The numbers for 2026 are painting a grim picture for the industry’s “frontrunners.” We’re looking at an operational model where, for every dollar of revenue brought in, the lead horse—OpenAI—is burning an additional $1.22 just to keep the lights on. A -122% operating margin isn’t growth; it’s a masterclass in fiscal defiance.

The reality is capital destruction on an industrial scale. We see 95% of enterprise AI projects failing to deliver measurable returns, yet organizations remain trapped in endless, budget-draining experimentation. Meanwhile, once-promising ventures are shuttering as the crushing cost of compute outpaces the value customers are willing to pay, and firms find that the cost of managing and integrating AI output has become an expensive tax that exceeds the manual effort it was meant to replace.

Basic math—logic, really—says you don’t double down on an investment unless there’s a definitive payoff on the horizon. There is a payoff on the horizon, right?

The industry and its global investors are caught in a FOMO-fueled race to incinerate capital. It’s an arms race where major players are determined to lose money and refuse to check the fuel gauge. And it’s not just the hardware vendors; the beneficiaries include a widening circle of power producers and infrastructure firms who are happily cashing the checks while the primary models burn.

When will this breathless, rudderless investment end? At what point does a board, a major investor, or a regulator finally have the spine to state the obvious? Are we really going to wait for a market crash to acknowledge what the P&Ls have been shouting for nearly two years.