrial value chains.

Humanoid Hype May Not Lead to Widescale Manufacturer Investment /Assembly/
The Rise and Fall of a 3-D Printing Empire /NYT/. Desktop Metal, a billion-dollar start-up, promised to revolutionize manufacturing. It went bankrupt, and now has much humbler ambitions as the 3-D printing industry takes a sober turn.

🇯🇵💿 Rapidus Secures 267.6 Billion Yen in Funding from Japan Government and Private Sector Companies

Rapidus Corporation announced that it has completed a funding round totaling 267.6 billion yen (equivalent to $1.7 billion USD) from the Japan government and private sector companies. Rapidus will continue to secure funding through capital increases and loans from both public and private sources. This strategic funding plan will enable the company to steadily progress from its current R&D phase to mass production of 2nm logic semiconductors by 2027.

Read more at PR Newswire

🇺🇸💿 Ayar Labs Closes $500M Series E, Accelerates Volume Production of Co-Packaged Optics /Business Wire/
🇳🇱💿 Axelera Raises $250M in Largest EU Semi Round Ever /EETimes/
🇸🇬💿 Allen Wu on Disrupting $100M Cost of Building Custom AI Chips /EETimes/. Wu’s CoreLab Technology wants to slash development entry-level cost by 75% using RISC-V as a baseline to enable heterogeneous custom AI chips.
🇮🇱💿 proteanTecs Receives Strategic Investment from TOPPAN Group Venture Arm TGVP /proteanTecs/

🇪🇸🚀 PLD Space secures €180M Series C to accelerate global access to space

PLD Space, an international space transportation company, has closed a €180 million Series C equity funding round led by the renowned Japanese manufacturer Mitsubishi Electric Corporation, alongside with other investors. The Spanish Ministry of Science, Innovation and Universities, through the Centre for the Development of Technology and Innovation (CDTI) and its INNVIERTE fund, and the Spanish public funds management company COFIDES, through its FOCO investment fund, have co-invested in this round. Ultimately, the European renowned Spanish fund Nazca Capital, via Nazca Aeroespacial y Defensa INNIVERTE I FCR Fund, close the round.

With over €350 million raised to date, this financing advances PLD Space’s strategic roadmap, supporting its transition to commercial operations and the scaling of its industrial and launch capabilities. PLD Space now consolidates its position among a select group of private companies worldwide developing complete launch systems, enabling global access to space supported by a strategic network of worldwide locations.

🇺🇸🧑‍💻 Nominal Valued at $1B as Founders Fund Leads $80M Acceleration Round

Nominal, the connected testing and operations platform for hardware engineering teams, announced an $80 million Series B-2 Acceleration Round led by Founders Fund, with participation from existing investors Sequoia Capital, Lux Capital, General Catalyst, Lightspeed, and Red Glass. The round values Nominal at $1 billion.

Nominal provides a unified platform that connects the entire hardware data supply chain — from instrumentation and data acquisition through analysis, reporting, and operational decisions. Its customers include teams building satellites, aircraft, autonomous and performance vehicles, energy infrastructure, and advanced defense systems — programs where reliability, speed, and operational clarity are critical.

Read more at GlobeNewswire and /Josh Wolfe on X/

🇺🇸🦾🧠 OpenAI’s Former Research Chief Aims to Automate Manufacturing With AI

✍️ Authors: Meghan Bobrowsky, Kate Clark

Arda, the new startup co-founded by Bob McGrew, is raising $70 million at a valuation of $700 million, according to people familiar with the matter. Founders Fund and Accel are coleading the round, with participation from Khosla Ventures and XYZ Venture Capital. The company’s round hasn’t yet closed, and terms of the deal could still change.
Arda is developing an AI and software platform, including a video model that can analyze footage from factory floors and use it to train robots to run factories autonomously, the people said. The company’s software will coordinate machines and humans across the entire production process, from product design and manufacturability to finished goods coming off the line.

🇦🇺⚗️ PlasmaLeap Raises A$30 Million to Accelerate Zero-Emissions Production of Fertilisers and Fuels

PlasmaLeap Technologies, the Australian company pioneering zero-emissions production of ammonia and nitric acid, has secured almost A$30 million (US$20 million) in new funding from a group of strategic and institutional investors, which was led by the Gates Foundation, Investible and Yara Growth Ventures, the venture arm of Yara International, a world leading integrated nitrogen fertilisers producer.

