OpenAI's Numbers: Good or Bad?

There's a peculiar irony at the heart of the most hyped technology in human history. The company leading the artificial intelligence revolution — OpenAI, creator of ChatGPT, arguably the fastest-growing consumer product ever built — is hemorrhaging money at a pace that would make any traditional investor faint.

Not millions. Not hundreds of millions. Billions. Every single month.

And yet, the capital keeps pouring in — from SoftBank, from Microsoft, from the White House, from sovereign wealth funds you've barely heard of. So what's really going on? Is this a calculated moonshot, or the most expensive gamble in Silicon Valley history?

Let's follow the money.

The Numbers You Need to Know

Here's the brutal arithmetic. OpenAI generated approximately $13.1 billion in revenue in 2025 — a staggering 118% jump from $6 billion in 2024. Its annualized revenue run-rate hit $20 billion by end of 2025.

That sounds like a rocketship. Until you look at the other side of the ledger.

OpenAI posted a $13.5 billion net loss in just the first half of 2025 alone. The company's own internal projections show losses tripling to $14 billion in 2026, against roughly $13 billion in sales, with total spending clocking in at ~$22 billion.

HSBC's analysts went further, and their conclusion is sobering: OpenAI likely won't make money by 2030, and still faces a $207 billion funding shortfall to power its own growth plans.

Read that again: $207 billion more to raise, even after everything already committed.

How Did We Get Here?

To understand the cash crisis, you have to understand what it actually costs to build and run frontier AI.

The Compute Problem

AI doesn't run on hope. It runs on GPUs — thousands, then tens of thousands, then millions of them — packed into data centers consuming as much electricity as mid-sized cities.

OpenAI's inference-related costs (the cost of running models when users type queries) quadrupled in 2025 alone, causing its adjusted gross margin to collapse from 40% in 2024 to 33% in 2025. The more users ChatGPT gets, the more it costs to serve them — and ChatGPT now has 800 million weekly active users as of late 2025, with 20 million paid subscribers and 5 million+ paying business customers.

Every query. Every image generated. Every reasoning chain. Each one costs money.

And that's before training the next model generation. OpenAI reportedly spent north of $100 million training its O1 reasoning model — at a time when Chinese lab DeepSeek had achieved comparable results for just $5.9 million.

The Infrastructure Commitment Is Staggering

OpenAI has committed to spending that goes beyond anything in tech history:

• $250 billion committed to Azure (Microsoft's cloud) over 6 years (2025–2030)
• $38 billion to AWS (Amazon's cloud) over 7 years
• $90 billion in AMD Instinct GPU purchases
• $350 billion in custom AI accelerators with Broadcom

Project Stargate, announced by President Trump on January 21, 2025, is a joint venture between OpenAI, SoftBank, Oracle, and MGX — with a stated commitment of $500 billion over four years, $100 billion deployed immediately. The project is now at nearly 7 gigawatts of planned data center capacity.

By OpenAI's own revised estimates, the company targets $600 billion in compute spending by 2030. That's not a typo. Six hundred billion dollars in compute alone.

The Energy Problem Nobody Talks About Enough

There's a hidden tax on every AI model: electricity.

As per research in 2024, AI data centers are expected to consume 90 terawatt-hours of electricity annually by 2026 — roughly a tenfold increase from 2022 levels. Global critical power supporting data centers is projected to nearly double to 96 gigawatts by 2026, with AI operations alone consuming over 40% of that total.

In 2025, the world spent approximately $580 billion on AI-focused data center infrastructure. Wholesale electricity prices near data center clusters have risen as much as 267% versus five years ago. Residential electricity bills in parts of the US have risen by $16–18/month specifically because of data center power draws.

OpenAI isn't just building software. It's building a new energy-hungry civilization underground.

The DeepSeek Earthquake

In January 2025, a Chinese AI lab most Americans had never heard of dropped a model called DeepSeek-R1. It wasn't just competitive with OpenAI's flagship reasoning model — it was reportedly 20–50x cheaper to run, and cost only $5.9 million to train versus OpenAI's reported $100M+ for a comparable generation.

The market reacted instantly. Nvidia lost nearly $600 billion in market cap in a single day — the largest single-session loss in US stock market history.

