Meta Halts Olympus Chip, Secures 6GW AMD GPUs and Google AI Deal
Meta scrapped its second-generation MTIA training chip project 'Olympus', originally due by Q4 2026, citing technical complexity and manufacturing risks after abandoning the Iris design. Meta struck a multibillion-dollar deal to rent AI chips from Google, agreed to deploy 6 gigawatts of AMD Instinct GPUs and plans a Nvidia rollout.
1. Meta Abandons Olympus AI Chip Development
Meta halted its second-generation MTIA training chip code-named Olympus, originally slated for Q4 2026, after facing technical complexity and manufacturing risks. The Olympus program follows the earlier cancellation of the Iris design as internal teams struggled with performance and software stability. This shift underscores the challenges of competing with established AI hardware ecosystems.
2. Meta Leans on External AI Chip Partners
The company agreed to a multibillion-dollar rental deal with Google for AI chips and will deploy 6 gigawatts of AMD Instinct GPUs under a multi-year partnership. Meta also confirmed plans to integrate Nvidia hardware across its data centers in a phased rollout. These external partnerships aim to support Meta’s AI training capacity while it reevaluates its in-house chip strategy.