The Quote

“It’s more surprising to me how many people still today think that these are horribly unprofitable businesses and that the whole market’s going to zero. It’s absurd to me that the meme of these, OpenAI, Anthropic, etc. are just burning cash… If they stopped training, they’d be massively profitable overnight.”

— @trsohmers, Positron AI Co-founder · The Twenty Minute VC (20VC)

Why It Matters

The speaker’s argument is precise: cash burn at frontier AI labs is a choice, not a condition. Training spend is investment-mode spending — deliberately front-loaded to build capability moats.

Confusing capex-for-growth with structural unprofitability is an analytical error. It’s the same mistake once made about Amazon’s retail margins.

The Structural Read · FourWeekMBA Analysis

Through the lens of the FDE Framework, frontier AI labs sit in the Founder tier — companies building core infrastructure that others will harness. That tier demands sustained reinvestment. Profitability is deferred by design, not by dysfunction.

The Product Overhang Doctrine applies here too. Capability accumulates invisibly during training runs. The market sees the burn; it doesn’t see the optionality being purchased. When that optionality surfaces — in new model generations, new product lines — the economics shift fast.

The real question isn’t are these businesses profitable — it’s what are they buying with the burn, and does it compound?

The Meme Problem

The speaker’s word choice — meme — is deliberate. A meme spreads because it’s sticky, not because it’s accurate. “AI companies burn cash” is a true statement that’s being used to support a false conclusion: that the unit economics are broken.

Inference revenue (running the model for paying customers) and training spend (building the next model) are two separate P&Ls living inside the same company. Conflating them is the error driving the meme.

The Bottom Line

Spending aggressively to build a capability lead is a strategy. Mistaking that strategy for structural failure is a meme. The speaker’s argument — that the inference business is already profitable and training is a deliberate bet — is the frame serious analysts should be using.