Time Scaling Industrial Attraction 💓

This week: inference-time scaling reasoning AI, ERP becomes core to AI strategy, world's biggest crane ship, rare-earth magnets, humanoids raise funds, tariffs impact automation.

Feb 16, 2025

∙ Paid

Shop Talk

Capturing this week's zeitgeist

Bottoms up conferences like ProveIt! (this Tuesday!) and Reindustrialize are proving vital for the next wave of innovation industrial technology. These grassroots gatherings enable operators to share their hands-on experiences cutting through marketing hype by showing solutions to practical challenges in Industrial DataOps, AI, robotics, 3D printing, and supply chain optimization. Register now to join the community forging the future of heavy industry.

Last week's 'Custom Quote' will provided insights into humanoid robots rapid growth trajectory. This week we discuss operational transparency, so we can track manufacturing like we do pizza. Upgrade 👇

Assembly Line

This week's most influential Industry 4.0 media.

⏳ Automating GPU Kernel Generation with DeepSeek-R1 and Inference Time Scaling

✍️ Authors: Terry Chen, Bing Xu, Kirthi Devleker

As AI models extend their capabilities to solve more sophisticated challenges, a new scaling law known as test-time scaling or inference-time scaling is emerging. Also known as AI reasoning or long-thinking, this technique improves model performance by allocating additional computational resources during inference to evaluate multiple possible outcomes and then selecting the best one, neural network. This enables AI to strategize and systematically solve complex problems in a similar fashion to how humans dissect complex problems and solve them individually to arrive at a final solution.

In this post, we talk about an experiment done by NVIDIA engineers who used one of the newest open-source models, the DeepSeek-R1 model, together with additional computing power during inference to solve a complex problem. The experiment was to automatically generate GPU attention kernels that were numerically correct and optimized for different flavors of attention without any explicit programming.

Epicor’s Josh Roberts on Harnessing Data and Reskilling the Manufacturing Workforce for a Tech-Driven Future

🖥️☁️ It goes down in the ERP:

SAP Databricks, is sold by SAP as part of SAP Business Data Cloud on Azure, AWS, and GCP. [Databricks]

♻️ AMP CEO shares insights into the company’s waste sortation strategy

✍️ Author: Brianna Wessling

AMP’s goal is to divert more waste from landfills by ensuring the removal of more recyclables and organics out of the waste stream. This includes understanding and being able to pick anything that comes through the line, which requires advanced AI. AMP is working to build all of its technology in-house.

“The technology is homegrown, I think we have 60 or 70 patents over the years that Matanya Horowitz and his engineering team have built and put into practice,” said Stuart. “The brains are all AMP, and we build some aspects of the equipment. But we also go procure some other parts of it to really put that whole sortation line in.”

The Logistics Inside The World's Biggest Crane Ship

🛥️ Going to the boat show:

The Life of a Shipping Container [YouTube]
How to Build Out a Machine Shop for Submarine Parts [Modern Machine Shop]
Pushing boundaries of ship design and simulation with Generative AI [Siemens]

Time Code for multifunctional 3D printhead controls

✍️ Authors: Sarah Propst, Jochen Mueller

Direct Ink Writing, an extrusion-based 3D printing technique, has attracted growing interest due to its ability to process a broad range of materials and integrate multifunctional printheads with features such as shape-changing nozzles, in-situ curing, material switching, and material mixing. Despite these advancements, incorporating auxiliary controls into Geometry Code (G-Code), the standard programming language for these printers, remains challenging. G-Code’s line-by-line execution requires auxiliary control commands to interrupt the print path motion, causing defects in the printed structure. We propose a generalizable time-based synchronization approach called Time Code (T-Code), which decouples auxiliary control from G-Code, enabling uninterrupted print path enrichment. We demonstrate the method’s effectiveness with both high-end and affordable 3D printers by fabricating functional gradients and parallelizing printhead auxiliary devices for mass customization. Our method reduces defects, enhances print speed, and minimizes the mechanical burden on 3D printers, enabling the rapid creation of complex multimaterial structures.

