Meta just wrote the largest check in AI infrastructure history. On Tuesday, the company announced it would purchase up to $100 billion worth of AMD chips over five years, enough to power roughly six gigawatts of data center capacity. As part of the deal, Meta gains the option to acquire up to 160 million AMD shares - approximately 10% of the chipmaker’s total stock.
The deal marks a dramatic shift in the AI chip market, which Nvidia has dominated with over 80% market share. More importantly, it signals what Meta is actually building: what CEO Mark Zuckerberg calls “personal superintelligence.”
The Numbers
Meta will begin receiving AMD’s latest MI450 GPUs and 6th Gen EPYC “Venice” CPUs in the second half of 2026. The six-gigawatt commitment represents a staggering amount of computing power - for context, that’s roughly equivalent to the electricity consumption of a city of 4 million people dedicated solely to AI inference.
The deal structure ties Meta’s investment directly to AMD’s success. If AMD hits certain performance and delivery milestones, Meta can exercise options to acquire 160 million shares. At Tuesday’s closing price of $213.84 (up 8.77% on the news), that option is worth roughly $34 billion.
This comes on top of Meta’s existing Nvidia relationship. Just last week, Meta expanded its agreement with Nvidia to purchase millions of AI chips including standalone Grace CPUs. Meta isn’t abandoning Nvidia - it’s building redundancy.
Breaking the Nvidia Lock
The timing isn’t coincidental. Nvidia’s Blackwell GPUs have been on backorder for months, and the next-generation Rubin chips only recently entered production. Every major AI company faces the same problem: you can’t build AI infrastructure if you can’t get chips.
AMD’s AI accelerator market share is projected to climb from 9% in 2025 to over 15% by year-end 2026. Meta’s deal accelerates that trajectory considerably. More importantly, it establishes AMD as a credible alternative for the largest AI deployments in the world.
For AMD CEO Lisa Su, this validates years of investment in AI hardware. AMD has been the perpetual underdog to Nvidia in GPUs, but this deal puts them at the center of one of the most ambitious AI projects ever attempted.
What Meta Is Actually Building
Zuckerberg has been explicit about Meta’s goal: “personal superintelligence” - AI that understands your context, sees what you see, hears what you hear, and helps you achieve your personal goals. Not a general chatbot, but something integrated into every aspect of your life through glasses, earbuds, and eventually neural interfaces.
“This is going to be a big year for delivering personal superintelligence, accelerating our business, building infrastructure for the future, and shaping how our company will work going forward,” Zuckerberg said in January.
Meta’s vision document describes AI that helps you “create what you want to see in the world, experience any adventure, be a better friend to those you care about, and grow to become the person you aspire to be.” It’s an extraordinarily ambitious - and potentially invasive - vision of AI integrated into daily life.
To build this, Meta has pledged at least $600 billion in U.S. data centers and AI infrastructure over the coming years, including $135 billion in 2026 alone. The AMD deal is the first concrete step in executing that plan with diversified chip supply.
Who Wins, Who Loses
Winners:
- AMD gains validation as a Nvidia alternative and a massive guaranteed revenue stream. The stock option gives Meta skin in AMD’s success while aligning incentives.
- Meta reduces single-vendor risk and gains leverage in future Nvidia negotiations. If AMD delivers, Meta has a reliable alternative supplier.
- Other AI companies benefit from a more competitive chip market. More suppliers means better pricing and availability.
Losers:
- Nvidia loses its near-monopoly position, though remains dominant. More competition means lower margins and less leverage over customers.
- Meta employees - the company announced it’s slashing equity compensation for the second straight year to fund AI infrastructure. The money for $100 billion in chips comes from somewhere.
- Privacy advocates will have concerns about Meta’s “personal superintelligence” vision, which requires extensive personal data collection to function as described.
The Real Play
Meta is making a bet that the future of computing isn’t phones or PCs - it’s AI agents that know you intimately and operate continuously in your life. To build that, they need massive inference capacity deployed globally, which requires breaking Nvidia’s supply constraints.
The AMD deal is about resilience as much as cost. When you’re spending $600 billion on AI infrastructure, you can’t afford to depend on a single supplier that’s perpetually sold out. Meta is building optionality.
Whether “personal superintelligence” is something people actually want - AI that sees everything you see and knows everything about you - remains an open question. But Meta is now positioned to build it at scale, with or without Nvidia’s cooperation.