Anthropic, OpenAI Pursue IPOs as Enterprise AI Spending Faces Pushback

TLDR:

• OpenAI’s 2025 net loss hit $38.5 billion despite revenue tripling to $13.07 billion overall.
• Uber, Amazon and JPMorgan now restrict employee AI usage after costs spiraled unexpectedly.
• Anthropic and OpenAI filed confidentially for IPOs, both targeting valuations near $850 billion.
• Chinese models DeepSeek and Kimi undercut Anthropic and OpenAI pricing in benchmark cost tests.

Anthropic and OpenAI are pushing toward public markets while facing mounting questions about AI spending sustainability.

OpenAI posted a $38.5 billion net loss in 2025, even as revenue tripled to $13.07 billion. Rising pay-per-token costs have prompted major employers to limit staff usage, raising doubts about near-term profitability for both companies.

Enterprise Costs Spark Internal Crackdowns

Several large corporations have begun restricting employee access to AI tools after expenses climbed sharply. Uber reportedly exhausted its 2026 AI budget by April and now caps spending at $1,500 per employee monthly.

Amazon told staff to avoid using AI tools without clear purpose, following reports that engineers ran automated agents to climb internal usage leaderboards.

JPMorgan circulated an internal memo this month addressing excessive AI spending across departments. Some employees reportedly generated AI bills exceeding their own monthly salaries.

These examples reflect a broader pattern among companies that adopted AI tools aggressively over the past two years.

One pricing shift illustrates the scale of the problem. Workato saw its Anthropic bill increase 700% in a single day after the company moved from flat-rate to pay-per-token pricing in May.

Workato’s chief information officer said earlier subsidized pricing had encouraged widespread adoption before actual costs became visible.

IPO Timing Collides With Spending Concerns

Anthropic and OpenAI filed confidentially for public offerings this month, both reportedly targeting valuations near $850 billion.

Neither company has reached profitability, and OpenAI’s losses nearly tripled year over year. In 2024, the company lost $5.09 billion, a figure that grew to $38.5 billion in 2025.

This timing creates friction for both firms as they court investor confidence. Public offerings typically require evidence of sustainable revenue growth, yet enterprise clients are simultaneously scaling back usage. The same trend driving corporate cost-cutting threatens the growth narrative needed for successful IPO valuations.

OpenAI is reportedly considering token price reductions to retain customers shifting toward Anthropic’s offerings. According to the Wall Street Journal, Anthropic’s Claude Code product helped push annualized revenue from $9 billion to $47 billion within five months. That growth has intensified competitive pressure between the two firms.

Competitive Pricing Pressure Intensifies From Chinese Models

Artificial Analysis tested major AI models on identical benchmark tasks, comparing total operational costs. Anthropic’s flagship model cost $4,811 to complete the test suite, while OpenAI’s model cost $3,357 for the same workload.

Chinese alternatives showed substantially lower costs in the same testing. DeepSeek completed the benchmark for $1,071, while Kimi finished for $948.

These figures suggest Chinese developers prioritize cost efficiency over matching premium-tier performance metrics.

Bain surveyed nearly 1,000 companies regarding AI return on investment, finding that 40% reported actual cost savings below 10%.

One investor told Axios that a corporate finance officer spent $500 million on Claude access in a single month before anyone noticed.

As Anthropic and OpenAI prepare investor pitches, enterprise customers are demonstrating measurable resistance to current pricing structures.