Meta has entered a multibillion-dollar, multi-year agreement with Amazon to use custom-designed processors for its expanding artificial intelligence operations, signalling a deepening shift towards AI infrastructure.
The deal gives Meta access to Amazon Web Services Graviton5 chips at scale, with deployment expected to begin with tens of millions of processing cores and grow over time. The agreement was confirmed by Nafea Bshara, vice-president at Amazon and co-founder of its Annapurna Labs chip division.
AI investment drives strategic shift
The partnership comes as Meta accelerates long-term investment in artificial intelligence. The company plans to spend between $115 billion and $135 billion by 2026 on data centres, advanced chips and AI talent.
To support this expansion, Meta is restructuring its workforce. The company has announced plans to cut around 8,000 jobs from May 2026 and freeze hiring across 6,000 open roles, reflecting a broader shift in resources towards infrastructure capable of supporting large-scale AI systems.
According to Meta’s Head of Infrastructure, Santosh Janardhan, diversifying compute resources has become a strategic priority.
Why Graviton chips matter
At the centre of the agreement is Amazon’s Graviton5 processor, a CPU built for high-performance, cloud-based workloads. Unlike GPUs, typically associated with AI model training and dominated by companies such as Nvidia, CPUs play a critical role in running and scaling deployed AI systems.
These workloads include real-time inference, search, code generation and multi-step AI agent orchestration, all of which require efficient, low-latency processing.
Graviton5 is built on advanced 3-nanometre architecture and features 192 cores alongside a significantly larger cache system. This enables faster data handling, improved bandwidth and reduced communication delays between processing units—key requirements for emerging “agentic AI” systems designed to operate continuously and autonomously.
Expanding a multi-partner chip strategy
The AWS agreement adds to Meta’s growing network of chip partnerships. The company already collaborates with Advanced Micro Devices, Nvidia and Arm Holdings to meet its computing needs.
By diversifying suppliers, Meta aims to reduce dependence on any single vendor while maintaining flexibility as global demand for AI computing power continues to rise.
Industry-wide shift towards AI infrastructure
Meta’s restructuring reflects a wider trend across the technology sector. Companies including Microsoft, Oracle and Block are also cutting costs in traditional areas while increasing investment in AI systems and infrastructure.
This shift highlights a fundamental change in how technology firms operate. Artificial intelligence is enabling leaner teams to deliver greater output, prompting organisations to rethink workforce structures while investing heavily in compute-intensive capabilities.
No comments:
Post a Comment