Nvidia and OpenAI: a $20 billion alliance (and it may not stop there)

Imagine selling shovels during a gold rush, then investing your takings in the biggest mine next door. That is exactly what Nvidia is doing now. According to sources, Jensen Huang's company is close to putting an astronomical $20 billion into OpenAI. In today's industry, it looks like an attempt to build a perpetual motion machine: Nvidia sells chips to OpenAI, OpenAI trains models on them, and then Nvidia returns part of its profit to the company as investment so that OpenAI can buy even more chips.

This funding round looks like an arms race in which the stakes passed the bounds of reason long ago. A couple of years ago, $20 billion would have seemed like the annual budget of a small country. Today it is merely the entry ticket to the top league of artificial intelligence.

OpenAI needs money more than ever. Developing GPT-5 and training the Sora video generator demand not just talent but endless farms of H100 GPUs and the upcoming Blackwell.

Why is Nvidia making this move now? The answer lies in Sam Altman's ambitions. The head of OpenAI has repeatedly hinted that he wants to reduce dependence on outside hardware suppliers, and has even considered building his own network of chip fabs. For Nvidia, that prospect is a direct existential risk. By pouring billions into OpenAI, the "greens" are not just buying a stake in a business; they are buying loyalty. It is soft power, meant to convince Altman that building his own fabs is slow and expensive when ready-made solutions are at hand from the best friend in the leather jacket.

Microsoft should not be forgotten either. For a long time, Redmond was OpenAI's only major patron. Now the shareholder structure is turning into a real "Avengers" club of the tech world. If Apple and Nvidia join the round, we get a unique alliance that would control practically every layer of the stack: from hardware and operating systems to end-user consumer services.

For smaller startups, this is bad news. The barrier to entry is now measured not by code quality but by access to capital on a scale that even many public corporations cannot reach.

Of course, regulators loom behind the scenes of this deal. In the US and Europe, antitrust authorities have already begun asking uncomfortable questions about ties between chipmakers and model developers. A $20 billion investment is not just financial support; it creates a vertically integrated giant that could strangle competition in the cradle. But Silicon Valley seems to be following a simple rule right now: build the monopoly first, and deal with the lawyers later.

The bottom line: Nvidia is finally moving from "hardware supplier" to "architect of the industry." Can anyone challenge the OpenAI-Nvidia tandem under these conditions, or are we watching the final shaping of the AI market for decades to come?