So here's what happens when a chipmaker lands the kind of deal that makes investors do a double-take: Advanced Micro Devices Inc. (AMD) shares rocketed up more than 10% in Tuesday's premarket session. The catalyst? A massive, multi-year partnership with Meta Platforms Inc. (META) that's all about building out artificial intelligence infrastructure at a scale that's frankly hard to comprehend.
The deal centers on deploying 6 gigawatts of AI compute capacity. To put that in perspective, that's enough power to run a small city. Meta's basically saying it needs enough computing muscle from AMD to train and run AI models that could one day power what CEO Mark Zuckerberg calls "personal superintelligence." For AMD, this isn't just another customer win—it's being positioned as a primary beneficiary of one of the biggest infrastructure builds in tech.
Building the AI Engine, One Generation at a Time
The agreement isn't for a one-time purchase. It's structured around multiple generations of AMD's Instinct GPUs, the company's specialized chips designed for AI workloads. Shipments for the initial one-gigawatt phase are scheduled to begin in the second half of 2026.
These won't be off-the-shelf components either. The systems will feature a custom AMD Instinct GPU based on the upcoming MI450 architecture, specifically optimized for whatever unique AI tasks Meta is cooking up. They'll be paired with AMD's 6th generation EPYC server CPUs, codenamed "Venice." This suggests the partnership goes beyond just selling hardware—it's about deep technical integration to make everything work seamlessly together.
Why This Partnership Matters
"We are proud to expand our strategic partnership with Meta as they push the boundaries of AI at unprecedented scale," said AMD CEO Lisa Su in a statement. That "unprecedented scale" part is key. Meta is in an arms race with other tech giants to build the most powerful AI systems, and it's betting heavily on AMD's technology to get there.
The partnership will utilize the AMD Helios rack-scale architecture, which was developed through the industry-wide Open Compute Project. This suggests Meta values the flexibility and efficiency of this open-standard approach, rather than being locked into a single vendor's proprietary system.













