Meta Eyes Billions in Google AI Chips, Taking Aim at Nvidia
Meta is in advanced talks to pour billions into Google’s custom AI chips, a move that could upend the hardware landscape and directly challenge Nvidia’s stranglehold on the market.
The deal, first reported by The Information, outlines a massive strategic partnership: Meta would rent Google’s Tensor Processing Units (TPUs) via the cloud starting in 2026, before buying the silicon outright for its own data centres in 2027.
A Radical Pivot
This marks a historic shift for Google. After a decade of guarding its TPUs for internal use, powering Search, YouTube, and its Gemini AI models, Alphabet is poised to enter the merchant semiconductor market.
By selling directly to a rival tech giant, Google aims to break Nvidia’s 80% market share. Internal estimates suggest this strategy could eventually capture up to 10% of Nvidia’s annual revenue, a slice worth billions.
The Hunt for Alternatives
The alliance underscores the industry’s desperate hunt for alternatives to Nvidia’s pricey, supply-constrained GPUs. Bolstered by the success of its TPU-trained Gemini 3 model, Google is proving its hardware can handle massive scale.
However, hurdles remain. Meta would need to adapt its software stack, currently optimised for Nvidia’s CUDA platform, to fully embrace Google’s architecture. Neither company responded immediately to requests for comment.
