$700 Billion in Capex. $50 Billion in Revenue. AI’s Math Is Broken.

2026-05-07 · Source: High ROI AI · Field: Finance & Economics — Capital Markets & Investment Management, Corporate Finance & Treasury, Economic Analysis & Policy · Depth: Intermediate, medium

Summary

Major AI labs like Anthropic and OpenAI are experiencing unprecedented revenue growth and user adoption, yet face significant profitability challenges due to rapidly escalating costs. Anthropic, with $72 billion in cumulative funding and a $380 billion Series G valuation in February 2026, projects a $14 billion loss for 2026 and negative free cash flow until 2028. OpenAI, valued at $852 billion, achieved a $24 billion annualized run rate by April 2026 but posted a $13.5 billion net loss in H1 2025, with internal projections indicating a $74 billion operating loss in 2028. Both companies exhibit unit economics where costs scale faster than revenue, with gross margins around 48% after inference. The article highlights that exponential adoption is not a differentiator, switching costs are minimal, and open-source models offer competitive, lower-cost alternatives, creating a price ceiling for frontier APIs that drops 30-50% annually.

Key takeaway

For CTOs and VPs of Engineering evaluating AI investments, recognize that current frontier model valuations are likely unsustainable given their negative unit economics and the rapid commoditization driven by open-source alternatives. Prioritize solutions with proven profitability or a clear path to cost recovery, and factor in the declining price ceiling for API services when forecasting long-term expenditures. Avoid committing to platforms solely based on adoption curves, as switching costs are minimal and capability leads are short-lived.

Key insights

Despite exponential growth, major AI labs struggle with profitability due to costs scaling faster than revenue and rapid commoditization.

Principles

• P&L is ground truth for businesses.
• Adoption curves do not equal defensible value.
• Hardware deflation is outpaced by workload inflation.

In practice

• Self-host open-weights equivalents to reduce API costs.
• Evaluate agentic workflows for disproportionate compute consumption.

Topics

• AI Lab Valuations
• Unit Economics
• Open-Source Models
• Hyperscaler Capex
• Model Commoditization

Best for: Entrepreneur, CTO, VP of Engineering/Data, Investor, Director of AI/ML, Consultant

Related on AIssential

• $650 Billion on AI Capex — Hyperscalers are aggressively investing in AI infrastructure, leading to record capital expenditures and shifting market dynamics.
• AI's Brokenomics — AI labs' business models are unsustainable due to massive subsidies, high costs, and unproven ROI, leading to customer revolt and price cuts.
• Weekly Top Picks #112: Besides Moltbook — Frontier AI models are rapidly depreciating assets, with R&D costs often outpacing revenue generation before obsolescence.
• The Economics of AI Usage and What's Next For SaaS | Benedict Evans on a16z — Foundation models are commoditizing; value shifts to specialized AI applications.
• AI Is Too Expensive — AI's current economic model is unsustainable due to massive hyperscaler investments, high operational costs, and unproven profitability of AI labs.
• Two AI Companies Pass The Sustainability Test. The Rest Are Burning Cash. — AI business sustainability requires costs to scale slower than returns, a challenge due to high CapEx and OpEx.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by High ROI AI.