NVIDIA, Microsoft and Amazon are circling OpenAI with what could be one of the biggest private funding packages in tech history, and the headline number is almost surreal: up to $60 billion. For an AI lab that was a niche nonprofit less than a decade ago, the idea that three of the world’s most valuable companies may collectively write cheques on that scale says a lot about where the power in the AI economy is concentrating.

The rough contours of the talks are starting to come into focus. NVIDIA, already the indispensable supplier of the GPUs that train and run OpenAI’s models, is said to be considering as much as $30 billion on its own. Microsoft, OpenAI’s longest‑standing partner and current 27% shareholder, is reportedly looking at a smaller top‑up, under $10 billion, after pouring tens of billions into the company and its infrastructure over the past few years. Amazon is the wild card: a new investor that could still come in with more than $20 billion, depending in part on side deals over cloud capacity and how OpenAI’s products, like enterprise ChatGPT, might be sold through Amazon’s channels.

None of this is signed yet; the key phrase is “in talks”. OpenAI is described as being close to receiving term sheets, which would move the discussions from speculative to concrete commitments, but the numbers and mix of investors can still shift. NVIDIA, Amazon, Microsoft and OpenAI have all declined to comment so far, and Reuters notes it has not independently verified the details, which underlines how much of this is still based on briefings from people close to the negotiations.

Even if you cut the headline figures, the direction of travel is obvious: OpenAI is trying to assemble a war chest on a scale that looks more like a sovereign wealth fund than a startup round. Reports around the same time say SoftBank is also in talks to add up to $30 billion more, on top of an existing stake of roughly 11%, as part of a broader push that could bring total new funding closer to $100 billion and value OpenAI in the $750–$830 billion range. For context, that kind of valuation would put a privately held AI lab in the same conversation as the world’s largest publicly traded tech companies, despite having a far shorter operating history and a business still in hyper‑growth mode.

Why does OpenAI need this much money? The core reason is brutally simple: compute. Training and running models at GPT-5‑class scale, plus image, video and multi‑modal systems, requires staggering amounts of hardware, power and engineering talent. OpenAI has already disclosed compute capacity in the ballpark of 1.9 gigawatts, and that needs to keep rising if it wants to stay ahead of Google, Anthropic and a wave of open‑source models that are getting better, faster and cheaper. Each new generation of model is more expensive to train than the last, yet users have been conditioned to expect chatbots and copilots that feel instant and always‑on, which pushes up ongoing inference costs as well.

From NVIDIA’s perspective, the logic of a giant equity cheque is straightforward. It already dominates the AI chip market, and OpenAI is both a marquee customer and a public proof‑point for what NVIDIA’s hardware can do at scale. By tying itself even more closely to OpenAI’s future, NVIDIA isn’t just selling GPUs; it’s effectively buying into the downstream value that those chips unlock in software and services. That kind of vertical alignment makes sense if you believe this is still the early innings of generative AI, and that the real monetisation — from agents, enterprise automation, and entirely new app categories — is still ahead.

For Microsoft, the calculus is a bit different. It already has a deep, multi‑layered partnership that gives it exclusive rights to resell OpenAI models on Azure and to embed them across Windows, Office, GitHub and more. Adding a single‑digit‑billion cheque on top of its existing stake is less about changing the ownership structure and more about keeping a seat at the table while other giants pile in. Microsoft also has to balance its own AI investments — in in‑house models, datacenters and software — with regulatory scrutiny over how intertwined it can become with a key supplier like OpenAI.

Amazon’s angle is arguably the most interesting. Unlike Microsoft, which already enjoys exclusivity on certain OpenAI services, Amazon runs its own Bedrock platform, backs Anthropic, and is trying to convince enterprises that AWS is the most neutral, flexible place to build AI workloads. A big cheque into OpenAI would give Amazon a share in a rival that is currently more tightly coupled to a competing cloud, but the reported negotiations suggest Amazon wants something tangible in return: more OpenAI workloads on AWS and a clearer commercial path to selling OpenAI‑powered products. If that happens, OpenAI would no longer be as closely identified with a single cloud platform, which could appeal to large customers who worry about lock‑in.

Zoom out, and what’s emerging around OpenAI looks less like a classic startup cap table and more like a syndicate of strategic superpowers. Between NVIDIA, Microsoft, Amazon and possibly SoftBank, you have the world’s leading AI chipmaker, two of the biggest cloud platforms and a financial conglomerate that has made “all‑in on AI” its new mantra. If the round reaches the mooted $100 billion, OpenAI will be a cornerstone asset simultaneously tied into their supply chains, cloud platforms and investment theses.

That raises uncomfortable questions about the concentration of power. A handful of companies already control most of the infrastructure underpinning modern AI — from chips and datacenters to the largest foundation models. Layering gigantic equity stakes on top of that, especially if they come with preferred access to future models or influence over commercial strategy, could make it even harder for smaller firms or open‑source projects to compete. Regulators in the US and Europe are already probing big‑tech AI deals; any move that deepens cross‑ownership between a major model provider and its primary cloud and chip partners is likely to get a long, hard look.

On the other hand, it’s not obvious that anyone else could realistically write cheques of this size right now. Training frontier models has become so capital‑intensive that only a short list of players — hyperscalers, sovereign wealth funds, and firms like SoftBank — can fund them at the pace the industry is moving. OpenAI sits in the middle of that, needing enough money to keep pushing the frontier while arguing that its work will spin off broader economic benefits in productivity, new tools and new companies.

For customers and developers building on OpenAI, the immediate signal is that the company is unlikely to run out of runway any time soon if even a fraction of this funding materialises. A bigger balance sheet can mean more datacenters, better uptime, faster models and potentially new pricing tiers, although it can also entrench the idea that the “real” AI happens on a few closed, very expensive platforms. For competitors, it’s a reminder that keeping up at the very high end of the model race is no longer just about research talent — it’s about assembling coalitions of capital, clouds and chips that look more like global infrastructure projects than traditional software startups.

All of this, remember, is playing out while OpenAI is still evolving its own governance structure and public story about safety, openness and profit caps. The company began life as a capped‑profit entity designed to keep commercial incentives in check, even as it took outside money; layering tens of billions more from the world’s most profit‑driven firms on top of that will only intensify questions about how those caps work in practice and who ultimately steers the ship. If the talks with NVIDIA, Microsoft, Amazon and SoftBank progress to signed term sheets, the next phase of the AI era will be shaped not just by the capabilities of the models, but by this dense web of financial and strategic ties that now surround them.