OpenAI is losing about three times more money than it's earning (www.theregister.com)

OpenAI is burning through cash far faster than it’s earning it: according to reporting, the company pulled in about $4.3 billion in revenue in H1 2025 but posted a net loss of roughly $13.5 billion (more than half driven by complex “remeasurement of convertible interest rights”). OpenAI reportedly books about $13 billion in annual recurring revenue based on recent monthly figures, yet roughly 70% of that ARR comes from ChatGPT and only ~5% of ChatGPT’s roughly 800 million users currently pay. The mismatch between massive free usage and a small paying cohort leaves OpenAI dependent on large partner investments and future monetization levers while it commits to buying over 26 GW of datacenter capacity and more than $1 trillion in partner deals through the rest of the decade. For the AI/ML community this underscores two big realities: first, freemium generative AI drives enormous demand for compute and drives platform dominance (OpenAI’s products account for ~80% of generative-AI web traffic), but converting users to revenue remains a hard product and pricing problem. Second, the capital and vendor dynamics (Nvidia’s reported $100B injection, GPU credits, massive datacenter commitments) amplify both rapid infrastructure growth and bubble concerns—large vendor-financed deployments may mask unsustainable unit economics unless new revenue streams (e-commerce commissions, ads, higher-paid tiers) scale. Technical implications include sustained heavy GPU/accelerator demand, continued investment in model-serving infrastructure, and pressure on product teams to create compelling paid features.