AMD, Flex to Build AI Chips in US
Flex and AMD have announced a US manufacturing collaboration to build AMD's AI accelerator platforms, starting with the Instinct MI355X GPU. The move is a direct response to customer demand for onshore production and a more resilient domestic supply chain. This gives AMD a powerful selling point for public sector and other regulated industry contracts.
The AMD Instinct MI355X enters a competitive field, targeting Nvidia's dominance in the AI accelerator market. Compared to Nvidia's B200, the MI355X boasts a significant advantage in memory capacity, offering 288GB of HBM3E memory per GPU versus the B200's 192GB. This larger memory is critical for training and running massive AI models, potentially reducing the need for complex parallelism and sharding techniques. This collaboration between AMD and Flex is a direct response to the strategic push for a more resilient domestic semiconductor supply chain. The onshoring of manufacturing to Flex's facility in Austin, Texas, aligns with the objectives of the US CHIPS and Science Act, which provides over $52 billion in incentives to boost domestic research and production. This move not only mitigates supply chain risks but also strengthens AMD's position for securing contracts with the public sector and other regulated industries that prioritize US-based manufacturing. The decision by hyperscalers to "build or buy" AI compute is a central theme in the data center infrastructure landscape. While giants like Google, Amazon, and Microsoft are investing heavily in their own custom silicon (ASICs) to optimize performance and cost for their specific internal workloads, they remain significant buyers of high-performance GPUs from merchant suppliers like AMD and Nvidia. These enterprise-grade GPUs offer the flexibility and general-purpose performance required to serve a wide array of external cloud customers and diverse AI development needs. For AI startups and enterprise ML teams, the availability of powerful, memory-rich accelerators like the MI355X from competitive vendors is crucial. The intense demand for AI compute has created a bottleneck, with startups often struggling to access sufficient GPU resources to train their models. The growth of the AI infrastructure market, which saw funding increase tenfold to nearly $12.8 billion between 2022 and 2025, underscores this critical need. AMD's expanded, onshore production offers another high-performance option, fostering a more competitive and accessible market for these innovators.
