Over the past few years, Nvidia has cornered the AI hardware market. But the competition has not stood still. According to internal sources, Google has changed her strategy: instead of keeping them for herself, is going to trade its Google Cloud Tensor Processing Units (TPU) AI chips. And your first client is going to be Meta.
Los Google Cloud TPU are artificial intelligence accelerator chips that are optimized for training and inference of large AI models.
As revealed The Informationvia Tom’s Hardware, Google and Meta are finalizing a “billion dollar” deal through which Mark Zuckerberg’s company will rent and buy Google Clouds TPUs to train and use with its AI.
Google partners with one of its biggest “enemies”, Meta
Until now, these customized chips were used by Google only for itself, as a competitive advantage, so it is surprising that it now decides to trade with them, and also with one of its biggest competitors. AI does not understand logic.
The reason is, of course, Nvidia. Jensen Huang’s company is doing big business in AI data centers. A few days ago it presented the results of the last quarter, where it has earned 60% more than last year. With the data center business it earned no less than 51,000 million dollars, in just three months.
Google wants a piece of that pie, and is willing to do whatever it takes. Even sell AI chips to your most direct competitionas is Meta.
According to the sources consulted, with this change in strategy Google hopes to take 10% of the data center market from Nvidia.
The agreement apparently implies that Meta will lease thousands of Google Cloud TPUs from Google in 2026, and purchase them in 2027.
The Stock Market has welcomed the leak. The shares of both companies have risen, and Google (Alphabet) could exceed 4 billion in trading today.
All this, when No one guarantees that the memories and semiconductors that all AI companies need can be manufactured.
The ones who are going to pay dearly are us, the ordinary users. DDR5 RAM already costs twice as much, and next year the price of RAM, SSDs, graphics cards, and other components that use memory chips will skyrocket.