Peter Hoeschele, OpenAI's Vice President of Compute, has joined Meta in a significant infrastructure leadership hire. At OpenAI, Hoeschele was responsible for overseeing the massive compute infrastructure required to train frontier models including GPT-4 and its successors, managing relationships with cloud providers and coordinating the deployment of tens of thousands of GPUs for training runs.
His expertise in scaling AI compute operations made him one of the most strategically important executives at OpenAI. The move comes as Meta has committed over $65 billion in AI capital expenditure, building out enormous data center capacity to support its Llama model family and AI-powered products across Facebook, Instagram, and WhatsApp.
Hoeschele's deep knowledge of large-scale AI training infrastructure positions him to help Meta optimize its unprecedented investment in AI compute. His departure is a notable loss for OpenAI at a time when compute resources remain one of the most critical bottlenecks in developing next-generation AI systems, and represents Meta's continued ability to attract top-tier talent from rival labs.
