Jensen Huang Says AI Grid Investment Will Drive Reindustrialization
NVIDIA CEO Jensen Huang stated that investments in energy and the grid surrounding artificial intelligence in the U.S. are creating jobs and could drive reindustrialization. There is an opportunity for manufacturing, which has been lost over the past fifty years, to return, and now is the time to invest in the energy sector and the grid. Without energy growth, there is no industrial growth, and without industrial growth, there is no job growth. He referred to data centers as AI factories, alongside chip factories and supercomputer manufacturing, being developed simultaneously across the U.S.
He mentioned that related infrastructure has created about 500,000 jobs, with rising demand for electricians, welders, and construction workers. In the collaboration between Sherman, Texas, and Coherent, NVIDIA has invested about $2 billion, with the factory expected to create around 1,000 jobs, of which about 550 will be in advanced manufacturing and engineering; the production target is to quadruple output within 12 to 18 months. Regarding Trump's energy policy, he stated that without increased production, it would be difficult to build chip factories, supercomputer factories, and AI factories. He advocates for a mix of energy sources, distributed generation, and accelerated nuclear power, as relying solely on the existing grid is insufficient. He has previously stated that the energy required for computing could be three orders of magnitude higher than current levels, setting a ceiling on intelligent output due to energy constraints.
Reindustrialization is framed as a closed loop of electricity, factories, and jobs, rather than just a tariff slogan. NVIDIA sells chips and complete systems, while the grid determines how much can be shipped. The 500,000 jobs are for the construction phase and do not equate to long-term positions after factory mass production. The Texas construction site serves as a test: can AI bring atoms back to U.S. production lines?
The power gap is estimated by consulting firms to be in the hundreds of gigawatts. Whoever secures electricity first will be able to convert capital expenditures into revenue. Huang describes national industrial policy as a solution to his own supply constraints.
In market mechanisms, this is about capital expenditures in electricity hedging against the limits of chip shipments. Buyers are cloud and sovereign customers looking to build factories; sellers are chip companies and the White House framing energy policy as a reindustrialization narrative. Funds are flowing from data center budgets into generation, transmission, cooling, and packaging. Beneficiaries include electrical contractors, U.S. assembly plants, and NVIDIA's systems business; those under pressure are projects queued behind the grid and research teams still viewing algorithms as the sole bottleneck. The event-driven narrative is that the CEO is harvesting jobs and repatriation from the national narrative following the earnings report.
Source: Public Information
ABAB AI Insight
Jensen Huang shifts the bottleneck from model parameters to substations. Chips can already be sold, but electricity cannot be sold; reindustrialization means integrating power plants and packaging plants into the same construction blueprint. The fifty years of manufacturing outflow is a political statement, while the 500,000 infrastructure jobs are a current reality; the gap between these two statements is filled with nuclear power permits and transmission corridors. The Sherman factory is the tangible evidence he must present: laser and optics must quadruple in the U.S. to signify that 'atoms have returned.'
The capital pathway involves cloud vendors renting gigawatts, NVIDIA selling complete systems, and energy companies selling electricity and land. A site in Ohio is reported to be planned for 4.25 gigawatts, with OpenAI as a tenant, potentially generating revenue equivalent to 150,000 to 200,000 GPUs. Distributed generation acknowledges that the public grid cannot keep up. Those who can self-generate power can change their waiting list into a start date. Tariffs aimed at forcing manufacturing back have only shifted the gap from overseas to the construction site.
The analogy is drawn to 20th-century hydroelectric dams driving the aluminum industry and the shale gas boom that once lowered U.S. electricity prices to attract chemicals. The current phase is about seizing electricity: algorithms are no longer the constraint; land, shells, and grid connectivity are. If reindustrialization remains just a speech, jobs will remain in the construction phase; if electricity truly reaches the factories, jobs will shift from pouring concrete to operational roles.
Structural judgments pertain to the reconstruction of the industrial chain. Smart production is sinking from the software layer to the power layer, making the premise for manufacturing repatriation the return of power generation. The mechanism is: without new gigawatts, chip factories are just empty buildings; without empty buildings, tariffs cannot translate into jobs. Huang frames the national grid as his own capacity function, while the state frames chip companies as billboards for reindustrialization. Whether the function and the billboard can coexist depends on when the next substation is energized.