Nvidia was set to pump up to $100 billion into OpenAI, but is now backpedaling. A simpler investment of $30 billion will take its place.
Nvidia will not be investing $100 billion in OpenAI as previously announced. The AI chip specialist is still setting money aside but is capping it at $30 billion, according to the Financial Times.
Optimism in 2025
The two parties announced the $100 billion agreement with great enthusiasm in September of last year. OpenAI and Nvidia were to jointly deploy at least 10 gigawatts of Nvidia systems. The agreement initially concerned Nvidia’s new Vera Rubin platform, but was intended to span several years and include new generations of hardware.
September 2025 was the month of major contracts: Oracle previously signed a $300 billion deal with OpenAI. Even then, economists noted that the financial picture behind these billion-dollar investments didn’t quite add up. Despite the hype and optimism surrounding AI, the technology currently does not generate enough revenue to make such investments profitable within a realistic timeframe. The agreements relied on the hope that revenue and profit would follow in the near future.
Caution in 2026
Investors are now starting to behave a bit more cautiously, and Nvidia is playing it safe. The initial $100 billion agreement was never finalized, and that is no longer going to happen. Furthermore, the deal was somewhat complex and tied to infrastructure.
read also
Nvidia introduces Rubin and Vera: 5 times more inference than Blackwell and custom CPU
The new formula is simpler: Nvidia will simply invest $30 billion in OpenAI in exchange for shares. The funds will still be used to finance new HPC clusters. These will be powered, at least for a very large part, by Nvidia chips, as originally planned. In this way, Nvidia is indirectly investing in the purchase of its own chips. The smaller agreement is said to be nearly finalized.
This smaller (but still massive) agreement is more manageable than the previous one. Moreover, there is no reason to assume this is the final investment. Nvidia and OpenAI can further deepen their collaboration over time if it makes sense. However, Nvidia is committing itself less in the medium term with this approach.
More money
OpenAI is also receiving funds from other sources. Softbank is on the verge of finalizing a $30 billion investment, and Amazon is looking at a partnership that could be worth up to $50 billion. OpenAI needs the money: the company wants to spend about $600 billion on hardware by 2030. Revenue alone won’t get the company there, let alone profit. OpenAI only recently reached a revenue of $20 billion for the first time.
Based on all these investments, the fear among major tech companies of an AI bubble appears minimal. Based on orders placed, the fear of such a bubble bursting indeed seems misplaced. On a three-year horizon, investments in AI infrastructure look solid. Furthermore, they have already been partially paid for.
On the other hand, AI, however useful, has very high financial expectations to meet. Whether it will succeed in a timeframe that satisfies less rational stock market investors remains to be seen. The fact that Nvidia and OpenAI cannot finalize an investment announced with such enthusiasm as planned is, in any case, significant.