A cache of audited financials seen by Ed Zitron and confirmed by the Financial Times reveals that OpenAI posted a $38.53 billion loss attributable to the company in 2025. The documents are not public, but the numbers circulated widely after an Ars Technica story highlighted them on Tuesday. The loss dwarfs prior year deficits significantly. The leak cuts through the usual opacity surrounding AI funding cycles.

For the same year, the same filings show $13.07 billion in revenue, meaning operating expenses exceeded income by a staggering margin. Analysts infer that the bulk of the shortfall stems from massive compute expenditures required to train and serve large language models at scale. Such spending pressures margin targets the sector. Those costs, while fueling rapid model iteration, now appear unsustainable without fresh capital.

The figures force investors to reassess the economics of frontier AI labs that rely on ever‑larger clusters of GPUs and custom ASICs. If OpenAI cannot narrow its burn rate, future fundraising rounds may demand equity concessions or shift toward cost‑sharing arrangements with cloud partners. Stakeholders will watch cost discipline closely now. The leak underscores that scaling breakthroughs now carry a hefty price tag.