Meta is deep in discussions to secure a massive supply of Google’s custom AI chips for its future projects. The negotiations reportedly focus on a plan for Meta to rent Google’s Cloud Tensor Processing Units (TPUs) throughout 2026, before switching to buying the hardware directly in 2027.
This potential partnership marks a sharp pivot for both tech giants. Historically, Google kept its TPUs strictly for internal use, refusing to sell them to outsiders. Meanwhile, Meta has typically relied on a mix of standard processors and graphics units from various outside vendors. Now, Meta seems determined to diversify even further. Beyond the Google talks, the company is looking into processor designs from Rivos, showing a clear desire to stop relying on a single source for its computing needs.
The news immediately hit the stock market. Alphabet’s value climbed steeply, nearing a reported $4 trillion valuation, while Meta’s shares also rose. Conversely, Nvidia saw its stock dip as investors worried about the long-term consequences of major cloud companies moving away from Nvidia’s hardware. Google executives believe this deal could help them capture a significant slice of Nvidia’s data-center revenue, which currently tops $50 billion a quarter.
However, signing a deal is easier than delivering the goods. The market faces a serious bottleneck in manufacturing capacity. Data centers are already reporting shortages of essential parts, such as memory modules, and prices will likely keep rising through next year. Logistics chains are stretched thin, and every major tech company is racing to lock down long-term supplies.
Even if the deal goes through smoothly, uncertainty remains. Technology evolves fast; Google releases new TPUs annually, but Nvidia updates its designs just as quickly. By the time Meta receives its first bulk shipment, the competitive landscape might look completely different. This rapid pace of change explains why companies are scrambling to try different architectures—betting everything on one type of hardware carries too much risk.
