Chinese models charge 5-9x less than frontier labs. Enterprise 'advisor model' cuts premium usage. OpenAI IPO filing may show eroding pricing power.
This earnings season, rising AI costs showed up in the numbers. Meta, Shopify, Spotify, and Pinterest all flagged higher inference spending as a margin drag. Shopify stated that economies of scale were "partially offset by increased LLM costs."
The pricing model behind OpenAI and Anthropic's expected IPO valuations – each projected above $800 billion – assumes those labs will hold market share and charge a premium because alternatives cannot match capability. That assumption faces a direct challenge from a wave of cheaper, capable models. Chinese labs charge a fraction of American prices for comparable work. Western challengers – Nvidia, Cohere, Reflection, Mistral – are building smaller, more efficient models for enterprises that avoid Chinese providers. By the time OpenAI files its confidential prospectus, expected as soon as this week, the core valuation premise may already be eroding.
AI benchmarking firm Artificial Analysis runs each major model through ten evaluations and tracks the total cost. For each lab's most capable model, the gap is stark:
Claude is nearly nine times more expensive than the cheapest Chinese alternative for the same workload.
Google made the same point at its I/O developer conference. CEO Sundar Pichai said "many companies are already blowing through their annual token budgets, and it's only May." He pitched Google's cheaper Flash model as the solution. If the largest Google Cloud customers shifted 80% of workloads from frontier models to Gemini 3.5 Flash, Pichai claimed, they would save more than $1 billion a year. The company is acknowledging that enterprises need cheaper options.
Enterprise AI budgets have surged. Some 45% of companies surveyed by cloud cost firm CloudZero said they spent more than $100,000 a month on AI in 2025, up from 20% the prior year. Where that money goes increasingly matters. The cost gap is driving a shift in purchasing decisions.
Databricks CEO Ali Ghodsi has a real-time view of enterprise behavior. Databricks' AI gateway sits between thousands of customers and the models they use. Revenue from that product is climbing sharply. The technique enterprises deploy is called an "advisor model." A cheap open-source model handles the bulk of work as the default. When it hits a task it cannot solve, it calls out to a frontier model from OpenAI or Anthropic for help.
Key insight: The advisor model proves that enterprises can cut frontier spend without losing capability. That directly caps the addressable market for premium APIs.
The speed of adoption is visible on OpenRouter, a marketplace where developers access hundreds of AI models through a single interface. Chinese models went from about 1% of usage in 2024 to more than 60% in May.
Figma CEO Dylan Field described three phases of AI adoption: first, nobody uses it; second, everyone has to, with some "literally holding competitions of who can spend the most with tokens"; third, the realization that "everyone's spending too much" and must cut back. Many enterprises, he said, are now entering that third phase. Figma is selling features that cut customers' token consumption by 20% to 30%.
The cost gap reflects structural differences. American frontier labs run on hundreds of billions in capex, training larger models on the most expensive Nvidia chips inside a power grid straining to add capacity. Those costs pass through to customers. Chinese labs face chip export restrictions and have turned constraint into strategy: training competitive models with less compute and running them more efficiently.
DeepSeek, whose earlier model triggered a U.S. tech selloff in 2025, released a preview of its next-generation model last month. It matches or nearly matches the latest from OpenAI, Anthropic, and Google on coding, agentic, and knowledge benchmarks. Models from Moonshot, Xiaomi, and Zhipu have shipped at similar capability in the past four months.
On OpenRouter, Chinese models now account for over 60% of usage. The U.S. government's AI Safety Institute documented that DeepSeek downloads have risen nearly 1,000% since the R1 release in January 2025, even as officials flagged security concerns.
Cohere CEO Aidan Gomez, whose company sells AI models specifically to banks, defense agencies, and other regulated industries, says those buyers will not touch Chinese models regardless of price. Cohere's revenue grew sixfold last year selling into exactly that segment. That is a relatively narrow slice of the broader enterprise market. Outside regulated industries, where security and compliance rules are looser, the case for paying a premium gets harder to make.
Nvidia (NVDA), the company that has profited most from the AI boom, now publicly pushes a different model. Nvidia releases its own AI systems that any company can download and run on its own servers free of charge – an alternative to both Chinese options and the locked-down models from OpenAI and Anthropic. Reflection AI raised at a multibillion-dollar valuation specifically to build American open-source models for enterprises that want a domestic alternative. Both are well-capitalized and target the same gap: capable models, cheaper than the frontier, deployed on infrastructure U.S. enterprises already trust.
(AlphaScala's proprietary data gives NVDA an Alpha Score of 66/100, labeled Moderate, with a current price of $224.40, up 1.72% today. For the full profile, see the NVDA stock page.)
Anthropic itself acknowledged the pressure in a policy paper released in May. The company said U.S. models are only "several months ahead" of Chinese ones and warned that Beijing is "winning in global adoption on cost."
OpenAI sees it differently. A person familiar with the company's thinking said every release of a new frontier model, including GPT-5.5 last month, has driven a surge in API and product usage. Enterprise demand is growing in what they described as a "vertical wall." Open source has a role in low-stakes tasks, this person said. Pricing pressure is not on the company's top ten list of concerns.
An enterprise AI CEO, who asked not to be named to protect customer relationships, offered a different read. The growth is real. It would expand even faster for frontier if the advisor technique was not used.
This is the market OpenAI and Anthropic are expected to ask public investors to value. At nearly trillion-dollar valuations each, the S-1 must show enterprise revenue growth and concentration that justifies the multiple. The premium that justifies the valuation is eroding fastest in exactly the segments the labs need to dominate.
What would reduce the threat:
What would make the risk worse:
The next concrete marker is the confidential IPO filing from OpenAI, expected as soon as this week. The prospectus will reveal whether enterprise revenue concentration and growth rates support the valuation assumptions that the market has priced in. For investors tracking the AI sector, the stock market analysis page provides ongoing coverage.
Prepared with AlphaScala research tooling and grounded in primary market data: live prices, fundamentals, SEC filings, hedge-fund holdings, and insider activity. Each story is checked against AlphaScala publishing rules before release. Educational coverage, not personalized advice.