The Big AI Companies Are Going to See Their Margins Disappear

Why It Matters

Margin erosion threatens the financial sustainability of the biggest U.S. AI firms and could shift the industry toward a commodity infrastructure, affecting investors, developers, and enterprise adopters.

Key Takeaways

• Anthropic and OpenAI still unprofitable, relying on investor cash
• Metered pricing introduced to boost AI revenue as token costs rise
• Open‑weight Chinese models expected to match US leaders by 2026
• Enterprise AI spend dominated by coding (~$3 B), other sectors lag
• OS and cloud platforms poised to capture most AI distribution value

Pulse Analysis

The current business model for frontier AI labs hinges on high‑margin token sales, yet both Anthropic and OpenAI are burning cash while offering subscription tiers that can cost hundreds of dollars per month. To offset the deficit, they have shifted toward metered usage pricing, a strategy that extracts more revenue per token but also exposes customers to volatile costs. This approach buys time for investors but deepens the urgency for broader adoption, because without a surge in enterprise demand the labs cannot achieve sustainable profitability.

At the same time, the competitive landscape is tightening as open‑weight models from China and other regions close the performance gap. Researchers estimate a seven‑month lead for U.S. firms, but by the end of 2026 models such as GLM‑5.1 and Qwen3‑Coder‑Next are expected to rival Anthropic’s Claude Opus 4.7 and OpenAI’s GPT‑5.5. As these alternatives become readily available at a fraction of the price, the pricing power of U.S. labs will diminish, forcing a market correction toward commoditization. This shift mirrors earlier software cycles where once‑proprietary platforms became infrastructure services.

The downstream impact will likely favor companies that control distribution channels. Operating‑system giants like Apple, Google, and Microsoft, together with cloud providers such as Amazon Web Services, already host AI workloads and can bundle them with existing services, creating sticky ecosystems. Enterprise AI spending remains concentrated in coding (~$3 B annually) while legal, support, and health sectors lag behind, indicating ample room for growth if pricing stabilizes. Ultimately, the firms that can embed AI into ubiquitous platforms will capture the most value, while pure‑play model providers may see margins shrink to near‑zero as the market matures.