Key Highlights
- Meta partners with Amazon Web Services in a multibillion-dollar agreement spanning three to five years for Graviton CPU deployment.
- The partnership involves tens of millions of Graviton cores, with primary deployment across U.S. data centers.
- Premarket trading showed Meta shares climbing 0.6% while Amazon stock advanced 1.4% after the announcement.
- Amazon joins Meta’s diversified chip procurement strategy alongside Nvidia, Broadcom, and AMD.
- Meta ranks among the top five AWS Graviton enterprise clients following this agreement.
Meta has entered into a substantial multibillion-dollar partnership with Amazon Web Services, securing access to tens of millions of Graviton CPU cores to support its expanding AI agent infrastructure.
According to Nafea Bshara, Amazon VP and Annapurna Labs co-founder who leads AWS’s custom chip division, the arrangement extends between three and five years. The majority of these Graviton cores will operate from U.S.-based facilities.
Market reaction proved favorable, with Amazon shares rising 1.4% and Meta climbing 0.6% during premarket sessions.
This partnership centers on AWS Graviton central processing units rather than graphics processing units. While GPUs dominated recent AI development cycles, CPUs are experiencing renewed importance.
AI agents have catalyzed this CPU resurgence. These processors manage distinct operations and channel workloads to GPUs, creating a complementary relationship that serves diverse AI computing requirements.
CPUs prove essential during the post-training stage of large language models, when pretrained systems undergo refinement for targeted applications.
Meta selected Amazon’s Graviton5, built on 3-nanometer technology, based on its price-to-performance characteristics. “Meta has access to so many options from the supply side. But they chose Graviton5,” Bshara stated.
Expanding Meta’s Chip Vendor Portfolio
This Amazon partnership extends Meta’s existing relationships with Nvidia, Broadcom, AMD, and Arm Holdings. The social media company has emphasized maintaining flexibility across multiple chip platforms.
“No single chip architecture can efficiently serve every computational task,” Meta said in a statement tied to the announcement.
Initial rollout encompasses tens of millions of Graviton cores, with expansion capacity aligned to Meta’s evolving AI requirements.
AI Expansion Amid Organizational Restructuring
Meta’s AI initiatives continue accelerating. The company completed its acquisition of AI startup Manus for over $2 billion last December. Manus develops sophisticated AI agents designed for complex task execution, creating additional CPU infrastructure demands.
To offset AI infrastructure investments, Meta announced Thursday it would reduce its workforce by approximately 10%—affecting around 8,000 employees—with implementation scheduled for May.
Meta launched Muse Spark, its first new AI model in twelve months, earlier this month, signaling additional releases on the horizon.
The collaboration between Meta and AWS traces back to approximately 2016, though previous engagements primarily involved cloud services, the Bedrock platform, and GPU cluster arrangements. This agreement represents significant expansion into customized semiconductor solutions.
For AWS, securing Meta as a Graviton customer delivers substantial validation. Earlier this week, Amazon revealed a $5 billion commitment to Anthropic, which similarly incorporates tens of millions of Graviton CPU cores.
AWS began developing proprietary chips before 2018, launching the original Graviton on Arm architecture.
Bshara confirmed Meta’s position among AWS’s five largest Graviton enterprise customers following this partnership.
