Meta Agrees Multibillion-Dollar Chip Deal With AMD — What It Means for the AI Arms Race

Meta will deploy 6 gigawatts of AMD processors across its AI data centres in a deal worth up to $100 billion, gaining the option to buy 10 per cent of the chipmaker. Days after committing to millions of Nvidia GPUs, Zuckerberg is building a multi-vendor strategy that reshapes the economics of AI infrastructure.

QUICK ANSWER

Meta has struck a multiyear deal with AMD to deploy up to 6 gigawatts of customised AI chips across its data centre network, in a partnership estimated at up to $100 billion over five years. AMD has issued Meta a performance-based warrant for up to 160 million shares — roughly 10 per cent of the company — at an exercise price of $0.01, vesting in tranches tied to GPU shipment milestones and share price thresholds reaching $600. The deal, announced on 24 February 2026, comes days after Meta committed to deploying millions of Nvidia’s Blackwell and Rubin GPUs in a separate multiyear partnership. Meta’s head of infrastructure said the company needs all three sources of silicon — Nvidia, AMD and its own custom chips — to support its $135 billion capital expenditure plan for 2026.

How big is the Meta-AMD deal?

The numbers are staggering even by Big Tech standards. Meta will purchase custom AMD Instinct GPUs based on the MI450 architecture and next-generation EPYC CPUs, enough to power 6 gigawatts of AI compute — equivalent to the electricity consumed by 5 million US households in a year. AMD CEO Lisa Su told analysts that each gigawatt of compute is worth “double-digit billions,” with industry analysts at TechCrunch estimating the total at up to $100 billion over the life of the agreement. Shipments supporting the first gigawatt deployment are expected to begin in the second half of 2026.

The financial structure goes beyond a standard procurement deal. AMD has issued Meta a performance-based warrant for up to 160 million shares of common stock at an exercise price of $0.01 per share. The warrants vest in tranches as AMD hits GPU shipment milestones, with the final tranche tied to AMD’s share price reaching $600 — more than three times its closing price of $196.60 on the day before the announcement. If fully exercised, Meta would own approximately 10 per cent of AMD. The structure mirrors a deal AMD struck with OpenAI in October 2025, where the ChatGPT maker was offered a similar stake. AMD shares surged 14 per cent in premarket trading on the news.

Why is Meta diversifying away from Nvidia?

The deal makes explicit what many in the industry suspected: Meta is building redundancy into its AI supply chain. Just days before the AMD announcement, Meta committed to deploying millions of Nvidia’s latest Blackwell and Rubin GPUs in a separate multiyear partnership. When that Nvidia deal was announced, AMD shares actually fell — investors interpreted it as Nvidia crowding out the competition. The AMD deal reverses that narrative entirely.

Meta’s head of infrastructure, Santosh Janardhan, was blunt about the strategy. “We don’t believe that a single silicon solution will work for all of our workloads,” he said. “There’s a place for Nvidia, there’s a place for AMD, and there’s a place for our own custom silicon as well. We need all three.” The customised nature of the AMD chips — tailored specifically for Meta’s inference workloads — is a key differentiator. Analysts briefed on the deal noted that Nvidia is not offering comparable customisation. For AMD, this is a competitive wedge that could reshape how hyperscalers think about chip procurement.

What does this mean for the AI infrastructure arms race?

Meta’s 2026 capital expenditure guidance of $115 to $135 billion is nearly double its 2025 spending. It is not alone. The five largest US cloud and AI infrastructure providers — Microsoft, Alphabet, Amazon, Meta and Oracle — have collectively committed to spending more than $650 billion on capital expenditure in 2026, nearly doubling 2025 levels. The scale of spending is now measured in gigawatts, not server racks.

For AMD, the deal validates its position as the most credible alternative to Nvidia in the AI chip market. AMD’s data centre segment generated $5.38 billion in Q4 2025, up 39 per cent year-on-year, with full-year revenue reaching $34.64 billion. But Nvidia remains dominant: it is expected to report revenue growth of 68 per cent to $66 billion when it announces quarterly earnings this week. The Meta deal narrows the gap — not by matching Nvidia’s scale, but by offering something Nvidia has not: deep customisation and an equity alignment that ties AMD’s success directly to its biggest customers.

For Lisa Su, the deal is strategic positioning. “Meta has a lot of choices,” she told reporters. “I want to make sure that we are always a clear seat at the table when they think about what they need next.”

What are the implications for Europe?

The deal has significant implications for European competitiveness. As we reported in our 2026 outlook on AI, cyber and the new capital cycle remaking European markets, the most powerful investment theme in Europe is no longer apps or platforms — it is AI infrastructure. European companies like Siemens Energy, Schneider Electric and ABB are benefiting from the data centre boom as suppliers of power management, cooling systems and grid equipment. But Europe lacks the chip manufacturing capacity to compete directly.

The EU’s response has been to invest in the enabling layer. The European Commission committed €20 billion through its InvestAI initiative to establish AI gigafactories, each running on roughly 100,000 advanced chips. The EIB launched TechEU, deploying €70 billion in loans and equity to mobilise €250 billion in technology investment by 2027. As we examined in our coverage of EIB and European Commission financing for AI gigafactories, Europe is building compute capacity — but at a fraction of the scale that Meta alone is now committing to a single chip supplier.

The power dimension is equally telling. Meta’s AMD deployment requires 6 gigawatts — roughly the output of six nuclear power stations. Europe’s grid constraints, planning delays and energy costs mean that comparable deployments face structural barriers that do not exist in the United States. As we explored in our analysis of Europe’s sovereign AI strategy, the continent’s AI ambitions depend on securing both chips and power at scale. The Meta-AMD deal is a reminder of how far ahead the US is moving — and how fast.

Mark Zuckerberg said he expects “AMD to be an important partner for many years to come.” For Europe, the question is whether it can build partnerships of comparable ambition before the compute gap becomes permanent.

Frequently Asked Questions

How much is the Meta-AMD chip deal worth?

The deal is estimated at up to $100 billion over five years, covering 6 gigawatts of AMD AI compute infrastructure. AMD CEO Lisa Su said each gigawatt is worth “double-digit billions.” AMD has also issued Meta warrants for up to 160 million shares — roughly 10 per cent of the company — at $0.01 per share, vesting as GPU shipment milestones are met.

Why is Meta buying chips from both AMD and Nvidia?

Meta is building a multi-vendor AI chip strategy to avoid dependence on any single supplier. The company committed to millions of Nvidia GPUs in a separate deal days before the AMD announcement. Meta’s head of infrastructure said the company needs Nvidia, AMD and its own custom silicon to support different workloads across its $135 billion 2026 capital expenditure plan.

What does the Meta-AMD deal mean for the global AI chip market?

The deal validates AMD as the most credible alternative to Nvidia in AI infrastructure and introduces a new financial model — shares-for-chips — that aligns chipmaker and customer interests. It also underscores the scale of the AI infrastructure buildout: the five largest US cloud and AI providers have collectively committed to over $650 billion in capital expenditure in 2026, nearly doubling 2025 levels.