Meta taps AMD in $60 billion, five‑year AI chip deal to power its infrastructure

Meta’s partnership with AMD will see the chipmaker deliver its next‑generation CDNA 3 GPUs and custom AI accelerators across a multi‑year rollout that is expected to underpin the social‑media giant’s generative‑AI services, according to Reuters. The agreement, valued at up to $60 billion over five years, marks AMD’s second “mega” supply contract after a similar deal with Microsoft earlier this year, and it positions the company as a primary hardware supplier for Meta’s expanding AI workloads (Reuters).

The contract calls for AMD to ship a mix of its Radeon Instinct MI300X and MI300A GPUs—devices built on the CDNA 3 architecture that combine high‑bandwidth memory (HBM3) with dedicated matrix cores for tensor operations. AI Magazine notes that Meta will also receive custom‑designed ASICs derived from AMD’s “MI300‑based” silicon, enabling the social network to tailor inference pipelines for its LLaMA‑2 models and upcoming multimodal offerings. By leveraging AMD’s heterogeneous compute stack, Meta aims to reduce reliance on Nvidia’s H100/H200 GPUs, which have dominated the market but face supply constraints, according to Tech Wire Asia.

From a capacity standpoint, the deal is expected to deliver several hundred thousand GPU units, enough to fuel Meta’s internal AI research labs and the compute backbone for products such as the “Meta AI” assistant and AI‑enhanced content moderation tools. New Electronics reports that the agreement includes a “flexible volume schedule” that will scale with Meta’s projected 30‑percent year‑over‑year increase in AI compute demand. The partnership also grants Meta early access to AMD’s roadmap, including the forthcoming “MI400” series slated for 2025, which will feature next‑generation compute units and further improvements in power efficiency—a critical factor for Meta’s data‑center strategy.

Financially, the $60 billion figure translates to an average annual spend of $12 billion on AMD silicon, a sum that dwarfs AMD’s prior AI‑related revenue streams. Reuters highlights that the deal will boost AMD’s non‑CPU earnings, which have been growing at double‑digit rates but remain a modest share of the company’s total revenue. Analysts cited in the Reuters piece expect the contract to lift AMD’s gross margin on its data‑center segment, given the higher price points of AI‑optimized GPUs versus traditional graphics cards.

Strategically, the alliance underscores Meta’s drive to diversify its hardware supply chain amid ongoing geopolitical tensions and the broader industry scramble for AI compute. By locking in a long‑term, high‑volume source of chips, Meta can better control costs and mitigate the risk of bottlenecks that have plagued other firms reliant on a single vendor. The partnership also signals a shift in the AI hardware landscape, where AMD is emerging as a credible alternative to Nvidia’s dominance, a trend first noted by Reuters when it described the deal as AMD’s “second mega chip supply” contract.