Meta has secured a $60 billion multi-year AI chip deal with Advanced Micro Devices, locking in large-scale supply of MI300 data center accelerators as competition for artificial intelligence compute capacity intensifies worldwide.
It positions Meta to scale large language model training and AI deployment across its hyperscale data centers amid ongoing constraints in advanced semiconductor supply.
Demand for AI accelerators has surged over the past two years as generative AI systems require increasingly complex training clusters. Securing long-term chip capacity has become a strategic priority for major technology firms.
AI Infrastructure Arms Race Accelerates
Meta has placed artificial intelligence at the center of its growth strategy, expanding generative AI tools, recommendation systems, and advertising optimization engines across its platforms.
Training frontier AI models requires massive parallel processing and high-bandwidth memory.
According to the Stanford AI Index, compute requirements for leading AI systems have risen sharply, driving record capital expenditures across hyperscale cloud operators.
Meta previously guided annual capital expenditures between $35 billion and $40 billion, largely tied to AI infrastructure expansion. The AMD agreement reinforces that trajectory.
AMD Gains Ground in AI Chip Battle
The deal centers on AMD’s MI300 series GPUs, designed for AI training and inference in large-scale data center environments.
The chips compete directly with Nvidia’s dominant AI accelerators, which currently lead the market.
Multi-vendor sourcing has become increasingly common among hyperscalers seeking supply resilience and pricing leverage.
The agreement provides AMD with long-term revenue visibility while strengthening Meta’s supply chain diversification strategy.
In the current AI cycle, access to advanced accelerators can materially affect development timelines and product rollout speed.
