The biggest AI companies are running out of runway to prove they can actually make money. That’s the central tension explored in a recent Decoder episode on The Verge AI, where senior AI reporter Hayden Field breaks down why 2026 feels like a make-or-break moment for Anthropic, OpenAI, and the broader industry.
The core problem is straightforward: hundreds of billions in capital investment have flowed into AI infrastructure, data centers, and chips. Now the profits need to show up, or the whole thing unravels.
Agents Changed the Math
The catalyst? AI agents. Products like Claude Code, Codex, and the open-source OpenClaw framework have proven genuinely valuable to users. But they also consume compute at rates these companies didn’t anticipate. More usage should be good news. Instead, it’s forcing painful trade-offs.
The evidence is already piling up:
- OpenAI killed Sora, its video-generation app, walking away from a $1 billion Disney licensing deal. The reason: it costs too much to run, and OpenAI needs that compute for Codex.
- Anthropic blocked OpenClaw users from burning through resources on standard subscription plans, pushing them onto pay-as-you-go pricing that costs substantially more.
- Revenue projections leaked to the Wall Street Journal this week show both companies promising hundreds of billions in revenue and profitability by decade’s end.
These aren’t subtle signals. They’re companies actively reshuffling priorities because the economics of their most popular products don’t work yet.
The IPO Pressure Cooker
Both Anthropic and OpenAI are barreling toward what could be two of the biggest IPOs in history. OpenAI recently raised another $122 billion at an $850 billion valuation. That kind of number demands a credible path to massive, sustained revenue.
What stands out here is how the pressure is already reshaping product strategy. These companies aren’t just building the best technology anymore. They’re making hard calls about which products survive and which get cut, which users get cheap access and which don’t. The Sora shutdown is the clearest example: a flagship product with a billion-dollar partnership, killed because the compute was needed elsewhere.
As Nilay Patel notes, most CEOs he’s interviewed acknowledge that some companies will “fail in spectacular fashion” while others succeed. The opportunities are too big to ignore, so everyone keeps pushing forward. “We’re doing this, whether we want to or not,” is essentially the industry’s collective stance.
What This Means for Practitioners and Businesses
If you’re building on top of these platforms, pay attention to the pricing signals:
- Expect costs to rise. Flat-rate subscriptions for heavy agent usage are likely going away across the board. Budget for pay-as-you-go.
- Don’t bet on product stability. If a major AI company can kill a billion-dollar product overnight, your favorite API feature isn’t safe either.
- Watch the IPO timelines. Once these companies go public, shareholder pressure will accelerate the shift from growth-at-all-costs to margin protection.
- Diversify your stack. Relying on a single provider is riskier than ever when that provider might reprioritize compute away from your use case.
What Comes Next
The next 12 to 18 months will determine whether the AI industry’s staggering valuations were justified bets or the peak of a bubble. The companies that figure out how to make agents profitable, not just popular, will define the next era. The ones that don’t will become cautionary tales.
For a deeper dive into the dynamics at play, the full Decoder conversation with Hayden Field is worth your time over at The Verge.