OpenAI’s $110 Billion Raise Ties Amazon, Nvidia and SoftBank Into a Closed-Loop AI Infrastructure Bet

On a Friday afternoon in late February, the company behind ChatGPT announced a deal more commonly associated with utilities than software startups: a $110 billion investment led by Amazon, Nvidia and SoftBank that values OpenAI at roughly $840 billion.

The Feb. 27 funding round is being described by bankers and analysts as the largest private tech raise on record, easily eclipsing OpenAI’s own $40 billion round last year. But the headline number obscures what may be the defining feature of the deal: much of the money is structured as phased commitments and is expected to cycle back to the very firms investing it, in the form of cloud bills, chip purchases and data-center contracts.

In effect, some of the world’s most powerful technology companies are financing a piece of AI infrastructure they will also be paid to build and operate.

OpenAI said the round values the company at about $730 billion before the new money and roughly $840 billion afterward. Amazon committed about $50 billion, Nvidia and SoftBank about $30 billion each, with smaller investors expected to join as the financing remains open. The company said it will use the capital to expand its computing capacity, build out data centers and accelerate deployment of new AI systems.

“We are entering a new phase where frontier AI moves from research into daily use at global scale,” OpenAI Chief Executive Sam Altman said in a statement announcing the deal. “Leadership will be defined by who can scale infrastructure fast enough to meet demand and turn that capacity into products people rely on.”

Largest private round yet, structured like a supply deal

Financial data firms and venture analysts say no private tech company has previously raised a single round of this size. OpenAI’s 2025 financing, a $40 billion SoftBank-led investment at a $300 billion valuation, held the prior record. Large late-stage rounds for companies such as SpaceX, Stripe and ByteDance have typically been measured in the low tens of billions at most.

The new deal is different not only in magnitude but in structure.

Amazon’s $50 billion commitment, people familiar with the terms said, includes about $15 billion in immediate equity and about $35 billion in additional funds to be deployed over time if OpenAI hits preset milestones. Some coverage, citing deal documents and people briefed on the talks, has reported that those triggers include either achieving a widely defined benchmark for artificial general intelligence or completing an initial public offering by the end of 2026. Those specific conditions have not been confirmed by the companies.

Alongside the equity, Amazon Web Services expanded an existing multiyear cloud agreement with OpenAI by roughly $100 billion, bringing the total value of that contract to about $138 billion over eight years, according to people familiar with the deal. OpenAI has agreed to consume about two gigawatts of Amazon’s custom Trainium chip capacity on AWS infrastructure during that period, the companies said.

Nvidia’s $30 billion stake is paired with its own supply commitments. The chipmaker will provide OpenAI with about 3 gigawatts of dedicated inference capacity and 2 gigawatts of training capacity using its forthcoming “Vera Rubin” GPU systems, on top of existing deployments of its Hopper and Blackwell chips across cloud providers.

SoftBank’s $30 billion investment, largely through its Vision Fund 2, is described as a follow-on that lifts its ownership in OpenAI to around 13%, according to people familiar with the cap table. The Japanese group has tied the deal to its broader push into “artificial super intelligence” and massive AI data-center projects in partnership with OpenAI and others.

One venture investor who has analyzed the agreements described the structure as “a supply chain deal dressed up as a venture round,” noting that the same companies providing equity are also locking in revenue streams over the next decade.

A capital-hungry business model

The round underscores the vast capital requirements of training and operating large-scale AI models. OpenAI’s revenue has grown quickly—analysts estimate it generated around $13 billion in 2025 and is on a much higher run rate this year—but the company remains unprofitable. People familiar with its finances say it posted a net loss of roughly $9 billion last year and expects more than $100 billion in expenses over the next four years.

OpenAI has told partners and investors that it plans to spend around $600 billion on computing infrastructure through 2030, including long-term contracts with cloud providers and commitments to data-center projects around the world. That figure is comparable to the capital spending of large telecommunications or utility operators, rather than conventional software firms.