The Series A round, which closed in January, also included Twynam, GrainCorp Ventures, Uniseed/UniSuper, Artesian, SVG Ventures and Agnition Ventures, part of New Zealand fertiliser co-operative Ravensdown. The proceeds will be used to progress first-of-a-kind fertiliser hubs in New South Wales and Tasmania, expand field trials, and further develop PlasmaLeap’s core technology. Funding will also support longer-term applications in sustainable fuels and energy systems.

Read more at PR Newswire

🇮🇱♨️ Lime Kilns: From 8% of Global Emissions to CDR Facilitators /CarbonBlue/
🇳🇱♨️ RIFT raises EUR 114m to produce iron fuel for industrial heat /Renewables Now/

Looking for additional content? Paid subscribers get more stories delivered each week as part of Overcapacity!

Constraint Capital #3 - Mar 2026

David Rogers — Mon, 02 Mar 2026 14:01:33 GMT

February 2026 was defined by a sharp pivot: the "Asset Light" era of software dominance is being overwhelmed by a gritty, physical reality. While generative AI continues to accelerate, investors are seeking refuge in tangible moats (HALO) and bracing for the systemic fallout of a world where human intelligence is no longer a scarce resource (The 2028 Global Intelligence Crisis). Let’s take a look at the origins of each of these.

Heavy Assets, Low Obsolescence, or HALO, was coined by Josh Brown in a blog post on February 8th. For fifteen years, the market worshipped “Asset Light” business models—software companies with high margins and recurring revenue. However, the rise of agentic AI (like Claude Code and Codex) has turned these recurring revenues into disruptable targets. But the physical assets needed to power agentic AI are significantly constrained. Here’s a passage from Josh which sums up the rationale behind Constraint Capital:

Energy stocks, with a few exceptions, are HALO - Heavy Assets, Low Obsolescence. You get it now? This is the sweet spot of the stock market.
These are the names you can buy and not worry about. They can be thought of as potential beneficiaries of AI given the extensive costs of their existing operations, marketing, service delivery and decision-making processes. Or you don’t even have to go that far - you can just look at them and apply the following simple test: Will chatbots and LLMs lessen or eliminate the need for (blank) in the near future?

And some additional commentary from around the Internet

HALO: Heavy Assets Low Obsolescence

Josh Brown coined the term on Super Bowl Sunday. Within two weeks, it was in the Wall Street Journal…

2 months ago · Dan Mottram

The Halo effect: Heavy asset, low obsolescence . . . high returns? /FT/
Every previous technology wave concentrated more value in the orchestration layer and less in the physical layer. The most valuable electricity-era companies were the ones building appliances and media on top of cheap power, not the plants generating it. The most valuable internet-era companies were the ones organizing information and attention, not the ones laying fiber. AI follows the same playbook. The models capture more value than the data centers running them, and that spread is growing every quarter. /Aakash Gupta on X/

While HALO represents one strategy for understanding the markets, the 2028 Global Intelligence Crisis published February 22th, is a chilling thought exercise regarding the systemic consequences of AI being too successful. It describes a phenomenon coined the “intelligence displacement spiral.” Read the piece in full below:

Citrini Research

THE 2028 GLOBAL INTELLIGENCE CRISIS

Preface…

2 months ago · 7164 likes · 84 comments · Citrini and Alap Shah

Ultimately, the crisis isn't caused by AI failing, but by AI working perfectly. The article outlines a sequence of events that lead to a Ghost GDP—high productivity numbers that don't circulate in the real economy.

But the thought exercise was met with some rebuttals:

Gavin [Baker]: "The world is fundamentally short both watts and wafers and it may take years to resolve these shortages. The shortage of watts and wafers may prevent an overbuild - hyperscalers would overbuild if they could, but they simply cannot... My best guess is that we would need roughly 1000x more compute for the unlikely hypothetical scenario described by Citrini to be remotely possible and the time it takes us to get there will give humans time to adjust and maximize the many potential benefits of AI."

