Every AI Subscription Is a Time Bomb for Enterprises

Major AI companies like OpenAI, Anthropic, and Google are currently taking losses while serving enterprise customers. This is a deliberate strategy.

The Gap Between Current Subscription Prices and Actual Costs

Claude Pro costs $20 per month, but when converted to actual API rates, it consumes tokens worth $200-400 per month per knowledge worker. In the case of GitHub Copilot, Microsoft is losing over $20 per user per month on the $10 subscription, with power users costing up to $80. According to some analyses, Anthropic users generate $8 in compute costs for every $1 of subscription revenue.

Agentic AI Is Destroying the Economics

During the chatbot era, token consumption was predictable, but agentic AI has completely changed the situation. Claude Code sessions run autonomously for long periods, consuming dozens of times more tokens compared to conversational use. There are reported cases of users exhausting their 5-hour limit in just 90 minutes.

GitHub announced that due to this problem, it will switch Copilot to usage-based billing starting June 1, 2026. This is because flat-rate models cannot sustain agentic workloads.

IPOs Will Trigger Price Increases

Both OpenAI and Anthropic are preparing for IPOs. Anthropic has surpassed $30 billion in annual revenue, while OpenAI stands at $25 billion. However, OpenAI is projected to incur $115 billion in cumulative losses by 2029, and has committed to $665 billion in compute spending through 2030.

Private companies can subsidize inference costs, but once public, investors will demand profitability. Price increases, usage limits, and conversion to metered billing are inevitable.

What Enterprises Should Do Now

• Audit actual token consumption by team (not just seat counts)
• Model 2x, 5x, and 10x pricing scenarios relative to current prices
• Minimize vendor lock-in
• Start proactive discussions with the CFO

According to a KPMG survey, average AI spending among U.S. enterprises is projected to reach $207 million over the next 12 months, double last year's figure. A Goldman Sachs survey found that many large enterprises are already exceeding their AI budgets by several times over.