The financial environment for artificial intelligence is changing rapidly, and the consequences for major players like OpenAI and Anthropic could be significant. As this earnings season progresses, several companies, including Meta and Shopify, have pointed to rising AI and inference costs as a strain on their profit margins. Shopify mentioned that while economies of scale have provided some relief, they have been "partially offset by increased LLM costs."
This financial pressure coincides with the expected IPOs of OpenAI and Anthropic, both projected to surpass valuations of $800 billion. These expectations rely on the belief that these companies will retain their market share and pricing power. However, mounting evidence suggests that this belief may be faltering. The market is evolving quickly, with more affordable alternatives emerging, particularly from Chinese labs, which provide competitive models at significantly lower prices.
The Cost Gap Widens
Recent data shows a growing disparity in AI operational costs. A survey from CloudZero found that by 2025, 45% of companies expect to spend over $100,000 monthly on AI, a notable rise from just 20% the previous year. As organizations evaluate their AI spending, the total cost of adopting various models has become a key consideration.
Artificial Analysis, a benchmarking firm, has assessed the leading AI models based on their costs. For example, Anthropic's Claude is priced at $4,811 per workload, while OpenAI's ChatGPT costs $3,357. In comparison, Chinese models like Zhipu's GLM offer similar capabilities for only $544 — nearly nine times cheaper than Claude. This stark pricing difference is attracting attention from businesses seeking budget-friendly solutions.
Google's CEO Sundar Pichai echoed these concerns at the recent I/O conference, noting that many companies are quickly surpassing their annual token budgets. He highlighted the cost-saving potential of Google's less expensive Flash model, which could save large enterprise customers over $1 billion annually if they shifted significant workloads to it. The demand for affordable AI solutions is becoming increasingly urgent in the corporate world.
Emergence of Competitive Alternatives
The rise of cheaper alternatives is not merely a hypothetical issue. Chinese AI labs have rapidly improved their model performance. DeepSeek, which previously sparked a tech selloff in the U.S., recently showcased a model that competes with the latest versions from OpenAI and Anthropic in coding and knowledge benchmarks. Other Chinese labs, including Moonshot and Zhipu, have also introduced competitive models.
Databricks CEO Ali Ghodsi shared insights into how businesses are adapting to this shift. He described the development of an "advisor model," where companies primarily use a cost-effective open-source model for most tasks but turn to advanced models from OpenAI or Anthropic only when necessary. This hybrid strategy enables firms to manage and lower costs effectively.
The speed of this transition is striking. Data from OpenRouter reveals that the usage of Chinese models jumped from just 1% in 2024 to over 60% by May 2025, reflecting a significant change in enterprise preferences.
American Response and the Trust Factor
In light of these trends, American companies are beginning to adjust. Nvidia, for example, is promoting its own AI systems that can be run on private servers, providing an alternative to both Chinese solutions and the proprietary models of OpenAI and Anthropic. Companies like Reflection AI are also gaining traction, focusing on developing American open-source models to meet the needs of enterprises seeking domestic options.
Trust is a critical factor for U.S. enterprises, especially in regulated industries. Cohere's CEO Aidan Gomez pointed out that while many clients are cautious about Chinese models, this concern might not apply across the entire enterprise market, where justifying a premium for U.S. models is becoming increasingly difficult.
The Future of AI Valuations
As OpenAI and Anthropic gear up for their public offerings, they face rapidly changing market dynamics. The anticipated premium that supports their high valuations is diminishing, particularly in sectors where they need to maintain their edge. An anonymous enterprise AI executive noted that while demand for AI is growing, it could be even higher if companies were not turning to cheaper alternatives.
OpenAI offers a different view, claiming that each new model release has driven a surge in API and product usage. They argue that while open-source models play a role, they do not significantly threaten their core business. Nonetheless, the pressure from lower-cost competitors is clear, and the implications for future valuations and market positioning could be significant.
As the AI market continues to evolve, the upcoming IPO filings from OpenAI and Anthropic will serve as a crucial test for how investors evaluate the changing dynamics of pricing, competition, and enterprise demand in the AI sector.