Shanu Mathew on X

The 2026 Global Intelligence Crisis /Citadel Securities/
Citrini’s Dystopian AI Vision Draws Global Investor Criticism /Bloomberg/

So where does this leave us? Constraint Capital Issue #3 breaks down HALO opportunities and the constraints blocking a global intelligence crisis. 👇

[Overcapacity] Cards Against AI-driven Factories

David Rogers — Sun, 01 Mar 2026 15:02:35 GMT

Thank you for being a paid subscriber! This week’s bonus media is available until 2025-03-22.

Click to access ⤵️

This weeks digest:

Cards Against AI-driven Factories

David Rogers — Sun, 01 Mar 2026 14:03:45 GMT

Shop Talk

Capturing this week’s zeitgeist

Get ready to level up, a brand-new industrial planning sim is dropping on March 31st, and it feels like the spiritual successor we’ve been waiting for. If you’ve been here a while, you probably remember our chaotic experiment playing Production Line using the early versions of ChatGPT. Well, it’s time to move from cars to tokens. The new game is called Data Center, and it’s essentially an AI Factory builder.

I’m dying to know: Can the latest generation of AI agents actually manage a data center built for itself? It’s a great meta-challenge. If you want to see me turn the keys over to the agents for a community play through, drop a comment below! If there’s enough interest, we’ll coordinate a deep-dive session and see if the AI can keep its own home running or if it all ends in a thermal meltdown.

Last week’s answer: Viral monkey Punch was clinging onto his stuffed orangutan while the US EXIM Bank, created during the Great Depression, helps American companies sell products overseas by financing foreign buyers.

More Thoughts on AI in Lab Automation

Assembly Line

This week’s Industry 5.0 breakthroughs and frontier technologies of the built world.

🚕🗣️ Elon Musk Drops News About CyberCab, FSD and Optimus

🚕🎨 Could Cybercab Be Petrochemicals’ Unlikely Lifeline? /Seti Park on X/ “If Cybercab’s paintless panels prove out — durability, cost, scalability — the incentive for other OEMs to follow is enormous. The moment one automaker eliminates its paint shop and lowers per-unit cost, competitors face two options: make the same switch, or compete with a structurally higher cost base. In automotive manufacturing, processes that demonstrate clear cost advantages tend to spread fast. Unibody construction did. Gigacasting is doing so right now.”
🛻 Tesla unveils cheaper Cybertruck variant, cuts Cyberbeast price to drive demand /Reuters/
🚕 Uber launches autonomous vehicles services venture in robotaxi push /FT/
🗣️ Elon Musk explains exactly when a company should vertically integrate /Jawwwn on X/

💿🏭🇺🇸 Inside Apple’s Multibillion-Dollar Push to Make Chips in the U.S.

Apple is beginning to bring its semiconductor manufacturing supply chain back to the United States. Nearly all of the most advanced chips are made in Taiwan, which China has threatened to annex. Concentrating chip supplies on an island that could be invaded, or face steep U.S. tariffs from Trump, is a big risk to Apple’s business. To find out how far the company still has to go, WSJ reporter Rolfe Winkler visited several of the company’s suppliers including TSMC, ASML and Foxconn in the Southwest.

🎛️🧠 Mike Labhart: On the shopfloor of AI-driven factories

Can AI really spot risks faster than any sensor? How far are we from the autonomous supply chain? In this episode of the AI at Scale podcast, Mike Labhart, North America Global Supply Chain Director for Smart Operations & Data Analytics of Schneider Electric, walks through the most valuable AI use cases in manufacturing and what’s next in smart factories.

👃🧠 Building Sigma: Lessons Learned Deploying Olfactory AI at the Edge /Kordel France on X/
🚗🧠 Cracking the Code: Scaling AI at Audi /Industrial AI Podcast/. Dive into Audi’s journey from piloting AI-based crack detection in sheet metal parts to scaling these solutions across multiple global sites. Andre shares practical insights on overcoming challenges like data consistency, model scalability, and workforce collaboration, all while aligning with Audi’s strategic vision.

🛒🦾🧠 Kroger’s (Simple) $2.6B Mistake

In 2018, Kroger partnered with UK automation company Ocado to roll out a nationwide network of “robot warehouses” — the famous “hive” system. It looked like the perfect corporate tech story… until the economics collided with American geography and American delivery expectations.

By late 2025, Kroger announced it would shut down multiple Ocado-powered automated fulfillment sites (including the Frederick, MD facility visited in this video), take a $2.6B charge related to the network, and pay Ocado a one-time $350M settlement as it “optimizes” the partnership.