New Product Introduction

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

MP Materials Restores U.S. Rare Earth Magnet Production

MP Materials (NYSE: MP) has reached a pivotal milestone in restoring the U.S. rare earth magnet supply chain. The company’s flagship Independence facility in Fort Worth, Texas, has commenced commercial production of neodymium-praseodymium (NdPr) metal and trial production of automotive-grade, sintered neodymium-iron-boron (NdFeB) magnets. These achievements bring the U.S. closer to reestablishing a fully integrated, domestic supply chain for these critical components for the first time in decades.

Yokogawa Releases OpreX Quality Management System to Accelerate DX in Quality Assurance Processes

Yokogawa’s OpreX Quality Management System is a cloud-based system that can accelerate the digital transformation of quality assurance processes for the manufacturing of pharmaceuticals and food & beverages. It streamlines operations and allows for continuous improvements by enabling the tracking and centralized management of quality assurance processes such as the handling of deviations from manufacturing procedures and specified standards, implementation of CAPA, changes to manufacturing procedures, and revision of documents.

Business Transactions

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

Lower early rounds and higher later stage rounds in 2024? [Carta]
VC funding in European defence and security tech surges to record $5.2bn [FT]
VC due diligence and analysis work being done by Perplexity Deep Research

🇺🇸 Apptronik Raises $350 Million to Scale Production of AI-Powered Humanoid Robots and Meet Significant Customer Demand

Apptronik, the AI-powered humanoid robotics company, announced the successful closing of a $350 million Series A funding round co-led by B Capital and Capital Factory, with participation from Google. The new funding marks a pivotal moment for Apptronik, fueling the deployment of Apollo—its groundbreaking humanoid robot—while enabling the company to scale operations, accelerate innovation, and grow its world-class team to meet the surging demand for next-generation, AI-powered humanoid robots. Backed by key partners and investors, this new funding fuels the continued development and deployment of Apollo, positioning Apptronik at the forefront of creating scalable, AI-powered solutions for industrial applications and consumer applications in the near future.

🇺🇸 Archer Raises $300M From Leading Institutional Investors To Accelerate Hybrid Aircraft Platform Development As Defense Opportunities Look Stronger Than Expected

Archer announced it has raised $301.75M, further reinforcing Archer’s strong financial position and strategically positioning it to accelerate the development of its hybrid aircraft platform for the defense market and beyond. Leading institutional investors participated in this financing, including funds and accounts managed by BlackRock. This raise brings Archer’s total liquidity position to ~$1B. Archer has long maintained one of the strongest balance sheets in the industry, and this additional capital further strengthens its position.

The financing provided for the purchase and sale of 35,500,000 shares of Archer’s Class A common stock at a price of $8.50 per share based on a volume-weighted average price of the Class A common stock, in a registered direct offering. The net proceeds from the offering announced will be used for the development of next generation aircraft manufacturing capabilities related to this effort, including batteries and composites, and the remainder for general corporate purposes.

🇺🇸 Karman Announces Closing of its Upsized Initial Public Offering, with Full Exercise of Over-Allotment Option

Karman Holdings Inc., a company specializing in the upfront design, testing, manufacturing, and sale of mission-critical systems for existing and emerging missile and defense, and space programs, announced the closing of its initial public offering of 26,450,000 shares of its common stock at a public offering price of $22.00 per share. Karman currently intends to use the net proceeds it receives from this offering, together with its existing cash, cash equivalents and short-term investments, for general corporate purposes, including additional development efforts, working capital and operating expenses.

🇨🇦 Hydrostor Announces $200 Million in Funding from Leading Investors to Accelerate Canadian and Global Deployment of its Advanced Compressed Air Energy Storage Projects

Hydrostor, a global long-duration energy storage (LDES) developer and operator, has secured a $200 million USD investment from Canada Growth Fund Inc. (“CGF”), Goldman Sachs Alternatives, and Canada Pension Plan Investment Board (“CPP Investments”). The transaction will support Hydrostor’s continued investment in Advanced Compressed Air Energy Storage (A-CAES) projects in Canada and around the world.

Hydrostor’s late-stage projects in New South Wales, Australia, and California, USA, are targeted to begin construction in 2025. Its A-CAES technology represents a critical solution underpinning the future reliability of energy systems globally, enabling utilities to replace end-of-life fossil generation with zero-emission storage and accelerate the integration of intermittent renewables.