The company says ChatGPT now has more than 900 million weekly active users, including over 50 million consumer subscribers and about 9 million paying business customers for its tools. A separate coding assistant has more than 1.6 million weekly users, according to recent metrics shared with investors.

Altman has argued publicly that such scale requires “massive capital and infrastructure” and has compared frontier AI projects to national-level undertakings.

Amazon, Nvidia and SoftBank’s motives

For Amazon, the deal is both an investment and a way to anchor OpenAI workloads on its cloud and custom silicon. AWS, which has faced stiff competition from Microsoft Azure in generative AI, stands to gain a large, predictable customer for its Trainium chips and data-center capacity.

Andy Jassy, Amazon’s chief executive, called the partnership “a major step forward” for customers building AI applications and agents, saying the companies would work together on “stateful runtime environments” that handle long-running AI tasks.

Nvidia, which already dominates the market for AI accelerators, appears to be using equity to reinforce what it views as a non-optional role in the AI stack. The company has previously framed data-center deployments in terms of about $35 billion in hardware and systems per gigawatt of capacity. At that scale, analysts say, its $30 billion investment in OpenAI could be recouped over time through chip and systems sales tied to the 5 gigawatts of capacity it plans to provide.

SoftBank, meanwhile, is concentrating more of its Vision Fund capital in a smaller set of AI bets after high-profile losses in companies such as WeWork. The Japanese conglomerate has already booked multibillion-dollar paper gains on its earlier OpenAI stake, bolstering recent quarterly earnings.

Microsoft’s role, even without new money

One conspicuous absence from the list of new investors is Microsoft, which has poured more than $10 billion into OpenAI since 2019 and is estimated to hold roughly 27% of the company.

Microsoft did not contribute additional cash to this round. Still, it stands to benefit from the higher paper value of its stake and from long-term cloud contracts with OpenAI that analysts say are worth hundreds of billions of dollars through the end of the decade.

In a joint statement after the new funding was announced, Microsoft and OpenAI said “nothing about the funding or new partners announced Friday in any way changes the terms” of their relationship and described their partnership as “strong and central” to both companies’ AI strategies.

Some industry observers say OpenAI appears to be diversifying its dependencies by adding Amazon and Nvidia as deep equity partners while maintaining its existing alignment with Microsoft and other infrastructure providers such as Oracle.

Pre-IPO positioning and regulatory questions

The scale and timing of the round have fueled expectations that OpenAI is preparing for a public listing. Several financial outlets have reported that investors view the financing as pre-IPO positioning for a potential offering as early as late 2026.

At a post-money valuation of about $840 billion, even a modest first-day trading gain would push OpenAI’s market capitalization toward the $1 trillion mark, placing it alongside the world’s most valuable companies from the outset.

Regulators are already scrutinizing the concentration of power around major AI developers and their cloud partners. The Federal Trade Commission has examined elements of the Microsoft–OpenAI arrangement, and antitrust experts say cross-equity stakes and exclusive distribution agreements among Microsoft, Amazon, Nvidia and OpenAI are likely to draw further attention.

Analysts also point to the systemic risks embedded in the “closed loop” funding model. If demand for generative AI services does not match today’s projections, the same firms could face losses on their equity positions, overbuilt data-center capacity and long-term compute contracts.

The infrastructure build-out carries environmental and social dimensions as well. The gigawatt-scale data centers tied to OpenAI’s commitments will require large amounts of electricity and water for cooling, raising questions from local communities and climate advocates about land use, emissions and resource allocation. Labor groups and researchers, meanwhile, are watching how a rapid expansion of tools like ChatGPT will affect white-collar jobs in fields such as programming, design, customer service and law.

For now, backers are betting that the demand will materialize. With nearly a billion people already using its products each week, OpenAI has little left to prove about interest in generative AI. The new capital is aimed at something different: building enough hardware, data centers and network capacity to make that interest a durable line of business—and, in the process, cementing a tight financial loop among the companies that supply the chips, run the clouds and now own a significant slice of the lab at the center.