In a surprising move that significantly reshaped the AI hardware market, Nvidia, the world’s most valuable chipmaker and the key enabler of the global AI boom, has made an unusual public statement on Tuesday, defending its market leadership after reports that one of its biggest customers, Meta Platforms is in advanced talks to invest billions in Google’s competing AI chips.
The information was so significant that it changed the whole market sentiment drastically and led to the drop of Nvidia’s stock value. In fact, the company lost around $250 billion of market capitalization – which is one of the largest single-day declines in the history of U.S. corporate – in one day.
The company issued an uncommon and rare in the industry defensive statement after The Information published a report revealing that Google has been very assertive in pitching its tensor-processing units (TPUs) to large consumers of technology, which comprises Meta, as well as some major banks. So, if the deal was done, a significant part of the expenditure on AI hardware accelerators would be diverted to Google, leaving Meta to go with Nvidia, a company that holds the bulk of the AI accelerator chip market.
Nvidia breaks from tradition with public statement
Typically, the company tends not to interfere by making statements and only lets its achievements and releases speak for themselves.
As a response to these allegations, Nvidia accordingly interacted with the public via a press release which aimed at reassuring investors and patrons.
Nvidia communicated that it was very happy for Google’s success, and they would like to point out they are still a supplier to Google. The text went from praise to positioning very quickly: – “Nvidia is far ahead of the next closest competitor – it’s the only platform that is capable of executing every AI model, and it does so everywhere there is the need for computation.”
The defensive position underscored how big the “attack” was that Google’s initiative constituted. In fact, for quite some time, TPUs have been considered as tools specifically designed and optimized for Google only use but not as a generally accepted standard across the industry. However, that understanding is now changing.
The point at which one could say that TPUs had become a real alternative to GPUs for the fact that Gemini 3 is far better than other models seems to be Google’s Gemini 3, the tech giant’s newest AI model – and the first blockbuster system trained entirely on TPUs rather than Nvidia GPUs. It is the model that has received a lot of positive feedback from AI leaders including the CEO of Salesforce, Marc Benioff, who in a public announcement said he was changing from ChatGPT after three years, and he named Gemini 3 as the best.
Industry analysts argue that the performance of Gemini 3 indicates that TPUs may not only challenge Nvidia’s GPUs technologically but also from the point of view of cost-saving, which is of great importance as companies are spending tens of billions on building AI data centers.
Per The Information report, Meta might start leasing TPU resources from Google Cloud as early as 2026, with the probable buying of actual TPMU machines for Meta data centers starting in 2027. Meta is already thought to be willing to invest as much as $72 billion in AI technology this year, thus it is one of the biggest customers of Nvidia. Just by taking a small part of that budget would mean a huge change in the AI chip industry for Nvidia.
Wall Street reacts to changing competitive dynamics
Investors didn’t waste time getting a grasp of what was at stake. After the report was published, the stock price of Alphabet went up by 4%. This, in turn, brought the company close to a $4 trillion valuation. At the same time, Broadcom, a company that makes processors for Google’s AI, was up 11%.
On the other hand, Nvidia, whose market cap reached $4 trillion for the first time in history just a few days ago, fell sharply as investors reevaluated their expectations of Nvidia’s continuous dominance. The decline wiped off close to $250 billion of the company’s market value.
Google’s reaction to the market turmoil was quite moderate. The company said: “The demand for our custom TPUs as well as Nvidia GPUs is increasing rapidly; we have no intention to discontinue one in favor of the other and, therefore, will continue supporting both.”
A seismic shift in the AI hardware arms race
For the most part of the last ten years, Nvidia has been able to retain almost complete control of the AI accelerator market which was the main contributor to the rapid growth of generative AI systems such as OpenAI’s ChatGPT and Meta’s Llama models. However, as the computing power needed for AI grows at an exponential rate, major tech companies are also looking for alternative ways of supply in order to lower their dependence on a single supplier and be able to cut costs in the face of the GPU shortage that is reaching extreme levels simultaneously.
In the event that Meta relocates a considerable portion of its work that is now running on Nvidia hardware to Google TPUs, this would be the first significant realignment in the AI chip market which consequently could lead to a wider industry acceptance of the rival platforms.
The analysts foretell that the coming 12 to 18 months may be pivotal in redefining the global AI infrastructure landscape and consequently deciding whether Nvidia will still be the leader or the AI race will become even more competitive than it has ever been.