The AI race is no longer primarily about who can buy the most advanced computer chips; it's about who can secure enough electricity to power frontier AI systems. While Washington's chip export restrictions remain a headline-grabbing policy tool, a more fundamental bottleneck is quietly reshaping which countries can actually host cutting-edge AI infrastructure: the ability to deliver gigawatts of continuous, reliable power.

Why Has Power Become More Important Than Chip Access?

For years, the US-China AI competition was framed almost entirely as a semiconductor story. The question was straightforward: who could buy enough Nvidia GPUs, and which countries would Washington's export-control regime permit to access them? That logic still holds some weight, but it no longer captures the full picture.

Training and running next-generation AI models now demands electricity at a scale that chips alone cannot explain. A single modern AI data center campus can draw between 500 megawatts to over a gigawatt of power continuously, equivalent to the electricity consumption of a mid-sized city. Since 2025, US grid operators have struggled to connect new large loads fast enough. Interconnection queues in AI hotspots like Northern Virginia, Phoenix, and Dallas now stretch four to seven years, while transformer orders take up to five years to fulfill.

The difference between a chip and a gigawatt of power is stark: a chip can be bought, leased, routed through a subsidiary, or smuggled. A gigawatt of firm, always-on electricity cannot. It requires transmission lines, transformers with multi-year lead times, and either an existing reactor fleet or a government willing to underwrite a new one. That difference is now the more decisive constraint.

How Are Tech Giants Responding to the Power Crisis?

Major technology companies have recognized that neither money nor chip access can accelerate grid connections. Microsoft, Google, Amazon, and Meta have each signed multi-billion-dollar, multi-decade deals to revive old nuclear reactors or pre-order ones that do not yet exist. The most visible example is Microsoft's $16 billion, 20-year agreement with Constellation Energy to restart Three Mile Island's Unit 1, the reactor infamous for the 1979 nuclear disaster. The renamed Crane Clean Energy Center is expected to generate power again in late 2027, with its output routed directly to Microsoft's data centers rather than the regional grid.

Even OpenAI's $500 billion Stargate initiative, despite having abundant capital and chip supply, showed no significant physical progress at its flagship Texas site as of April 2026. Nationally, roughly 12 gigawatts of AI data-center capacity was promised for construction in 2026, but only about 5 gigawatts actually broke ground. The best-capitalized, most chip-privileged companies on Earth are discovering that neither money nor an export license can buy a faster grid connection.

Steps to Understanding the New Nuclear-AI Nexus

• Nuclear Restart Strategy: Tech companies are signing long-term agreements to revive mothballed reactors because nuclear power is the only lever that can plausibly deliver firm, gigawatt-scale power this decade, bypassing years-long grid interconnection queues.
• International Cooperation Framework: US civil nuclear cooperation requires a "123 Agreement" with each partner government, and 2026's executive orders on nuclear set a target of twenty new agreements by 2028, effectively gating access to American reactor technology country by country.
• Grid Capacity as Strategic Resource: Power-delivery timelines of four to seven years, driven by transformer procurement and utility agreements, now set the schedule that chip access alone cannot accelerate, making grid capacity as decisive as chip export controls.

What Does This Mean for the Gulf States and Other Rivals?

Consider the Gulf region, where the constraint becomes even clearer. In July 2026, Washington eased chip export rules for the United Arab Emirates, elevating it to the top country tier for semiconductor access. Yet as of that same month, roughly 4,000 megawatts of announced Saudi and Emirati AI projects had zero documented energizations, industry shorthand for a project actually receiving grid power. Saudi Arabia's HUMAIN alone announced 6.6 gigawatts of AI capacity for 2034, but Riyadh currently operates only 467 megawatts of live data-center capacity against that target, a fourteen-fold gap.

Washington's permission to buy chips is necessary. It is no longer sufficient. The same constraint binds the United States itself, demonstrating that this is not a problem unique to countries facing export restrictions. The bottleneck is structural and global.

How Will the Nuclear Pivot Reshape Geopolitics?

