NVIDIA CEO Jensen Huang: AI is Re-Industrializing America
NVIDIA CEO Jensen Huang stated that artificial intelligence is bringing manufacturing back to the U.S., re-industrializing the country after decades of offshoring. He noted that AI is creating demand, driving investments in the aging U.S. power grid and sustainable energy, and is market-driven rather than subsidy-driven. He directly linked grid upgrades and sustainable power generation to the expansion of computing power, rather than separate industrial policy projects.
He also mentioned that AI is creating jobs in construction and manufacturing for energy plants, chip factories, and data centers. The jobs specifically target power plants, wafer fabs, and data centers, covering both construction and manufacturing, rather than just software positions.
He continued by stating that AI is creating new companies and industries. In just the past six months, $400 billion has been invested in AI startups. He used this funding to demonstrate that demand has overflowed from chips themselves into new companies and industries.
He urged builders to collaborate with communities to establish trust and create local benefits in their hometowns, stating that the U.S. has the opportunity to create lasting benefits for communities while leading the next industrial revolution. Community engagement is a prerequisite for establishing operations, rather than a post-facto public relations effort.
In terms of market mechanisms, the payers are model companies, cloud vendors, and energy developers that need to secure computing power and electricity, while the sellers are chips, packaging, systems, land, electricity, and construction capacity. This is an event-driven capital expenditure chain: tech capital expenditures flow into the grid, power plants, wafer fabs, and data centers, benefiting those who have secured factory sites, grid connections, and long-term power purchase agreements, while traditional users sensitive to electricity prices and lagging projects that have not secured power and land are under pressure.
Public reports also indicate that NVIDIA is advancing multi-gigawatt implementations around the same "AI factory" narrative, including an initial 4.25 gigawatts capacity in Ohio, built and operated by energy companies, leased by model companies, with NVIDIA providing computing power and credit support.
Source: Public Information
ABAB AI Insight
After transforming NVIDIA from a gaming graphics card company into a data center computing platform, Jensen Huang's next step is not just to sell accelerator cards, but to lock the CUDA ecosystem into wafers, packaging, systems, electricity, and factories. The company is first advancing Blackwell for mass production in the U.S. with TSMC in Arizona, then bringing Wistron into Fort Worth to produce GB300, and pushing Corning and Coherent into fiber optics, lasers, and materials, following the path of "first controlling supply bottlenecks, then relocating factories to where demand occurs."
Capital is shifting from purchasing GPUs to pre-purchasing land, electricity, and factories. NVIDIA is using long-term visible shipments to incentivize energy companies to expand production, funding projects like SB Energy and providing leasing and electricity payment guarantees in exchange for exclusive computing power positions at the Ohio site, with OpenAI signing long-term leases. The flow of money is clear: model training budgets turn into power purchase agreements, which then become power plants and substations, with power plants issuing orders to construction, electrical, cooling, and chip supply chains.
This is closer to Rockefeller controlling the oil barrel before controlling the pipeline, rather than Intel's past model of only retaining design and outsourcing manufacturing overseas. When viewed alongside TSMC's Arizona plant, SK Hynix packaging, and Foxconn system assembly, the industry has shifted from "chips exported for use in the U.S." to a stage where "chips, systems, and electricity must expand in the same location," with location being a control period, not a conceptual expansion period.
Structurally, this represents a reconstruction of the industrial chain. The reason is that the marginal output of artificial intelligence is constrained by electricity and factory space; whoever first integrates "land—electricity—shell—chip" into a balance sheet will gain pricing power. Subsidies are not the main engine; long-term computing power contracts are, as these contracts can discount future token revenues into today's operational power plants and wafer fabs.
ABAB News · Cognitive Law
- Wherever computing power expands, factories and grids will follow.
- Whoever secures electricity will lock in the next generation of pricing power.
- The cost savings from offshoring will ultimately be added to electricity prices and safety.