Meta Is Going Big on AMD — Billions per Gigawatt, 6 Gigawatts Planned

US Market & Sectors
1

This is not a small vendor agreement. This is one of the largest AI infrastructure commitments we’ve seen so far.

Meta Platforms Inc. is set to deploy 6 gigawatts of data center capacity powered by processors from Advanced Micro Devices Inc. (AMD). To put that into perspective, one gigawatt equals the output of a nuclear reactor — enough to power roughly 700,000 homes.

Now multiply that by six.

And here’s the scale that really stands out:

  • The deal is worth “double-digit billions of dollars per gigawatt”
  • Purchases will stretch over five years starting in the second half of 2026
  • Meta will receive warrants to buy 160 million AMD shares
  • Some of those warrants vest only if AMD’s stock reaches $600 (it closed at $196.60 on Monday)

What Meta Is Actually Buying

Meta is committing to:

  • AMD AI chips
  • AI servers and computing systems
  • Custom versions of AMD’s upcoming MI450 accelerator and future successors

These chips will largely support AI inference — the stage where trained AI models are deployed and used at scale.

Importantly, this isn’t an off-the-shelf relationship.

Meta will:

  • Influence chip design
  • Get customized silicon
  • Continue building its own in-house AI chips
  • Continue buying from Nvidia

As Meta’s infrastructure head Santosh Janardhan put it — at this scale, there’s room for all three.

Why This Matters

:one: A Major Win for AMD

AMD has been trying to close the gap with Nvidia in AI accelerators. This deal signals:

  • Hyperscalers are willing to diversify
  • AMD is competitive at large-scale AI workloads
  • Revenue visibility just improved significantly

Meta is already AMD’s second-largest customer. This deal deepens that dependency.

AMD reported:

  • $34.6B in sales last year
  • Expected 34% revenue growth this year

Even an additional $10B+ from Meta meaningfully accelerates AMD’s trajectory.

After rising 77% in 2025, AMD stock is down 8.2% so far this year — investor skepticism had been building around AI valuations. This partnership challenges that narrative.

:two: Meta Is Aggressively Front-Loading AI Capacity

CEO Mark Zuckerberg has made AI Meta’s top strategic priority.

Recently, he announced Meta Compute, aiming to build:

  • Tens of gigawatts this decade
  • Potentially hundreds of gigawatts over time

The company is deliberately spending early and heavily to secure long-term advantage.

This AMD deal is part of that strategy:

  • Secure supply
  • Influence hardware roadmap
  • Reduce over-reliance on a single vendor
  • Lock in long-term AI compute access

:three: AI Spending Is Not Slowing

Even as some investors worry about an AI investment bubble, this agreement signals:

  • Infrastructure buildout is accelerating
  • Capital commitments are getting larger
  • Hyperscalers are competing on compute scale

AMD shares jumped as much as 15% in early trading following the announcement. Meta shares rose 0.6%.

The Bigger Picture

This is about more than just chip purchases.

It’s about:

  • Vertical integration
  • Strategic supply control
  • Custom silicon influence
  • Long-term compute dominance

Meta is not just buying GPUs.
It is locking in compute capacity measured in nuclear-scale power equivalents.

And AMD just secured one of the most important AI customers in the world at scale.

Questions for the Community

  • Does this meaningfully shift the AI chip balance away from Nvidia?
  • Is AI infrastructure spending still in early innings — or nearing excess?
  • At what point does compute capacity become the moat?

This is not incremental spending.

This is infrastructure positioning for the next decade.