Financial projections in the generative artificial intelligence sector have long relied on creative accounting and aggressive comparisons, but the numbers are finally catching up with the hype. As TechCrunch reported, OpenAI recently informed investors that its annualized revenue is approaching $50 billion. That figure represents a rather glaring downward adjustment of $20 billion compared to earlier internal projections that floated the $70 billion mark. In our view, that earlier astronomical figure existed primarily to match Anthropic’s annualized revenues through generous, investor-crafted comparative math.
Methodological differences explain part of this discrepancy, though they do little to excuse the inflated baseline. OpenAI and Anthropic calculate their annualized revenue differently—Anthropic counts sales made by its cloud partners, while OpenAI historically excludes them. As TechCrunch noted, this revenue ambiguity has complicated the company's efforts to justify the gargantuan capital pouring into its infrastructure, including a staggering $122 billion raised during a single funding round in March.
Infrastructure Burn and Capital Realities
The gap between projected growth and actual operational intake highlights the brutal financial burden of scaling frontier models. The company’s leaked 2025 financials from earlier this year revealed that it had generated about $13 billion while spending significantly more to maintain operations.
The company’s leaked 2025 financials earlier this year showed it had made about $13 billion but spent significantly more.
These mounting compute costs and forced revenue adjustments are already rewriting corporate timelines. OpenAI’s initial public offering, previously whispered to materialize this year, has reportedly been pushed off until early 2027.
It reads as a classic collision between venture capital valuation models and the sluggish pace of enterprise software adoption. When top-line figures must be revised downward by twenty billion dollars just to maintain a nodding acquaintance with cash flow, the timeline for public market readiness inevitably stretches out into the distance. For decision-makers watching the AI infrastructure race, this correction is a stark reminder that burning capital on compute does not automatically translate to enterprise demand.