NVIDIA CEO Jensen Huang stated on August 30, 2026, that AI is not merely a digital revolution, but the beginning of the reindustrialization of the United States. After decades of offshoring, AI is creating new economic incentives to bring manufacturing and strategic supply chains back home. This is supported by US$400 billion invested in AI startups over the past six months, a figure that reflects strong capital confidence in AI as the next generation of economic infrastructure, not an indication of realized profits.

The fundamental shift is AI's transformation from a software product into a massive industrial project. Unlike in previous eras, AI requires physical elements that are not easily moved: semiconductor chips, massive power supplies, land for data centers, internet networks, and skilled engineering and construction workers. NVIDIA has built a manufacturing and supplier network across 43 US states, covering semiconductors, circuit boards, assembly systems, and AI infrastructure, creating an entirely new industrial ecosystem.

A crucial aspect that is often overlooked is AI's dependence on industrial-scale electricity. Behind a single chatbot question lies a long chain of demand: GPU → server → data center → electricity → cooling → network → chip → memory → power system. Huang highlighted that the AI boom is driving major investment in the aging US power grid and the development of new energy sources. This chain creates a multiplier effect: AI drives data centers, data centers drive electricity demand, electricity drives power plants, turbines, transformers, cables, and factories, ultimately creating construction and manufacturing jobs.

The US$400 billion figure needs to be viewed carefully; it represents venture capital bets, not profits. Funds are flowing not only to AI model makers, but also to AI *coding*, AI agents, robotics, *physical AI*, cybersecurity, semiconductors, and energy. Startup Lovable announced US$400 million in funding at a valuation of US$13.3 billion, while China's Zhipu AI reported a 400% surge in revenue to 953.9 million yuan, signaling that competition is shifting to competition between companies, countries, and ecosystems.

NVIDIA itself is transforming into an AI infrastructure company. In fiscal second-quarter 2027, revenue reached US$96.2 billion (up 18% QoQ), with the Data Center segment accounting for approximately US$89 billion (up 117% YoY), and next-quarter guidance at US$108 billion. Huang emphasized, *"compute is revenue"*. NVIDIA is working with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to mobilize more than US$500 billion in third-party capital for AI infrastructure development, marking the treatment of compute as an infrastructure asset.

Concrete evidence can be seen in Texas and Arizona, which have become new laboratories. Arizona is focused on chips and packaging (partner Amkor), while Texas is focused on supercomputers and AI systems (Wistron's facility in Fort Worth). SK Hynix announced a US$4 billion facility in Indiana for next-generation HBM production, with production targeted for 2029 and approximately 7,000 jobs to be created. This demonstrates that AI is drawing semiconductor supply chains back to the US.

The next phase is Physical AI, in which AI operates in the real world through robots, automated factories, vehicles, and digital twins, transforming factories into "thinking machines" capable of detecting defects, optimizing electricity use, and simulating production lines. If successful, US manufacturing costs could become competitive despite high wages.

However, the jobs question is not simple. Data center construction creates demand for electricians, welders, HVAC workers, and construction workers. Labor unions support it. However, post-operational automation could reduce the workforce. Brookings cautions that construction jobs are temporary and do not necessarily create long-term technology employment in every region. The key question is what types of jobs are being created and who has the skills to fill them.

The greatest irony is that the main constraint is not GPUs, but electricity. The US has capital and talent, but without sufficient electricity, AI factories cannot operate. Huang warned that energy constraints are a major problem, causing the AI race to shift into an energy race. Whoever has cheap and stable electricity will have the advantage.

This opens broad opportunities in semiconductors, HBM, networking, cooling, transformers, power generation, the grid, construction, robotics, and cybersecurity. However, the Financial Times warned of financing risks: if monetization is slow, massive capex will lead to low utilization, disappointing cash flows, rising debt, and valuation repricing. US$400 billion is not profit, but capital betting on the future of AI.

Over the next 5-10 years, there are three scenarios. Bullish scenario: AI boosts productivity, mass robotics emerge, compute costs fall, manufacturing becomes competitive, and AI becomes the largest economic engine. Moderate scenario: AI develops rapidly, but productivity does not grow as quickly as expected; consolidation occurs, and investors distinguish between AI that generates money and AI that burns capital. Bearish scenario: excessive capex, AI efficiency reduces compute demand, monetization fails, energy costs rise, and a bubble correction occurs, while the infrastructure remains an asset.

In conclusion, Huang is talking about a revolution larger than ChatGPT. US$400 billion shows the scale of capital, but more important is what happens after the money enters: capital requires chips, chips require factories, factories require electricity, electricity requires grids and power plants, and all of it requires workers and construction. AI is not bringing back the old manufacturing era, but creating a new manufacturing era controlled by software, robots, and compute. The US does not need to bring back old factories; it only needs to build the next generation of factories designed for the AI world. If Huang's thesis is correct, the real battle is over who controls physical infrastructure. AI does not live in the cloud; it requires electricity, chips, buildings, networks, machines, and people. The digital revolution is gradually becoming the next industrial revolution.

#AIStartupsRaise400BInSixMonths $NVDA