AI Pulse by Inblix

AMD inks 6-gigawatt GPU deal with OpenAI — and a 160M share kicker

OpenAI Blog · Jul 12, 2026 · 2 min read · Read original article →

Curated by the Inblix editorial team


Featured image for article: AMD inks 6-gigawatt GPU deal with OpenAI — and a 160M share kicker

The GPU wars just got a lot more interesting. AMD and OpenAI announced a multi-year, multi-generation agreement that will see the chipmaker deploy 6 gigawatts of its Instinct GPUs to power OpenAI’s infrastructure. That’s an eye-popping number — enough wattage to dwarf most individual data center campuses — and it starts with a 1-gigawatt slug of MI450 series accelerators in the second half of 2026. The deal formalizes a relationship that’s been simmering since the MI300X days, but the scale here signals something new: OpenAI is dead serious about diversifying its compute supply chain beyond Nvidia.

And there’s a financial twist. AMD issued OpenAI a warrant for up to 160 million shares, structured to vest as specific milestones are hit. The first tranche triggers with that initial gigawatt deployment. More shares unlock as purchases scale toward the full 6 gigawatts, and — critically — vesting is also tied to AMD hitting certain share-price targets and OpenAI clearing technical and commercial hurdles. CFO Jean Hu framed it bluntly, calling the deal “expected to deliver tens of billions of dollars in revenue” while being highly accretive to non-GAAP earnings per share. That’s not the language you use for a toe-in-the-water experiment.

Lisa Su called it a “true win-win enabling the world’s most ambitious AI buildout.” Sam Altman said the partnership is “a major step in building the compute capacity needed to realize AI’s full potential.” Greg Brockman chimed in with the obligatory “across every layer of the stack” nod. But the real story here is the deepening alignment between the hottest AI lab and the perennial underdog in AI silicon. AMD gets a marquee customer that could single-handedly reshape its data center GPU narrative. OpenAI gets leverage — and a hedge against the brutal lead times and pricing power Nvidia has enjoyed.

What’s less clear is whether the MI450 series will be genuinely competitive on AI training workloads by late 2026, or if this capacity is primarily aimed at inference and less demanding scale-out tasks. AMD’s software ecosystem has been the perennial question mark. A deal this large suggests OpenAI has seen enough in the MI350X generation to bet big on the roadmap. If AMD executes, this partnership becomes a gravitational force in the AI infrastructure market. If not, it’s a very expensive insurance policy.

💡 Key Takeaways

  1. OpenAI is making its largest-yet move to diversify away from Nvidia, locking in 6 gigawatts of AMD GPU capacity starting with MI450 deployments in H2 2026.
  2. The deal includes a warrant for up to 160 million AMD shares that vests in tranches — but only if AMD hits share-price targets and OpenAI meets technical milestones.
  3. AMD CFO Jean Hu explicitly pegged the revenue opportunity in the 'tens of billions,' signaling this is not a pilot program but a core strategic bet for the company.
  4. Success hinges on whether AMD’s MI450 series and its software stack can truly compete on training workloads by late 2026, not just inference.

Keep reading: See related articles below for more coverage on this topic.

Get smarter about AI

The sharpest AI news, curated daily. Delivered free to your inbox.

Learn more

Glossary terms

← Back to all articles