We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
OpenAI Blog·Oct 13·5 min read
OpenAI and Broadcom announce strategic collaboration to deploy 10 gigawatts of OpenAI-designed AI accelerators
TL;DR
Training and deploying large AI models requires massive computational infrastructure, but existing accelerators (specialized chips for AI) lack sufficient scale and energy efficiency. OpenAI and Broadcom partnered to co-develop OpenAI-designed AI accelerators and custom Ethernet networking solutions for deployment across 10 gigawatts of capacity by 2029.
✦ Why It Matters
Engineers can plan infrastructure investments around a credible 10-gigawatt deployment timeline and OpenAI-validated accelerator specifications.
Key Takeaways
- Track OpenAI accelerator specifications for infrastructure procurement planning through 2029.
- Evaluate Broadcom Ethernet solutions for AI cluster networking requirements.
- Plan data center capacity around 10-gigawatt deployment milestones.
- Monitor co-developed systems for energy efficiency benchmarks versus existing accelerators.
Full Summary
Large language models and AI systems demand enormous computational power, creating a bottleneck in infrastructure capacity and energy consumption. OpenAI and Broadcom announced a strategic multi-year partnership to address this gap by co-developing next-generation AI accelerators (specialized processors optimized for machine learning workloads) designed by OpenAI and custom Ethernet solutions (high-speed networking technology) from Broadcom.
The collaboration targets deployment of 10 gigawatts of total compute capacity—a unit measuring power consumption—by 2029. The approach combines OpenAI's expertise in AI system requirements with Broadcom's semiconductor and networking capabilities to create integrated, energy-efficient infrastructure.
This partnership directly addresses the hardware constraints limiting AI model training and inference at scale, establishing a concrete timeline and capacity target for next-generation infrastructure deployment.