OpenAI’s Latest Move Signals a New Phase in the Global AI Arms Race

Every startup has a point in its existence when it ceases to be a startup. When the modest beginnings of the nonprofit, which was established in a San Francisco office with the goal of helping people, subtly give way to something more difficult to define, something larger, and, depending on your point of view, either more thrilling or more concerning. That moment may have quietly arrived for OpenAI at the end of February when the company announced a $110 billion funding round that caused everyone in the technology industry to pause and reevaluate the figures.

SoftBank, Nvidia, and Amazon. Three of the most influential players in the world of technology are simultaneously writing huge checks to the same business. OpenAI agreed to run massive workloads on Amazon’s own Trainium chips as part of Amazon’s $50 billion commitment, which has far-reaching implications for the two businesses. Nvidia, which already provides the GPU engines that drive the majority of OpenAI’s infrastructure, contributed an additional $30 billion, solidifying a partnership that is becoming more and more dependent on one another. SoftBank made its third significant investment in OpenAI in as many years, bringing in an additional $30 billion. OpenAI’s valuation increased to $730 billion after the combined round, which is said to be the biggest private funding event in Silicon Valley history. Three years ago, this amount would have seemed absurd.

It’s hard to ignore the gravity of what’s happening as you watch this play out. This is not a capital-raising startup. This is more akin to an industrial alliance, the kind of strong structural ties that typically take decades to establish between platforms, developers, and infrastructure providers. The distinction between OpenAI’s suppliers and investors is now completely hazy. Amazon is a chip partner, a cloud provider, and a backer all at once. In addition to being an investor, Nvidia is the manufacturer of the hardware that OpenAI depends on.

This arrangement, which involves money going from big tech to OpenAI and then going back out through contracts for hardware and cloud services, has been dubbed “circular financing” by critics. Sam Altman has challenged that interpretation, contending that the system is sound as long as actual revenue is increasing. OpenAI reportedly made $13 billion in revenue in 2025 and is expected to double that amount this year. He might be correct. Additionally, $110 billion in new capital and $13 billion in revenue might be on completely different planets.

This statement of ambition is astounding. A $400 billion commitment has been made to OpenAI’s Stargate project, a collaborative data center buildout with Oracle and SoftBank. Altman mentioned plans to bring 17 gigawatts of computing capacity online when he discussed spending trillions on infrastructure in the upcoming years. That is about the same as seventeen nuclear power plants, each of which requires ten years to construct. This is anticipated to be absorbed in some way by the U.S. electricity grid, which is already nearing capacity in many areas. According to reports, gas turbines will be completely sold out by 2028. The speed at which new energy infrastructure permits are granted is unrelated to Altman’s desired construction pace. The execution timeline and the ambition are on different tracks, and it’s still genuinely unclear how they will intersect.

Additionally, OpenAI’s focus is shifting, which may have a greater impact than the raw capital figures. The ChatGPT product, which became a household name for the company, was a consumer play: quick, easily accessible, sometimes unreliable, and constantly discussed at dinner tables. The new actions are unmistakably corporate. OpenAI’s reach into enterprise workflows was expanded by the Databricks integration. The deliberate attempt to connect with hundreds of portfolio companies at once through private equity outreach, which is purportedly offering firms a guaranteed 17.5 percent minimum return through joint ventures, effectively uses financial incentives to accelerate adoption across entire industries in ways that direct sales could never. It’s a smart move, if a little out of the ordinary for a company that used to define its mission solely in terms of science.

All of this is being driven by a genuine and persistent competitive pressure. With Google’s strong support, Anthropic has been gaining significant traction with corporate clients and is pursuing similar enterprise partnerships, but OpenAI is dangling without the guaranteed-return sweeteners. Altman was compelled to publicly pledge to accelerate product releases after DeepSeek, a Chinese startup, shocked Silicon Valley at the beginning of 2025 by releasing a capable model at a fraction of the cost. OpenAI no longer operates from a comfortable lead; instead, it feels more like a sprint with no clear finish line.

Beneath all of this activity is a deeper question that is not sufficiently raised: what does it mean that the race is now more about who controls the infrastructure than it is about which model is smarter? There isn’t really a wager on any particular AI capability in the $110 billion round. It’s a wager on compute, cloud contracts, distribution networks, and who can integrate themselves most thoroughly into global enterprise workflows before anyone else. Nowadays, logistics take precedence over technology. Perhaps the most important thing that OpenAI’s most recent actions actually indicate is the change from AI as science to AI as infrastructure war. Depending on what you initially believed to be the goal, that may or may not be progress.