Automation only works when the process is right /The Fabricator/
🇰🇷🏭🦾🧠 Physical AI startup RLWRLD raises $26M /The Robot Report/ for models trained using high-precision, multimodal data collected directly inside real factories, warehouses, and service environments.

Stefan’s Substack

I have a lot of thoughts about how AI can be used in lab automation. I thought this was an opportune time since SLAS just passed and the Ginkgo cloud lab demo has a lot of people thinking about automation…

2 months ago · 4 likes · Stefan Golas

LUMI-lab: A foundation model-driven autonomous platform enabling discovery of ionizable lipid designs for mRNA delivery /Cell/

New Product Introduction

Highlighting new and innovative facilities, processes, products, and services

🧠🫲🦾 The Physical Intelligence Layer

Our first use case, e-commerce order packaging, has historically been impossible to automate with robots. Large variability in workflow, item types, deformable packaging, and external machinery have created a “long tail” of problems that have been intractable to solve with traditional automation techniques which are often too rigid to be practical. Vision-language-action models (VLAs) provide a way to solve this by providing a recipe which improves in performance with data scale rather than engineering hours without reducing flexibility in the ways customers can use the robot.

Throughout our time working with Physical Intelligence, we have seen considerable improvements in real-world autonomous performance at our customer deployments. With each new model generation (π0 to π0.5 to π0.6) we have observed a significant step up in intelligence, throughput, and reliability with no signs of slowing down. With π0.6, we observe noticeable improvements in cycle times and success rates which compound to mean higher throughput for our customers and less fatigue for our remote intervention team.

Read more at Physical Intelligence and discuss on X

🇪🇺🧠🫲🦾 Supporting the new generation of physical AI /Google DeepMind Accelerator/. Their three- month program for robotics startups across Europe. Build the next generation of physical AI with Google DeepMind. Applications are open until March 25, 2026.
🇨🇳🧠🫲🦾 Spirit AI Lands $280M to Scale Embodied AI Through “Dirty Data” to scale VLA models /PR Newswire/

💿🔦 ASML unveils EUV light source advance that could yield 50% more chips by 2030

This is the quest to generate EUV light with the right power and properties to turn out chips at high volume. The company’s researchers have found a way to boost the power of the EUV light source to 1,000 watts from 600 watts now. The chief advantage is that greater power translates into the ability to make more chips every hour, helping to lower the cost of each.
In the United States, at least two startups, Substrate and xLight, have raised hundreds of millions of dollars to develop American competitors to ASML’s technology, with xLight securing government funding from President Donald Trump’s administration.

Business Transactions

This week’s top funding events, acquisitions, and partnerships across industrial value chains.

📑 2026 State of Manufacturing & Supply Chain: AI, Resilience, and the Path Forward /Fictiv/
🇺🇸🚛🧠 Einride and Legato Merger Corp. III Announces $113 Million Oversubscribed Capital Raise in Support of Proposed Business Combination /PR Newswire/

🇺🇸🦾📊 ZaiNar, the Foundation Layer of Physical AI, valued at $1B+ with $100M investment

ZaiNar emerged from nine years of stealth with a breakthrough platform for Physical AI, tech that turns any wireless network into a sensing system that continuously knows where everything is, without satellites, cameras, or drain on device power or compute. The company has raised more than $100 million at a valuation exceeding $1 billion.

“Physical AI needs a live, continuous feed of where everything is, and that dataset simply did not exist,” said Daniel Jacker, CEO and Co-Founder of ZaiNar. “By solving time synchronization at the sub-nanosecond level, we’ve turned existing infrastructure into the foundation layer for Physical AI. This funding accelerates deployment with carrier and enterprise partners globally.”

Read more at PR Newswire

🇺🇸🦾📊 Encord raises $60M to challenge Scale AI in physical AI data infrastructure /TFN/

🇺🇸🎛️ Revel Raises $150M Series B to Modernize the Software Layer Behind Hardware Test and Control

Revel, a unified software platform for hardware test and control, announced $150 million in Series B funding to accelerate its expansion across aerospace, defense, robotics, and industrial markets. The round was led by Index Ventures, with major participation from Redpoint Ventures and returning investors Thrive Capital, Felicis, and Abstract Ventures, as well as prominent angels such as Dylan Field, Figma’s co-founder and CEO.

