The growing popularity of Chinese artificial intelligence models has emerged as a key risk factor for Anthropic ahead of its blockbuster initial public offering, according to US research firm Hedgeye Risk Management. “It’s all about Chinese competition when analysing Anthropic,” said Felix Wang, managing director and global tech sector head at Hedgeye.

“Many of my clients are asking the same question.”The assessment comes as Anthropic faces renewed investor jitters over whether its massive AI investments will deliver commensurate returns, with the company’s latest reported financial figures sparking sell-offs of leading AI chipmakers, including Nvidia and Intel, in recent days. Anthropic told investors at a private briefing last week that third-quarter revenue rose nearly 50 per cent from the previous quarter to US$16.8 billion, according to a person familiar with the matter. Despite the quarter-on-quarter jump, the figure marked a slowdown from the company’s explosive growth in the first half of the year, when revenue more than doubled in the second quarter as companies raced to purchase subscriptions for its powerful Claude Code coding agent tool.

Low-cost Chinese alternatives The deceleration has coincided with a sharp jump in the global popularity of Chinese open-weight models. According to AI research group Exponential View, open-weight models now serve around half of all AI “tokens” – the core unit of AI usage – globally.

While Chinese AI start-ups still significantly lag their US counterparts in total revenue, they have seen explosive growth in the last year. Beijing-based Z.ai, for instance, has seen its annual revenue run rate jump more than sixfold since March on the back of soaring demand for its GLM models. Crucially, the Chinese industry has managed to grow sales while steadily raising prices.

Z.ai, also known as Zhipu, has increased prices for its model services at least three times this year, while Hangzhou-based start-up DeepSeek more than doubled prices for its flagship V4 model in August. Even so, Chinese models remain among the cheapest on the market while trailing leading US models by only a few months in terms of capabilities, according to third-party benchmarking firm Artificial Analysis.

Total demand for your tokens may be going up, but if the price is flat or falling, then you’re not making more moneyFelix Wang, Hedgeye Risk Management This dynamic meant Chinese open models were not only taking market share from Anthropic but were also setting the floor for AI pricing, Wang said, pointing to Chinese start-ups Moonshot AI and DeepSeek as popular, cost-effective alternatives to Anthropic’s Claude.

Hedgeye’s internal models showed that Anthropic had experienced a third-quarter slowdown as growth in enterprise AI spending cooled, Wang said. Self-reported data from C-suite executives support this finding. Boston Consulting Group’s latest “Applied AI Index” found that nearly three-quarters of the world’s most sophisticated enterprise AI users had set usage limits or spending controls, even as they continued to promote internal adoption.

For Anthropic, this implied that even as token demand continued to increase, downward pricing pressure prevented the company from generating as much revenue as it might have without Chinese competition, Wang said.

Adding to Anthropic’s challenges is the recent resurgence of domestic rival OpenAI, which has seen rising demand for its coding tool Codex and new flagship model GPT-6 Astra.

Defence against distillation Anthropic’s IPO is expected next month, though some investors are now discussing a lower valuation of around US$1.5 trillion – a significant drop from the roughly US$2 trillion target the company previously sought, Reuters reported on Friday. Meanwhile, OpenAI has delayed its own IPO until next year. These headwinds have prompted some analysts to forecast a market meltdown for the US AI industry.

Earlier this month, Jefferies’ global head of equity strategy Chris Wood cited cheaper Chinese models as a key driver for his prediction of “massive capital destruction” in the US, arguing that cheaper Chinese alternatives were poised to dominate.

However, it remains unclear whether Chinese models can sustain their competitive edge against US rivals. Anthropic has repeatedly accused Chinese firms of illicitly improving their own models using Claude outputs – a practice known as “distillation” – but last week the company said it had made “substantial progress” in curbing the distillation of its models by as much as 99 per cent.

The firm also told investors that its crackdown on distillation had partly caused its recent revenue slowdown, with the campaign representing a 5 to 7 per cent hit to top-line figures, the Financial Times reported. Many China AI industry analysts have underestimated how reliant the country’s AI progress has been on distillation, according to Wang.

While Chinese companies would remain competitive, a successful clampdown by Anthropic could force the Chinese firms to invest more on research and development, hurting their margins, he added. “The China dynamic is very fast-moving when it comes to analysing Anthropic and OpenAI,” Wang said.