The implications for OpenAI are existential. If Chinese competitors can produce frontier-level AI at a fraction of the cost, the entire premise of "you need to spend tens of billions to stay ahead" collapses. DeepSeek has since launched V4 — a 1-trillion parameter coding model with a 1-million-token context window — in February 2026. Meanwhile, a broader AI price war has erupted, with Chinese models now priced at one-sixth to one-fourth the cost of comparable US systems.

OpenAI's response has been to race forward — not slow down. GPT-5.2 was introduced in December 2025, and the next generation aims for "cognitive density" — packing more capability into a smaller, cheaper architecture. But the fundamental question hangs in the air: can you out-spend your way to a moat?

The Fundraising Flywheel — And Its Risks

OpenAI has become the most aggressively funded startup in history:

• March 2025: $40B SoftBank-led round at $300B valuation — the largest private tech deal on record
• Early 2026: Finalizing $100B raise at $750–850B valuation
• IPO target: $1 Trillion valuation in 2026 or 2027
• Stargate JV: $500B infrastructure commitment

SoftBank fully delivered its $22B+ final tranche in late December 2025, closing the full $40B commitment. OpenAI completed its restructuring into a for-profit public benefit corporation, giving Microsoft a 27% diluted stake to facilitate the path to IPO.

But there is a troubling circularity that analysts have flagged. OpenAI raises money from Microsoft — and then spends much of it at Microsoft Azure. SoftBank invests $40 billion — and Stargate's proceeds partly flow back to Stargate's other corporate partners. Investors are, in part, funding their own future revenues.

The Bull Case: Revenue to the Moon by 2030

To justify any of these valuations, OpenAI needs revenue growth that makes its 2025 numbers look quaint. The company's own internal forecasts project:

• $280 billion+ in annual revenue by 2030 — roughly equal from consumer and enterprise
• Cash flow positive by 2029–2030, generating ~$40 billion in annual cash
• 44% of the world's adult population using OpenAI's services in some form by 2030

The bear case, per HSBC: Even with all of that, the cumulative free cash flow through 2030 remains deeply negative, requiring $207 billion in additional capital raises. Every revenue projection is based on continued dominance — in a market where DeepSeek, Anthropic, Google Gemini, Meta's Llama, and a dozen other well-funded competitors are all racing to take share.

Is This a Crisis, or a Calculated Bet?

The framing of "OpenAI running out of money" misses the larger picture — but only slightly.

OpenAI isn't in imminent danger of bankruptcy. With $40 billion raised in March 2025, another $100 billion on the way, and a Stargate infrastructure fund behind it, the lights will stay on for years. The fundraising machine remains world-class.

The real question isn't survival. It's whether the economics of frontier AI can ever work at the scale needed.

OpenAI's cumulative cash burn forecast through 2029 is $112 billion. Its total compute spending target by 2030 is $600 billion. That's not a startup burning cash to find product-market fit. That's a sovereign-scale infrastructure build dressed in a software company's clothes.

Every prior technology wave — the internet, mobile, cloud — went through this inflection point where the question shifted from "can we build it?" to "can it pay for itself?" For AI, that reckoning is arriving now, in 2026, even as the hype and capital commitments hit all-time highs.

The AI industry raised over $202 billion in 2025 alone — up 75% year-over-year. But as analysts put it: "AI took investors on a date in 2025. In 2026, it's time to foot the bill."

The Bottom Line

1. OpenAI is growing revenue explosively — $13.1B in 2025, on track for $20B+ — but spending faster than it earns, with $14B in projected 2026 losses.
2. The compute cost problem is structural, not temporary. Inference costs quadrupled in one year. Data center power demand is growing at a pace the grid struggles to match.
3. DeepSeek has fundamentally challenged the assumption that massive spending equals moat. The price war is on.
4. OpenAI needs capital the way rockets need fuel — not as an emergency, but as a continuous operational requirement.
5. The 2030 vision is real but fragile: it requires dominant market share, no regulatory catastrophe, successful AGI development, and revenue growth with no real historical precedent.

Whether OpenAI can convert its first-mover advantage into durable profits before the money runs thin — or before a Chinese competitor undercuts the whole model — is the defining business question of our era.

The billions are burning. The race is on. And nobody, including OpenAI, knows exactly how it ends.