Charting the OpenAI ‘ecosystem’ - Financial Times
OpenAI is no longer just a research lab with a popular chatbot. It is the center of a $730 billion bet — and everyone involved needs it to work.
The Financial Times this week published the most detailed map yet of OpenAI's sprawling ecosystem of capital, compute, and contractual obligation. The picture is striking: money flows in from Microsoft, Oracle, Nvidia, Amazon, and SoftBank. Infrastructure — GPUs, cloud capacity, data center floor space — flows back out. The whole arrangement is increasingly circular, and increasingly fragile. The question regulators, competitors, and partners are quietly asking is whether these are partnerships or dependencies wearing a different name.
Start with the scale. OpenAI raised $6.6 billion in equity in October 2024 at a $157 billion valuation. Less than five months later, it closed a $110 billion round at a $730 billion pre-money valuation — $50 billion from Amazon, $30 billion from Nvidia, and $30 billion from SoftBank — making it the most valuable private technology company in history by a wide margin, according to OpenAI's own announcement and reporting by TechCrunch and CNBC. At the same time, OpenAI is projecting roughly $25 billion in annualized revenue by 2026, per DigiDai, up from a standing start a few years ago. ChatGPT, the product that made all of this possible, went from around 400 million weekly active users in early 2025 to approaching 800 million by late 2025, according to TechCrunch. The growth is real. So is the burn.
The capital that built this was not OpenAI's own. It came from a small number of very large technology companies, each of which now has a structural interest in OpenAI's continued operation — and, in some cases, continued dependence.
Microsoft was first and deepest. Between 2019 and 2023, the company built its investment to nearly $14 billion, per TechCrunch's review of infrastructure deal data. For years, OpenAI ran its workloads exclusively on Microsoft Azure — an arrangement that gave Microsoft both a customer and a halo effect as the AI boom drove Azure growth. Microsoft held an observer seat on OpenAI's board after the November 2023 governance crisis, but dropped it in July 2024. OpenAI announced it would no longer use Microsoft's cloud exclusively as of January 2025.
That shift matters. When OpenAI was captive to Azure, Microsoft held significant leverage. Now it is one voice among several. Oracle signed a $30 billion cloud services deal with OpenAI in June 2025 and a separate five-year, $300 billion compute agreement beginning in 2027, per Oracle's SEC filings and reporting by TechCrunch. Nvidia committed $100 billion in AI systems investment to OpenAI in September 2025. OpenAI has invested $12 billion in CoreWeave, the cloud provider whose largest customer is Microsoft — a detail that creates its own small circular dependency. The Stargate project, a joint venture between SoftBank, OpenAI, and Oracle announced by President Trump in January 2025, targets $500 billion in AI infrastructure spending.
The numbers are large enough that they resist comprehension. Hyperscalers are planning to spend nearly $700 billion on data center projects in 2026 alone, per TechCrunch. OpenAI alone, according to analyst Tomasz Tunguz, has committed over $1.15 trillion in infrastructure spending across Broadcom, Oracle, Microsoft, Nvidia, AMD, Amazon, and CoreWeave between 2025 and 2035.
The question of leverage is therefore not abstract. Deloitte framed it directly in a March 2026 analysis: at what point does partnership become dependency? Gillian Crossan, Deloitte's global technology, media, and telecommunications industry leader, told the firm that 90 percent of all AI compute today is managed by U.S. and Chinese companies, making sovereignty "a bigger issue around the world." The risk is that companies mapping their AI strategy are not choosing partners freely — they are choosing from an increasingly narrow list of providers, any of which is simultaneously a customer, a supplier, and a competitor to everyone else in the loop.
Regulators are paying attention. The UK's Competition and Markets Authority has examined the OpenAI-Microsoft relationship specifically and identified something structurally important: Microsoft has acknowledged it has held the ability to materially influence OpenAI's commercial policy since 2019, per Stanford Law's CodeX center. The CMA has further found an interconnected web of 90 partnerships between Big Tech players and AI developers — suggesting the OpenAI-Microsoft dynamic is not an anomaly but the shape of an entire industry. The question is not whether these arrangements are scrutinizable but which ones cross lines that current antitrust frameworks were not designed to handle.
There is a coherent argument that this mutual dependency serves as its own check. Nvidia needs OpenAI as a showcase customer for its AI systems; OpenAI needs Nvidia for the hardware that runs its models; Oracle and Amazon need OpenAI to justify their own infrastructure buildouts; Microsoft needs OpenAI to stay competitive with Google and Amazon in the AI services market. Everyone has skin in the game, and the game requires OpenAI to keep working.
But coherence is not the same as stability. Closed-loop systems of this kind concentrate failure modes in a small number of nodes. If compute availability tightens further, if a major partner renegotiates terms, if regulatory intervention reshapes the arrangement — the cascade effects are not obvious to model because the system has never been tested at this scale. Deloitte's counsel to corporate leaders maps directly onto the OpenAI situation: treat these partnerships as critical infrastructure and potential single points of failure. Map the dependencies. Build contingency plans. The question for the industry is whether anyone is actually doing that.
OpenAI at $730 billion is too important to fail — which is precisely the condition that gives its partners leverage they may not want to exercise publicly, and its regulators a problem that existing law was not built to solve.