Meta inks $60‑$100 bn AI‑chip pact, offers up to 10% stake in AMD in 6 GW deal.

Meta’s first tranche of the agreement will arrive in the second half of 2026, delivering roughly one gigawatt of custom Instinct MI450‑class GPUs built expressly for the company’s massive language‑model training pipelines, according to a report from The Register. The deal calls for a total of six gigawatts of silicon over a five‑year horizon, a volume that dwarfs any prior AI‑chip procurement contract in the industry. Forbes notes that the MI450 GPUs will be “optimized specifically for Meta’s AI training and inference workloads,” suggesting a deep co‑design effort that goes beyond a simple off‑the‑shelf purchase. The scale of the commitment translates to a headline valuation of $60‑$100 billion, a range cited by Fortune and New Electronics, positioning the pact as the most financially ambitious chip supply deal ever announced.

The financial structure of the pact is equally unconventional. Meta will pay for the hardware with a combination of cash and equity, receiving an option to acquire up to a 10 percent stake in AMD at a price tied to the company’s market valuation at the time of exercise, as detailed by Fortune. This “chips‑for‑stock” mechanism mirrors a similar arrangement AMD struck with OpenAI last fall, a parallel highlighted by The Register. By tying a portion of the payment to equity, Meta secures a long‑term upside on AMD’s growth while AMD locks in a guaranteed buyer for a multi‑gigawatt supply of its next‑generation AI accelerators.

From a strategic perspective, the partnership gives Meta a direct line to the silicon that will power its next wave of AI services, from advanced recommendation engines to large‑scale generative models. Reuters emphasizes that AMD now counts Meta among its “next big AI chip customers,” a status that could accelerate the chipmaker’s roadmap for future Instinct architectures. The agreement also signals Meta’s intent to internalize more of its AI stack, reducing reliance on external cloud providers and potentially lowering the cost per inference—a critical metric as the company scales its AI‑driven advertising and content‑moderation tools.

Analysts observing the deal note that the $60‑$100 billion price tag reflects not only the raw hardware cost but also the value of the equity kicker. New Electronics points out that the five‑year supply window aligns with AMD’s projected production capacity expansions, meaning the chips will be fabricated on the company’s most advanced process nodes, likely 5 nm or below. This timing dovetails with Meta’s roadmap to roll out new AI features across its family of apps beginning in 2027, a rollout that will demand unprecedented compute density. By locking in a dedicated supply of MI450 GPUs, Meta can avoid the market volatility that has plagued other firms scrambling for AI silicon in recent quarters.

The broader AI ecosystem is likely to feel the ripple effects. With Meta committing to a multi‑gigawatt purchase, rival chipmakers such as Nvidia and Intel may see a contraction in the pool of available capacity for other large customers, a dynamic hinted at by the Reuters coverage of AMD’s “second mega chip supply deal.” Meanwhile, Meta’s equity option could align the two companies’ fortunes, potentially encouraging joint R&D initiatives that push the performance envelope of future AI accelerators. If the partnership delivers on its promise, the six‑gigawatt deployment could set a new benchmark for corporate AI infrastructure, reshaping how tech giants source and finance the compute engines that underpin the next generation of intelligent services.