Fidji Simo says OpenAI can't afford not to strike massive compute deals: 'What's much riskier would be to not lean into compute' (www.businessinsider.com)

OpenAI’s applications chief Fidji Simo told Wired that the company must keep “hyperscaling” compute capacity, arguing the bigger risk is failing to lean into massive GPU and data‑center deals rather than being cautious. She said internal compute scarcity already forces OpenAI to gate features—citing ChatGPT Pulse, a personalized Pro feature, as something she wants available to all users but can’t scale because there aren’t enough GPUs. The comments come as OpenAI faces roughly $1.4 trillion in data‑center commitments over the next eight years, ongoing operating losses, and public scrutiny about whether hyperscale capex is forming a bubble; Sam Altman and other hyperscaler CEOs (including Mark Zuckerberg) have similarly framed large bets on capacity as necessary risks. For the AI/ML community, Simo’s stance signals continued, aggressive demand for GPUs, networking and datacenter capacity—meaning cloud providers, chip vendors, and infrastructure builders will remain central to model development and deployment. The immediate technical implication is that product availability and model scale will be constrained by hardware and procurement timelines, amplifying pressure for more efficient models, compiler/runtime improvements, and alternative architectures to stretch scarce compute. It also raises competitive stakes: startups and researchers without deep cloud commitments may be squeezed, while broader capex flows could accelerate hardware innovation and supply‑chain expansion (and the macroeconomic debates about a potential AI investment bubble).