OpenAI announced a new strategic collaboration with Broadcom to co-develop custom AI accelerators and expand high-performance compute infrastructure supporting its next generation of models. The partnership aims to deepen OpenAI’s hardware ecosystem beyond GPUs, integrating Broadcom’s semiconductor expertise to deliver optimized silicon for AI workloads at unprecedented efficiency and scale. According to the joint release, the collaboration covers multi-year R&D programs focused on advanced packaging, network interconnects, and memory architectures tailored to the needs of frontier model training. OpenAI will work directly with Broadcom engineers to design and deploy purpose-built AI accelerators that complement existing GPU clusters, reducing latency and energy consumption while increasing throughput. The companies emphasized that the initiative aligns with OpenAI’s broader effort to diversify its compute stack across multiple partners and technologies. Sam Altman, CEO of OpenAI, said the collaboration will “accelerate innovation across the full AI stack—from silicon to software—while ensuring sustainable scaling.” Hock Tan, President and CEO of Broadcom, added that combining OpenAI’s model engineering with Broadcom’s chip design and networking capabilities will help unlock the next phase of AI infrastructure performance. The partnership arrives amid intensifying competition among AI labs to secure reliable, efficient compute supply as model complexity grows. Analysts view the deal as a signal that OpenAI intends to pursue a hybrid strategy—balancing GPU partnerships with NVIDIA and AMD alongside new custom silicon projects for long-term efficiency and resilience. Full announcement: OpenAI and Broadcom announce strategic collaboration.