OpenAI's Missed Targets Expose Generative AI's Unit-Economics Problem

The company, which has raised over $120bn at a valuation of roughly $852bn, is no longer growing at the pace its own leadership projected. For the hundreds of UK SMEs and scale-ups that have woven OpenAI's models into their products and workflows, the implications extend well beyond one firm's pre-IPO jitters. They touch pricing stability, platform risk, and the durability of the generative AI business model itself.

Where the numbers fell short

OpenAI set an internal goal of reaching one billion weekly active users by the end of 2025, the Wall Street Journal reported. It did not hit that mark. The company also missed its annual revenue target, having previously been cited at an annualised run rate of roughly $12.7bn in late 2025.

The misses were not confined to a single quarter. The Journal's reporting indicated that OpenAI fell short of various monthly revenue targets earlier in the year as well, and that subscriber churn has been higher than expected. That pattern suggests the gap is structural rather than seasonal.

The valuation trajectory makes the shortfall harder to absorb. In late 2024, OpenAI raised at a valuation of approximately $157bn. Within months, that figure had ballooned to $852bn, a more than fivefold expansion that far outstripped the growth in underlying revenue. The mismatch between multiple and monetisation is now impossible for prospective public-market investors to ignore.

Competition has also sharpened. OpenAI has lost ground in parts of the coding market, according to the Wall Street Journal, with Anthropic's Claude and Google's Gemini making notable gains in both enterprise API usage and consumer adoption. Neither rival has disclosed figures that allow a direct revenue comparison, but their market-share advances appear to have contributed directly to OpenAI's missed sales targets.

The cost problem: compute spending versus revenue reality

OpenAI's chief financial officer, Sarah Friar, has warned colleagues internally that the company may struggle to cover future computing costs if revenues do not accelerate, according to the Wall Street Journal. The statement is remarkable for its candour: a CFO preparing for a public listing does not typically flag existential cost pressures to peers unless the concern is acute.

The firm has committed to enormous investment in data centres and infrastructure. Board members have reportedly scrutinised major data-centre deals and questioned whether continued expansion is justified against softer growth, the Journal reported.

Despite those concerns, OpenAI's chief executive, Sam Altman, and Friar have said they remain "aligned on buying as much compute as we can," according to the Wall Street Journal. The tension between that ambition and the revenue trajectory is the central question facing the business.

Generative AI models are uniquely capital-intensive. Every inference, every API call, every ChatGPT conversation consumes GPU cycles that must be paid for. Unlike a traditional software product, where the marginal cost of serving an additional user is negligible, large language models carry variable costs that scale with usage. When user growth stalls and subscribers churn, the denominator in the unit-economics equation shrinks while the numerator, compute spend, remains fixed or grows.

For the broader sector, this dynamic is instructive. If the company with the largest installed base and the deepest pockets cannot make the maths work at scale, smaller model providers face the same physics with fewer resources.

Microsoft steps back from exclusivity

Alongside the growth slowdown, OpenAI has renegotiated its relationship with Microsoft (NASDAQ: MSFT), its largest strategic partner.

Under the revised terms, Microsoft will no longer hold exclusive access to OpenAI's models and intellectual property, the Wall Street Journal reported. Instead, it will retain a non-exclusive licence through 2032 while continuing as OpenAI's "primary cloud partner" via Azure.

Microsoft holds roughly a 25% stake in OpenAI's for-profit arm and is expected to continue generating billions from the partnership, though it surrenders meaningful exclusivity in return.

The restructuring was prompted in part by OpenAI's separate arrangement with Amazon Web Services, which had created potential legal tensions over cloud exclusivity. By moving to a non-exclusive model, OpenAI gains the freedom to distribute its technology more broadly, but it also signals that the original partnership structure, which gave Microsoft a decisive competitive advantage in the enterprise AI market, is no longer tenable.

For UK businesses procuring AI services through Azure, the practical effect may be limited in the short term. Microsoft remains the primary cloud host. But the shift opens the door to OpenAI models appearing on rival cloud platforms, which could alter pricing dynamics and reduce switching costs for organisations currently locked into Azure-based AI deployments.

What this means for businesses built on OpenAI's stack

The combination of missed targets, rising costs, and a restructured Microsoft relationship creates a set of risks that any business dependent on OpenAI's API should weigh carefully.

Pricing risk

If OpenAI's unit economics remain under pressure, the company has limited options: raise prices, reduce model quality to cut inference costs, or accelerate the shift toward tiered subscription models that push heavier users onto more expensive plans. All three outcomes affect the cost base of downstream businesses.

Platform concentration

Many SMEs and scale-ups have integrated OpenAI's models deeply into customer-facing products, internal workflows, or both. The missed growth targets and CFO warnings introduce a degree of platform risk that was easier to dismiss when the company appeared to be on an unbroken growth trajectory. Diversifying across model providers, including Anthropic and open-source alternatives, becomes a more defensible strategy.

IPO uncertainty

OpenAI has signalled its intention to pursue a public listing later in 2026. The broader IPO market for loss-making, capital-intensive technology firms remains cautious. A listing at or near the $852bn private valuation would require public-market investors to accept a multiple that the company's own revenue performance has not yet justified. If the IPO is delayed or repriced significantly, it could affect OpenAI's ability to fund the compute expansion that Altman and Friar have described as essential.

Competitive dynamics

The entry of Anthropic and Google into segments where OpenAI previously dominated, particularly coding tools and enterprise APIs, gives procurement teams more negotiating power. For scale-ups building on generative AI, this is broadly positive: more competition tends to drive down API pricing and improve service terms. But it also demands that technical teams stay current with a rapidly shifting landscape of model capabilities.

None of this means OpenAI is failing. A company generating revenue at a $12.7bn annualised rate, even if that figure fell short of internal projections, is operating at a scale most technology businesses never reach. The question is whether the gap between ambition and execution, between valuation and revenue, between compute spending and monetisation, can close fast enough to sustain the confidence of investors, partners, and the thousands of businesses that have made OpenAI a load-bearing part of their operations.