Mark Cuban says OpenAI won’t be able to repay $1T of capital

Mark Cuban’s point is not that OpenAI is fake. It’s almost the opposite. He thinks AI is real, useful, and inevitable — but that the economics of the current arms race may be broken. In a Big Technology Podcast episode published April 29, 2026, Cuban said OpenAI is “wasting” money at scale and that the investment “will never pay off.” The clip landed just weeks after OpenAI closed a $122 billion funding round, on top of earlier megadeals to finance even more compute. (podcasts.apple.com) ### What is Cuban actually arguing? He’s making an ROI argument, not a capability argument. Basically, he’s saying OpenAI may build amazing models and still fail to earn enough money to justify the mountain of capital going in. That is a very different critique from “AI is a bubble” or “the tech doesn’t work.” Cuban’s version is harsher in a way — the product can succeed and the investors can still get crushed. (youtube.com) ### Why does the number feel so huge? Because OpenAI’s spending story now comes from multiple giant buckets piled on top of each other. In January 2025, OpenAI and partners announced Stargate, a project that intends to invest $500 billion over four years in U.S. AI infrastructure, with $100 billion slated to start immediately. Then on March 31, 2025, OpenAI announced a separat(youtube.com) it had closed another round with $122 billion in committed capital at an $852 billion post-money valuation. (openai.com) ### Is that really “OpenAI owes $1 trillion”? Not exactly — and that’s the important cleanup. The trillion-dollar line is more like a shorthand for the scale of capital and infrastructure commitments orbiting OpenAI, not a literal bill due tomorrow from one checking account. Stargate itself is a separate company with SoftBank, Oracle, MGX, and OpenAI involved, and(openai.com)nsibility. So Cuban is talking about the broader investment thesis around OpenAI, not just one neat debt number on OpenAI’s balance sheet. (openai.com) ### So why might the math break? Because frontier AI is brutally capital intensive. OpenAI itself now frames “durable access to compute” as the strategic advantage that compounds across research, products, and delivery. That sounds strong, but the catch is obvious — compute is also the thing you have to keep buying in absurd quantities. If model quality improves s(openai.com)enue grows. That’s the heart of Cuban’s complaint. (openai.com) ### But isn’t OpenAI growing incredibly fast? Yes — very fast. OpenAI says it is now generating $2 billion in revenue per month. It says it reached $1 billion in revenue within a year of launching ChatGPT, hit $1 billion per quarter by the end of 2024, and is now aiming at far larger scale. CNBC also noted that the company made $13.1 billion in revenue last year an(openai.com)It’s a “can revenue outrun infrastructure?” story. (openai.com) ### Why does the structure matter? Because OpenAI is no longer just a scrappy lab raising venture money. It has become a financing machine, a distribution platform, and an infrastructure buyer all at once. The company also completed a recapitalization in October 2025, putting the for-profit business into a public benefit corporation controlled by the nonprofit. Tha(openai.com)rlines how large and permanent this whole operation has become. (openai.com) ### What would prove Cuban wrong? OpenAI would need to show that AI becomes less like a costly demo and more like a utility with expanding margins. That means consumer subscriptions hold up, enterprise adoption deepens, developers keep building on its APIs, and the cost to serve each new user falls fast enough to offset the infrastructure binge. OpenAI’s own pitch is ex(openai.com)d compute reinforcing one another. (openai.com) ### Bottom line? Cuban is attacking the business model, not the technology. And right now that’s the sharper question. OpenAI has clearly won attention, users, and capital. What it has not yet fully proved is that spending at this scale can produce returns at the same scale. (youtube.com)