Anthropic, OpenAI Dominate 89% of AI App Revenue; Musk's xAI Misses Top 5

Anthropic and OpenAI now control 89% of the paid AI app market, with combined annualized revenue of $80 billion, up 112% from six months ago. OpenAI generated $55 billion and Anthropic $15 billion, while the rest of the 34 major private AI startups split just 11% of the market. The article also highlights a deepening oligopoly driven by VC dual-bet investing and pricing actions ahead of planned IPOs, which supports the leaders' revenue and valuation momentum.

Analysis

The important implication is not just that model revenue is concentrated, but that pricing power is starting to migrate from application-layer software back to foundation models. That raises the hurdle rate for every AI-native wrapper: if the model toll continues to rise, many “high-growth” startups will see gross margin compression before they ever reach durable scale, which is a late-cycle VC dynamic more than a broad software bull case. It also means the next leg of winner-take-most may be in enterprise distribution and compute infrastructure, not in every new AI app.

For public names, MSFT faces the clearest second-order pressure because model resale and partner economics cap margin upside in Azure AI even as demand grows; the market often underestimates how much revenue can look accretive while contribution margin stays thin. GOOGL is more insulated than it appears: if the market is truly moving toward a small set of premium model providers, Gemini can win share on quality and distribution, but the bigger equity story remains that search/ads cash flow can subsidize model losses longer than private peers can. RAMP is the cleaner beneficiary in the data because procurement and spend management should monetize the proliferation of paid AI seats even if the ultimate model suppliers capture most of the value.

The contrarian read is that the oligopoly is both real and fragile. Real, because enterprise buyers are converging on two trusted vendors; fragile, because any step-change in open-source performance or a material price cut from a hyperscaler could reset willingness to pay within one or two quarters. The IPO window creates a near-term incentive to maximize reported revenue rather than optimize retention, so the next 90-180 days may look stronger than the underlying lifetime value economics would justify.

Key Decisions for Investors

• Long RAMP for 3-6 months: the best public-market expression of AI spend expansion without needing to pick the ultimate model winner. Use pullbacks to add; target a 15-25% upside if AI seat growth converts into durable procurement share.
• Underweight MSFT vs GOOGL over 1-2 quarters: MSFT is more exposed to partner rev-share and low-margin resale economics, while GOOGL has more flexibility to defend with first-party distribution. Structure as a pair trade if Azure AI margin commentary stays pressured.
• Short a basket of AI application enablers with weak differentiation on any strength over the next 1-3 months: the thesis is margin squeeze from rising foundation-model input costs. Favor names with low switching costs and heavy model-dependent COGS.
• Buy medium-dated call spreads on RAMP or other AI spend-management beneficiaries into any post-earnings selloff: volatility should stay bid as investors reprice enterprise AI adoption, but the upside is capped by broader SaaS multiple compression.
• Watch for a reversal catalyst in the next 90 days: major hyperscaler model price cuts or a high-performing open-source release. If that happens, rotate out of model-purity winners and back into distribution/platform names.