OpenAI moves to its own Broadcom chip while committing to 10 gigawatts of Nvidia systems
The bottleneck for AI is shifting from models to megawatts. OpenAI is moving toward its first custom chip built with Broadcom $AVGO, while committing to deploy 10 gigawatts of Nvidia $NVDA systems, with Nvidia signaling 100 billion dollars of investment. Source: nvidianews.nvidia
The competition in artificial intelligence is increasingly a competition for chips and electricity, and two recent moves make that clear. First, OpenAI is working with Broadcom to bring its first in-house AI accelerator into production. Designing its own silicon would reduce OpenAI's dependence on any single chip supplier and give it more control over cost and performance. Second, OpenAI and Nvidia announced a letter of intent for a strategic partnership to deploy at least 10 gigawatts of Nvidia systems for OpenAI's next generation of infrastructure. Nvidia intends to invest progressively as each gigawatt is deployed, with the first gigawatt targeted for the second half of 2026 on the Vera Rubin platform. Ten gigawatts is comparable to the output of about ten large power plants. That scale is why access to energy, not just access to chips, is becoming the defining constraint on how fast AI infrastructure can grow. Source: nvidianews.nvidia.com.