OpenAI Secures Major Chip Partnerships with AMD and Broadcom, Heightening Dependence on TSMC
Key Points
- OpenAI signed major chip supply deals with AMD and Broadcom.
- AMD will produce multi‑gigawatt GPU capacity, starting in late 2026.
- Broadcom will deliver custom AI accelerators and Ethernet systems through 2029.
- Both partners will rely on TSMC for advanced chip fabrication.
- TSMC’s capacity is described as very tight, creating industry‑wide supply risks.
- Diversifying design partners aims to reduce cost and improve performance for OpenAI.
- Experts note that TSMC remains the primary foundry for cutting‑edge AI silicon.
OpenAI has signed significant agreements with AMD and Broadcom to supply AI chips and accelerators for its data‑center deployments. The AMD deal focuses on large‑scale GPU production, while Broadcom will provide custom AI accelerators and Ethernet systems. Both partnerships underline OpenAI's strategy to diversify its silicon supply and reduce reliance on a single vendor. However, the majority of the chips are still expected to be fabricated by Taiwan Semiconductor Manufacturing Company (TSMC), whose capacity constraints and geographic concentration remain a critical bottleneck for the broader AI and semiconductor industries.
New Chip Deals Expand OpenAI's Supply Chain
OpenAI announced two high‑profile collaborations aimed at securing the hardware needed for its next generation of AI models. The first partnership with AMD will see the chipmaker build a multi‑gigawatt portfolio of GPUs, with initial deployments slated for the latter half of 2026. The second agreement with Broadcom involves the creation of custom AI accelerators and advanced Ethernet components, also targeting deployment beginning in 2026 and extending through 2029. Both deals are designed to provide OpenAI with a broader base of silicon suppliers and to tailor hardware to its specific inference workloads.
Industry Implications and Expert Views
Industry analysts note that the agreements illustrate a shift away from exclusive reliance on a single chip provider. By working with AMD and Broadcom, OpenAI aims to achieve cost efficiencies, performance gains, and supply‑chain resilience. Experts also highlight that the custom blocks Broadcom will assemble are based on its existing IP library, allowing faster time‑to‑market for specialized accelerators.
Continued Dependence on TSMC
Despite the diversification of design partners, the majority of the chips produced under these agreements are expected to be manufactured by Taiwan Semiconductor Manufacturing Company (TSMC). Commentators stress that TSMC remains the dominant foundry for advanced AI silicon, handling the high‑volume production for AMD GPUs, Broadcom accelerators, and many other industry leaders. The concentration of cutting‑edge manufacturing in Taiwan creates a strategic vulnerability, as capacity is already described as “very tight” and any disruption could impact global AI development.
Capacity Constraints and Future Outlook
TSMC’s limited capacity has already caused scheduling challenges for major customers, with some orders delayed by many months. The company is investing heavily in new facilities both in Taiwan and the United States to expand output, but analysts caution that alleviating the bottleneck will take years. In the meantime, OpenAI’s reliance on TSMC underscores the broader industry’s exposure to supply‑chain risks tied to a single geographic region.
Strategic Benefits for OpenAI
The dual‑partner approach offers OpenAI several strategic advantages. Custom silicon from Broadcom can be optimized for inference, potentially reducing power consumption and improving performance. AMD’s GPU production scale can support the massive compute demands of large language models. Together, the partnerships aim to deliver cost savings, diversified supply, and hardware tailored to OpenAI’s workloads, while still depending on TSMC’s advanced manufacturing capabilities.