On February 24, 2026, Meta Platforms (the parent company of Facebook) and AMD simultaneously announced a large‑scale strategic cooperation: Meta plans to deploy up to 6 gigawatts (6GW) of AMD Instinct GPUs over multiple years for its next‑generation AI infrastructure. Coming just a week after a similarly large agreement with NVIDIA, the deal is seen as another major move in Meta's compute build‑out and a sign that the AI infrastructure arms race is intensifying.
According to a joint statement, this is a definitive multi‑year, multi‑generation partnership. The first 1GW deployment will begin in the second half of 2026, using custom Instinct GPUs based on the AMD MI450 architecture and the Helios rack‑level architecture co‑designed by AMD and Meta. Subsequent deployments will phase up to a total of 6GW based on Meta's needs. AMD CEO Lisa Su said on a conference call that the agreement "places AMD at the core of Meta's next‑generation AI infrastructure," and estimated the value per gigawatt at "double‑digit billions of dollars," with the total agreement potentially exceeding $100 billion.
As a key component of the agreement, AMD will issue 160 million common stock warrants to Meta, vesting in stages and potentially resulting in Meta owning about 10% of AMD. This "chips‑for‑warrants" structure is highly similar to AMD's prior collaboration with OpenAI and reflects a trend of AI giants and chipmakers tightly binding together to co‑build ecosystems.
Meta's AI compute hunger: 6GW is just the beginning
Meta's investment in AI has grown explosively in recent years. On February 18, 2026, Meta Platforms announced a separate multi‑generation, multi‑year strategic infrastructure partnership with NVIDIA. According to that joint statement, Meta will deploy millions of NVIDIA Blackwell and next‑generation Rubin GPUs across its global data centers, along with NVIDIA Grace CPUs and Spectrum‑X Ethernet networking solutions to support AI training, inference and long‑term core business needs. The AMD 6GW agreement means Meta has opened another important, diversified supply line for compute beyond NVIDIA.
What does 6GW mean? Industry estimates suggest 1GW of AI compute roughly corresponds to the peak power of hundreds of thousands of high‑end GPUs. A 6GW deployment would elevate Meta's AI infrastructure to a new scale, large enough to support training next‑generation Llama models and the massive AI workloads across Facebook, Instagram and WhatsApp—such as recommendation systems, content generation and metaverse rendering.
AMD emphasized that the GPUs delivered under the agreement are custom versions optimized for Meta's workloads and, combined with AMD's Helios rack architecture, offer significant advantages in energy efficiency and large‑scale deployment. This is also the first time AMD's Instinct GPU series will be adopted at scale in the core of Meta's AI infrastructure.
AMD seizes momentum: a strategic turning point to challenge NVIDIA
For AMD, the agreement is highly significant. AMD has long been overshadowed by NVIDIA in the AI accelerator market, with a market share below 10%. Deep integration with Meta not only brings massive orders but also provides an excellent opportunity to validate and optimize the Instinct platform in production environments.
AMD CEO Lisa Su said in the statement: "This collaboration will accelerate AMD's growth in AI infrastructure and provide a clear roadmap for our future multi‑generation products." The market reacted positively, with AMD's share price rising sharply on the day the agreement was announced, reflecting increased investor confidence in its AI business prospects.
Notably, within a week Meta concluded large deals with both NVIDIA and AMD, creating a "double‑insurance" compute strategy. This reflects both Meta's voracious demand for AI infrastructure and its strategic intent not to rely entirely on a single supplier.
Industry impact: AI compute competition enters the "gigawatt era"
The 6GW agreement pushes the AI compute competition to a new level. Global AI leaders are expanding data centers at unprecedented speed:
- Meta plans to deploy GPUs at a scale of millions of units in 2026;
- Companies such as OpenAI, Anthropic and xAI are also building superclusters at the million‑GPU scale;
- Power, chips and cooling systems have become among the most constrained production resources.
The emergence of 6GW deals signals that the measurement unit for AI infrastructure has shifted from "tens of thousands of cards" to "gigawatts." This not only drives demand across the global semiconductor and energy markets but also intensifies supply‑chain tightness and geopolitical competition.
For Chinese AI companies, the event also offers lessons. As compute becomes a core competitive advantage, diversifying supply chains, developing customized chips and building ecosystem partnerships to reduce dependence on a single supplier have become key issues. Domestic AI chip vendors such as Huawei's Ascend and Cambricon may draw insights from the Meta–AMD cooperation model.
Conclusion: a gigawatt‑level signal in the AI arms race
The Meta–AMD 6GW agreement is more than a commercial transaction; it is a strong signal that the AI‑era compute arms race has entered a new phase. It shows that leading AI companies are locking in core resources for years ahead in units of tens to hundreds of billions of dollars.
In the fast‑evolving technology landscape of 2026, whoever controls sufficient compute will hold the keys to the next generation of intelligence. The deep cooperation between Meta and AMD may be only the latest chapter in a much larger contest.
▶️ Like this article? Creating it takes hard work!
Subscribe to our YouTube channel:
AI Rosetta
Your likes, subscriptions and shares are our greatest support! 🙏
For more AI news, visit: https://www.airosetta.com