In the most recent earnings calls, NVIDIA (August 26) and Broadcom (September 2) were the two executive teams that put land, power, and shell (LPS) or ready-to-energize buildings at the center of the AI data-center bottleneck. The industry phrase is usually “powered shell,” “warm shell.” “Energized shell” is the same idea: a building with contracted power and electrical and mechanical plant in place so racks can actually be plugged in.
NVIDIA went further than an earnings aside. On August 17, Jensen Huang published the note that named the scarce object: “Land, Power, Shell: The Next Strategic Resource”
For the vast majority of NVIDIA customers, he wrote, securing LPS has long been part of their infrastructure strategy. The world’s largest cloud service providers and investment-grade enterprises have the balance sheets, infrastructure expertise, and long-term contracts to secure LPS independently. They build and operate AI factories using NVIDIA accelerated computing, networking, systems, and software. That model will continue to represent most of NVIDIA’s business.
Frontier AI labs are different. They have extraordinary demand for training and inference compute, but many are growing faster than their balance sheets and long-term credit profiles can support. They may have strong customer demand and rapidly growing revenue and still lack the decades-long infrastructure contracts and investment-grade financing capacity needed to secure AI-factory infrastructure on their own. Their growth is increasingly constrained not by algorithms or customer demand, but by the availability of compute. For these companies, more compute means more intelligence, more products, more users, and more revenue. NVIDIA is helping provide the infrastructure that powers that flywheel.
The centerpiece of that August 17 note was a partnership with SoftBank’s SB Energy to secure LPS at the PORTS-Pike Technology Campus in Portsmouth, Ohio, with OpenAI as tenant /NVIDIA/. Initial capacity: 4.25 gigawatts of AI-factory infrastructure. Site potential if NVIDIA extends the arrangement: about 8 gigawatts. Huang’s economic point was explicit. The LPS commitment locks a long-lived site. The NVIDIA compute inside it can be upgraded every generation.
That is the frame. Both companies can see more chip demand than they can house. The physical market is not one pipeline. It is three inventories of buildings: energized now, ready to energize in the next few quarters, and planned to be energized in 2027–29. Where those buildings appear, and which rack geometries they can take, is what actually constrains NVIDIA and Broadcom.
So let’s dive in…