Sources: OpenAI's losses roughly doubled to ~$540M in 2022, and Sam Altman privately suggested the company may try to raise as much as $100B in the coming years
OpenAI's losses roughly doubled to around $540 million last year as it developed ChatGPT and hired key employees from Google …
Context & Ripple Effects
In May 2023, The Information put a number on ChatGPT's cost: OpenAI's 2022 losses roughly doubled to ~$540M, driven by model development and hiring away key Google staff, even before ChatGPT generated meaningful revenue. The striking part is Altman's private framing — that OpenAI may need to raise as much as $100B in coming years — which reads less like fundraising chatter and more like a thesis that frontier AI is a capital problem first.
That framing aged into the company's operating reality rather than an outlier remark: within months Altman was telling staff OpenAI had reached $1.3B in annualized revenue growing 30% in three months, and by late 2025 OpenAI was projecting $115B of cumulative cash burn through 2029 — roughly $80B above what it had told investors previously.
First-order effects
- OpenAI's cost structure is now public knowledge for every investor and partner: training compute plus poached-Google salaries doubled losses to ~$540M in a year when revenue barely registered, making any future raise a necessity, not an option.
- Google loses directly twice — talent to OpenAI's hiring push and competitive secrecy around its own lab economics — while Altman's $100B signal tells capital markets to price OpenAI as a decade-scale infrastructure bet.
Second-order effects
- Rival labs face a forced escalation: if Altman is privately contemplating nine-figure-millions-to-$100B scale raises, competitors must either match that capital intensity or cede the frontier — a dynamic visible later when Anthropic reached profitability on $11.6B of quarterly sales while OpenAI's losses jumped 32% to $12.3B in a single quarter.
- Investors gain the first real loss baseline for a frontier lab, letting them anchor valuations to burn rates instead of demos — and pressuring OpenAI to convert its rapid revenue growth ($300M/month by August 2024) into a credible path against mounting losses.
Third-order effects
- If the pattern holds — losses doubling early, then compounding into tens of billions of projected burn — frontier AI consolidates around a handful of labs able to raise at unprecedented scale, structurally separating well-capitalized players from the rest of the field.
- A lab whose unit economics depend on repeated mega-rounds becomes dependent on continued investor appetite for unprofitable growth, making capital-market sentiment itself a systemic risk to AI progress timelines.
The trend: Frontier AI labs are converging on a model where capability leadership is purchased with escalating, multi-tens-of-billions capital commitments rather than earned through near-term profitability.