OpenAI and Anthropic are growing at record speed, but massive computing costs are driving historic losses as the race to build advanced AI intensifies.

A new look into the financials of OpenAI and Anthropic reveals a striking reality: the AI boom is as expensive as it is fast-growing.

Explosive growth, but even bigger costs

Both companies are scaling rapidly, releasing new models faster and investing heavily in training them. But that progress comes at an enormous price.

  • OpenAI expects AI compute spending to hit $121 billion by 2028
  • Even with surging revenue, it could still spend around $85 billion in a single year
  • Losses at that scale would be unprecedented in corporate history

Anthropic is spending less overall, but faces the same trend: rising costs tied to increasingly complex AI models.

Profitability depends on how you measure it

Both companies report two different realities:

  • Without training costs → small operating profits possible
  • Including full AI compute → years of losses ahead

OpenAI does not expect to break even until the 2030s, while Anthropic may reach profitability sooner. This highlights a key issue: AI is profitable at the product level, but extremely expensive at the infrastructure level.

Anthropic pulls ahead in revenue

In a major shift, Anthropic has now overtaken OpenAI in annualized revenue.

  • Anthropic: $30B+ run rate
  • OpenAI: ~$24–25B run rate

The difference comes down to strategy.

  • Anthropic focuses on enterprise clients and APIs
  • OpenAI combines consumer products like ChatGPT with enterprise growth

Anthropic’s enterprise push is accelerating fast, with over 1,000 customers spending $1M+ annually.

Why costs are so high

The biggest driver is compute. AI companies spend heavily on:

  • Training new models
  • Running systems for users (“inference”)
  • Expanding infrastructure and data centers

Right now, inference costs alone consume over half of revenue for both companies. And there’s another challenge: Only a small percentage of ChatGPT users actually pay, meaning OpenAI supports a large base of non-paying users.

IPO plans and investor pressure

Both companies are expected to remain cash-burning for years, relying on investors to fund growth.

  • Anthropic is reportedly considering an IPO as soon as October
  • OpenAI is still debating timing internally

To support these massive capital needs, exchanges like Nasdaq are even adjusting rules to allow faster index inclusion, helping attract more investor money.

The AI race is entering a new phase. Revenue is exploding. Costs are exploding even faster.

  • OpenAI is betting on scale and consumer reach
  • Anthropic is winning through enterprise dominance

But for both, the same question remains: Can AI profits eventually outweigh the enormous cost of building intelligence itself?

Disclosure: This article does not represent investment advice. The content and materials featured on this page are for educational purposes only.

Related: What to Know About OpenAI’s Ideas for a World With ‘Superintelligence’