I think Anthropic and OpenAI have found product-market fit
Simon Willison argues that Anthropic and OpenAI have finally achieved true product-market fit, driven by high-cost coding agents used by enterprises. This shift, marked by recent price increases and rumored profitability, positions them for significant revenue, despite some companies pushing back on rising costs. The Hacker News discussion debates the sustainability of this model against open-source alternatives and the actual value generated by such high spending.
The Lowdown
Simon Willison asserts that Anthropic and OpenAI have reached product-market fit, citing their recent strategic shift to charging enterprise customers API-level prices for their advanced LLM services, particularly coding agents. This move, which comes alongside new model releases with higher price points, is expected to generate substantial revenue and potentially lead to their first profitable quarters. Willison highlights his own experience, where his $200/month subscriptions would equate to over $2,000 at API rates, demonstrating the potential for significant enterprise spending.
- Pricing Strategy Shift: Both Anthropic and OpenAI transitioned enterprise pricing from discounted seat-based plans to direct API token usage around April 2026. This means large organizations are now paying standard API rates, which are significantly higher than previous bundled offerings.
- Coding Agents as Revenue Drivers: The author identifies coding/general-purpose agents (like Claude Code/Cowork and OpenAI Codex) as the primary engine for this product-market fit. These tools consume vast numbers of tokens but are used by highly compensated professionals, making them a lucrative revenue stream far beyond consumer ChatGPT subscriptions.
- Job Market Indicators: An analysis of job postings at both companies reveals a substantial focus on enterprise sales and support roles, suggesting a concerted effort to scale their B2B operations.
- Challenging "AI Failure" Narratives: Willison disputes popular narratives about AI being too expensive or not providing clear ROI. He interprets instances like Uber's budget overruns and Microsoft's Claude Code cancellations as evidence of customers experiencing price sensitivity (the "suck air through their teeth" effect) rather than outright dissatisfaction or lack of utility.
- Massive Compute Investments: The article points to Anthropic's staggering $1.25 billion per month deal with SpaceX for compute capacity as an indicator of the enormous operational costs and scaling efforts involved, especially for inference.
- April 2026 Inflection Point: The author dubs April 2026 a new inflection point, following November 2025 (when agents became genuinely useful), as it marks the moment these advanced agents began generating significant, direct revenue for the frontier AI labs.
In essence, Willison posits that the era of discounted AI for enterprises is over, and major AI labs are successfully monetizing their powerful agent models by aligning enterprise costs with actual API usage. The true financial picture, however, awaits the transparency of their upcoming IPO S-1 filings.
The Gossip
Costly Competition and Open-Source Challengers
Many commenters expressed skepticism about the long-term viability of OpenAI and Anthropic's pricing model, particularly given the rapid advancements and cost-effectiveness of open-source models. While some believe premium models will maintain an edge by staying ahead of open-source alternatives, others predict that businesses will eventually opt for significantly cheaper, equally capable open-source solutions. The debate also touched on whether token costs truly reflect intrinsic value or are simply a mechanism for vendor lock-in.
Productivity Paradox and Enterprise Pennies
The discussion explored whether the reported productivity gains from AI agents (e.g., 20-40% faster development) genuinely justify the substantial increase in spending, which could amount to 5-20% of a knowledge worker's salary. Concerns were raised about how non-technical users' unrestrained use of AI agents could lead to 'astronomical' and economically nonsensical bills for companies, contrasting with engineers who are often more conscious of token usage. Companies are already responding by restricting access and emphasizing responsible token use.