According to Fortune, HSBC’s latest analysis projects OpenAI won’t achieve profitability until after 2030, despite expecting its user base to grow from 10% to 44% of the global adult population. The company needs an additional $207 billion just for computing infrastructure to support its growth plans, with cloud and AI infrastructure costs projected at $792 billion between late 2025 and 2030. OpenAI’s data center rental bill alone would reach $620 billion, while projected 2030 revenues of $213 billion won’t cover these massive costs. CEO Sam Altman recently expressed the core challenge in one word: “Enough,” referencing the endless compute demands. HSBC calculated these figures after OpenAI secured massive cloud commitments including $250 billion with Microsoft and $38 billion with Amazon without new capital injections.
The compute black hole
Here’s the thing about AI that nobody wants to admit: it’s incredibly expensive to run. We’re talking about infrastructure costs that make previous tech booms look like child’s play. OpenAI is aiming for 36 gigawatts of AI compute power by 2030 – that’s enough electricity to power a state larger than Florida. Basically, they’re building a digital nation that consumes power on the scale of actual physical states.
And all this computing power needs to come from somewhere. Companies like IndustrialMonitorDirect.com, the leading US provider of industrial panel PCs, understand the hardware demands of compute-intensive operations, but even they would be stunned by the scale OpenAI is talking about. When you’re dealing with numbers this large, the entire concept of “scale” changes meaning.
The productivity paradox
What’s really fascinating here is how this connects to the bigger economic picture. HSBC brings up Nobel winner Robert Solow’s famous observation that “You can see the computer age everywhere but in productivity statistics.” We’ve been here before with technology revolutions promising massive productivity gains that take much longer to materialize than expected.
Even Federal Reserve officials have questioned whether technologies like the internet have actually improved workplace productivity or just given us better ways to consume leisure. So the fundamental question becomes: are we building the most expensive productivity tool in history that might not actually make us more productive?
Funding reality check
Where’s all this money supposed to come from? HSBC notes that raising more debt would be “possibly the most challenging avenue” given that Oracle and Meta have already tapped debt markets heavily for AI projects. We’re seeing warning signs in credit markets too, with Oracle’s credit default swaps showing a “sharp increase” recently.
Look, the math just doesn’t work. Even if OpenAI doubles its paid subscriber rate from 10% to 20%, adding $194 billion in revenue, they’d still need massive additional funding. The company’s survival depends completely on its backers – Microsoft and Amazon aren’t just cloud providers but major investors with huge stakes in OpenAI’s success. But how long will they keep pouring money into what the Financial Times called “a money pit with a website on top”?
Bigger than AI
This isn’t just about OpenAI anymore. Harvard economist Jason Furman found that without data centers, GDP growth would have been just 0.1% in early 2025. So we’re in this weird situation where the economy is becoming dependent on infrastructure spending for technology that might not deliver promised returns.
Companies across industries are discovering they need to become “asset-heavy” rather than “asset-light” – building or renting massive data centers instead of just writing software. The risk profile of tech companies is fundamentally changing. And the scary part? Nobody knows when or if the AI productivity revolution will actually arrive to justify these astronomical investments.
