OpenAI’s $207 Billion Question: Can The A.I. Boom Ever Pay For Itself?
Most people know OpenAI as the company behind ChatGPT, the friendly chatbot that answers questions in seconds. What almost nobody sees is the industrial machine behind it – and the financial risk that now worries some of the world’s biggest banks.
Analysts at HSBC tried to answer a simple question: can this business ever really pay for itself? Their answer is uncomfortable. Even in 2030, they think OpenAI could still be losing money and needing around $207 billion more to fund its plans.
On paper, the growth story looks spectacular. By the end of the decade, OpenAI could be serving about 3 billion people, close to half of all adults on the planet.
Revenues might pass $200 billion a year as more users pay for subscriptions and companies build OpenAI tools into everything from office software to search and digital ads.
But the cost side looks like something from the energy industry, not from Silicon Valley’s old “light” tech model. OpenAI has signed gigantic cloud-computing deals with Microsoft and Amazon and is aiming for 36 gigawatts of AI power by 2030.
AI’s trillion-dollar bet
One gigawatt is enough electricity for roughly 750,000 homes. HSBC estimates that the company’s cloud and data-centre costs alone could reach $792 billion from late 2025 to 2030 and that total compute commitments could rise towards $1.4 trillion by 2033.
For hardware makers and cloud giants this is a dream: rivers of cash for chips, servers and new data centres. For expats and international workers, it is a distant boardroom decision that can still shape your pension, your next job and even parts of your electricity bill.
All of this rests on a bet that future productivity gains will justify today’s massive capital spending. The story behind the story is about what kind of capitalism we want.
One path is disciplined investment that proves its value in productivity, jobs and higher wages. The other is a race driven by hype, political promises and cheap money. OpenAI now sits exactly on that line – and so, indirectly, do all of us who use its tools.
This article was drafted with automated assistance and reviewed before publication. How we use AI · Report an error · Editorial responsibility: Matthias Camenzind, Editor-in-Chief