Meta, Facebook’s parent company, has entered into an agreement with Amazon to utilize Graviton chips for agentic artificial intelligence applications, marking another significant win for Amazon’s expanding custom silicon division.
According to Bloomberg, the agreement spans multiple years and is valued in the billions of dollars. The announcement comes just one day following Meta’s revelation that it plans to reduce its workforce by approximately 10%, eliminating roughly 8,000 positions, as technology companies industry-wide implement workforce reductions while simultaneously investing heavily in AI infrastructure.
Under the terms of the agreement, Meta will gain access to tens of millions of Graviton5 processor cores operating within AWS data centers, positioning Meta among the world’s largest Graviton customers, according to statements from both companies. The deal expands upon Meta’s current utilization of Amazon Bedrock, Amazon’s platform designed for AI models.
In a LinkedIn post, Amazon CEO Andy Jassy noted that agentic AI is “becoming almost as big a CPU story as a GPU story.” This observation highlights that while graphical processing units, predominantly manufactured by Nvidia, have dominated discussions around AI hardware, agentic systems require conventional central processing units to manage the reasoning and coordination occurring between operational steps.
The timing of this deal follows Intel’s announcement of strong quarterly results, featuring data center revenue growth of 22%, partially fueled by increased demand for CPUs designated for agentic AI workloads—the identical trend benefiting Amazon’s Graviton offerings. Intel’s stock price surged over 22%, while Amazon’s stock increased 3% following the Meta announcement.
Meta has adopted a diversified strategy, establishing agreements with both Nvidia and AMD, recently committing to utilize Google’s custom processors, while simultaneously developing proprietary in-house silicon in collaboration with Broadcom.
Santosh Janardhan, Meta’s head of infrastructure, stated in a release that “As we scale the infrastructure behind Meta’s AI ambitions, diversifying our compute sources is a strategic imperative.”
Amazon is positioning itself as a significant chipmaker in its own capacity. In his annual shareholder letter, CEO Andy Jassy revealed that Amazon’s custom silicon business generates over $20 billion annually in revenue, suggesting it’s “quite possible” that Amazon may sell racks of its chips to external parties in the future. Such a move would place Amazon in more direct competition with Nvidia.
The company’s chip customer base continues to expand. Anthropic made a commitment to operate its models on Amazon’s Trainium processors as part of a $25 billion expanded partnership revealed this week, while OpenAI agreed to utilize Trainium as part of a $100 billion cloud agreement finalized earlier this year.
This development demonstrates the shifting landscape of AI
infrastructure, where traditional CPU technology is regaining prominence alongside GPU solutions. The strategic partnership between Meta and Amazon underscores the growing recognition that diverse computing resources are essential for advancing AI capabilities, particularly in the emerging field of agentic AI systems.
The multibillion-dollar commitment represents a substantial
endorsement of Amazon’s chip development efforts and validates the company’s investment in custom silicon technology. For Meta, the arrangement provides the computational resources necessary to support its ambitious AI initiatives while maintaining flexibility through its multi-vendor approach to chip sourcing.