Revel’s platform enables teams to visually configure hardware systems, monitor live telemetry, and safely issue commands in real time. Their programming language, RevelCode, combines intuitive, Python-inspired syntax with deterministic execution, precision, and debuggability for high-consequence environments. Revel accelerates testing, prevents costly errors, and powers the hardware systems that keep critical industries running.

Read more at Business Wire

🇺🇸📐 Flux, the AI hardware engineer, announces $37M in new investment /GlobeNewswire/
🇬🇧📐🖥️ BeyondMath raises $18.5M to build the ChatGPT of physics simulation /TFN/

🇺🇸🖨️🪙 Freeform closes $67m Series B to fund Skyfall laser melting platform

Freeform, a trailblazer in AI-driven metal 3D printing, has closed a $67 million Series B funding round which will fund the development and deployment of Skyfall, the company’s next-generation factory platform, scheduled to go live in the first half of 2026. The California-based company secured funding from a series of investors, including Apandion, AE Ventures, Founders Fund, Linse Capital, NVentures (NVIDIA’s venture capital arm), Threshold Ventures, and Two Sigma Ventures.

“With this raise, we’re launching Skyfall, the world’s fastest laser melting platform,” stated Erik Palitsch, CEO and Co-Founder at Freeform. “Skyfall increases production capacity by 25x and solidifies our position as the industry’s most advanced software-defined factory, and the only one deploying Physical AI to enable entirely new classes of products.” The company claims Skyfall will be capable of producing thousands of kilograms of high-quality parts per day, broadening its material offerings by more than 10 times. It will serve as the foundation of Freeform’s first automated, software-defined factory.

🇺🇸🖨️🔋 MATERIAL Unleashes Formless Energy, Defying the Laws of Industrial Gravity /Business Wire/. MATERIAL’s proprietary HYBRID3D™ technology prints energy directly into a component of any shape and size, removing the barriers between design intent and manufacturing reality.

🇮🇳🏭📄 Wootzwork raises $6.6M to bring predictability to offshore manufacturing

As global manufacturing shifts across regions, supply chains, and regulatory environments, execution risk has quietly become one of the biggest constraints on industrial growth. For many OEMs, the challenge is no longer access to factories, but the complexity of coordinating dozens of suppliers, quality systems, timelines, and interfaces across borders. Wootzwork was built to solve that problem. Today, the company announced a $6.6 million Series A round to scale a new model of manufacturing execution built around single-point accountability.

[Overcapacity] Clinging On to an Industrial Policy Punch

David Rogers — Sun, 22 Feb 2026 15:00:57 GMT

Thank you for being a paid subscriber! This week’s bonus media is available until 2025-03-22.

Click to access ⤵️

This weeks digest:

Clinging On to an Industrial Policy Punch

David Rogers — Sun, 22 Feb 2026 14:01:26 GMT

Shop Talk

Capturing this week’s zeitgeist

For those long-time subscribers (some 5+ years!) and new subscribers, I want to take a moment and reinforce the mission of Exponential Industry.

Exponential Industry seeks to be the highest quality mixed media newsletter for the business of high-tech manufacturing, energy, and supply chain. We seek to build a community across engineers, managers, and capitalists innovating within physical AI, additive manufacturing (3D Printing), cyber-physical systems (CPS), industrial automation, energy systems, and materials science.

Our Tenets

Direct from the source - We bias to primary sources by synthesizing thousands of company blogs, press releases, podcasts, deep-dive articles, tweets, and videos directly from where they are originally published.
Credit and attribution - We prominently display the author(s) and publication to give credit where credit is due.
Straight to the point - Our excerpts seek to answer what and why. We want to highlight the value proposition and result so each reader is briefed on a wide range of topics and is encouraged to dive deeper if the content is enticing to explore more.
Private and opt-in by default - Substack is used to deliver this publication directly to your inbox.

My ask to you: Keep clicking on the content within the weekly digests to give insight on which media you find valuable to ensure our content continues to hit the mark.

Last week’s answer: Sweeping Innovation Bothers Industrial Incumbents paralleled the "broomgate" scandal in Olympic curling, where high-tech directional brooms disrupted the sport by allowing players to steer stones regardless of the initial throw, with the way digital manufacturers like Fuyao bothered glass manufacturing incumbents in the USA.