In recent weeks, OpenAI has forged significant partnerships with AMD and Broadcom to produce vast quantities of AI chips, essential for its ambitious goals. OpenAI’s agreement with AMD involves building 6 gigawatts (GW) of GPUs, starting with an initial deployment in late 2026. Concurrently, OpenAI and Broadcom will develop 10 GW of AI accelerators and ethernet systems to enhance data center connectivity. These deals aim to alleviate OpenAI’s dependence on single suppliers while optimizing chip costs and performance. However, TSMC remains the only capable manufacturer for these chips, creating a “bottleneck” in the industry. TSMC’s advanced production capabilities have led to its dominance, producing the latest 3nm chips, essential for powering modern technology. Despite efforts to expand production capacity in the U.S., the industry’s reliance on TSMC poses significant risks, making it a critical point of failure in global supply chains.
Source link