The nuclear pivot reproduces the export-control logic in a different institutional form. Restarting Three Mile Island, pre-ordering TerraPower's Natrium reactors for Meta, or committing Amazon to X-energy's small modular reactors are all bets that firm power, however slow to arrive, beats waiting on a grid that cannot expand fast enough. Crucially, exporting that solution abroad runs through the same kind of bilateral vetting as chip diffusion tiers: US civil nuclear cooperation requires a "123 Agreement" with each partner government.

Access to American reactor technology and fuel will be gated country by country, just as chip access already is. The predictable objection is that most small modular reactor designs remain commercially unproven, with timelines stretching to 2032 and beyond. However, that does not explain away the 2026-2030 window, which is the one that matters for the current capital cycle. Gas turbine order books are already sold out for years, renewables cannot supply the continuous baseload that frontier clusters need, and only a finite number of mothballed reactors exist to restart.

Industry analysts point to three possible scenarios. In the base case, roughly 55% likely, incremental progress occurs: Crane and a handful of other US restarts come online close to schedule by 2027-28, and a first wave of small modular reactors reaches commercial operation by 2030. The US and hyperscaler-run capacity in a small set of trusted partner states consolidates a real lead. Gulf states continue announcing multi-gigawatt targets while operating a fraction of that capacity.

In a downside scenario, the bottleneck compounds rather than eases. A high-profile restart or small modular reactor project slips, while transformer and turbine supply chains remain the binding constraint regardless of reactor type. Frontier capacity concentrates even further into the handful of jurisdictions with spare grid headroom, and the ratepayer backlash already visible in regional electricity auctions hardens into state-level restrictions on new data-center interconnections.

In an upside scenario, licensing genuinely accelerates. The Nuclear Regulatory Commission's streamlined framework and Department of Energy co-location on federal land produce operating small modular reactors ahead of schedule, while a faster-than-expected run of new 123 Agreements extends trusted nuclear-and-AI status to partners beyond today's shortlist. In this scenario, the 123 Agreement, not the chip license, becomes the document worth tracking.

What About China's AI Momentum?

While the US grapples with power constraints, China's AI sector is experiencing a dramatic acceleration. Chinese startup Moonshot AI's valuation quintupled to $50 billion between February and August 2026, following a $3.5 billion funding round that closed in late July. The catalyst was the July release of Moonshot AI's Kimi K3 model, which boasts 2.8 trillion parameters, making it the world's largest open-weight model to date.

Within 24 hours of its release, Kimi K3 decimated global leaderboards, performing on par with top-tier American models. On the developer platform Code Arena, Kimi K3 captured the global number one spot for front-end development capabilities, marking the first time a Chinese large model had claimed that crown. The launch caused a temporary computing power shortage, forcing Moonshot AI to pause new user subscriptions.

"Before the release of Kimi K3, the market generally believed that due to computing power constraints, the gap between Chinese large models and top US models had widened to eight to 12 months. When it finally came out, it exceeded expectations. This tells the world that Chinese model teams can maintain and even narrow the gap with advanced US models," said an investor familiar with Moonshot AI.

Investor familiar with Moonshot AI

The geopolitical implications were immediate. David Sacks, tech investor and co-chair of the US President's Council of Advisors on Science and Technology, publicly warned in July that while America "is tying itself in knots" with regulations, China was winning the AI race. Washington hardliners quickly proposed sanctioning Chinese open-source models. This sparked a revolt in Silicon Valley: nearly 200 tech companies drafted a letter urging the US government not to restrict access to China's open-weight AI. Shortly after, Nvidia CEO Jensen Huang led a coalition of over 270 companies, including Meta, Microsoft, OpenAI, and Google, in signing a letter emphasizing the critical importance of open-source collaboration.

Moonshot AI's founder and CEO Yang Zhilin is a star prodigy who graduated first in his computer science class at Tsinghua University before earning his PhD at Carnegie Mellon University in just four years. He authored foundational papers on machine learning with over 52,000 citations and declined lucrative offers from US tech giants to return to China and build his own company. In early 2023, following the global awakening sparked by ChatGPT, Yang launched Moonshot AI with the explicit goal of achieving artificial general intelligence.

The bottom line: the US-China AI race is being decided not just by who can access the best chips or build the most capable models, but by who can secure the power infrastructure to run them. As electricity becomes the new strategic bottleneck, the 123 Agreement may soon matter more than the chip export license.