OpenAI Triples to $5.7B Revenue as Q1 Costs Triple to $3.7B

OpenAI generated $5.7 billion in revenue in the first quarter of 2026 and burned through $3.7 billion in operating costs to do it. According to The Next Web, reporting figures from documents shared with shareholders first surfaced by The Information, both numbers tripled year-over-year. The symmetry is the story: this is not a company where costs are outrunning a struggling revenue line, but one where cost and revenue are scaling in lockstep — producing neither a runaway loss spiral nor the efficiency gains investors would prefer.

The company's own characterisation of the challenge is worth sitting with. The article notes that "the cost of producing it grew at the same rate" as revenue, with frontier-model inference as the structural culprit. Every query runs through expensive compute, and as the user base expands, so does the compute bill. This is essentially the inverse of the operating-leverage dynamic that made prior-generation software companies such appealing public investments: serving another user costs about the same as serving the last one.

That math sits awkwardly with OpenAI's IPO ambitions. The company has filed confidentially for a US listing potentially valued at up to $1 trillion, with a listing reportedly possible as early as September — a flotation that, at that level, would rank among the largest in history. The cash position of $73 billion at quarter's end, up from $40 billion in December, looks substantial until you note it primarily reflects the March funding round that valued OpenAI at $852 billion, not operational cash generation. The company expects to spend "tens of billions in a single year" on compute, research, and infrastructure, and has told investors it does not expect profitability until the end of the decade.

What the reporting does not give you is the revenue breakdown across products — subscriptions versus API versus enterprise — or any specific efficiency milestone that would show when inference costs might start growing slower than revenue. The honest caveat is that "end of the decade" is a long timeline to ask public investors to hold, and quarterly earnings calls have a way of concentrating attention on horizons far shorter than that.

The underlying bet is that model efficiency improvements — custom silicon, architectural gains, better training pipelines — eventually decouple cost from user growth. Inference costs across the industry have trended down sharply over the past two years, so the directional case is plausible. But OpenAI is making that bet at a scale and valuation where there is very little margin for the timeline to slip.