How China ‘risk factors’ loom over Anthropic’s hotly anticipated IPO filing

Open-weight models, China-US tensions, Taiwan – just a few of the more than dozen references to China-related risks cited in Anthropic’s much-anticipated listing prospectus, which was finally released on Monday after months of speculation.

Mostly concentrated within the 93-page “risk factors” section of the filing with the US Securities and Exchange Commission, the disclosures reveal how the Silicon Valley start-up assesses the commercial threats posed by China and its artificial intelligence industry.

The warnings come as Anthropic prepares for what could become the largest initial public offering in history – and lays out plans for more than US$500 billion of AI infrastructure investment over the next 20 years.

The open-weight challenge

A central focus of the risk assessment is the rapid rise of open-weight models, whose underlying numerical parameters are released publicly for users to download for free and modify. China has emerged as a global leader in this domain, propelled by the success of start-ups such as DeepSeek and Z.ai.

While Anthropic and its CEO Dario Amodei have repeatedly warned of the safety risks posed by open-weight systems, the prospectus marks the first time the company has formally acknowledged the direct commercial threat they pose to its business.

“[Open-weight] or open-source AI models may capture a portion of the market for certain use cases, which could exert pricing pressure on our offerings or lead certain customers to adopt open-source alternatives for those use cases,” the company said.

Poe Zhao, a Beijing-based technology analyst and founder of Hello China Tech, said the filing underscored how Chinese competitors were exerting pressure on Anthropic, particularly among enterprise users who “prioritise cost efficiency over intelligence”.

Anthropic’s flagship Claude Opus 5.5 model, for example, has a token price around five times higher than Z.ai’s top-tier GLM-5.3.

Geopolitical friction and supply chains

Anthropic also identified tensions between Beijing and Taipei as a supply-chain vulnerability, pointing to the concentration of advanced chip manufacturing at Taiwan Semiconductor Manufacturing Company.

Beijing regards Taiwan as part of China to be reunited by force if necessary. Most countries, including the United States, do not recognise Taiwan as an independent state, but Washington is opposed to any attempt to take the self-ruled island by force and is committed to supplying it with weapons.

Amodei has been among Silicon Valley’s most outspoken China critics, having described the country as a “hi-tech surveillance state”.

However, the filing acknowledged that US export controls on advanced chip-related technology – moves that Anthropic supported with its own lobbying – could rebound on its operations, noting that trade barriers “could affect the price or availability of compute”.

The prospectus also addressed Anthropic’s recent crackdown on Chinese companies allegedly using outputs from Claude to train their own models – a practice known in the AI industry as “distillation”.

Anthropic disclosed that up to 7 per cent of its revenue in the first nine months of the year was tied to accounts “we subsequently determine are using our models, platform and applications in violation of our usage policy or commercial terms”.

The company cautioned that this revenue stream was likely to shrink after heightened usage enforcement in August and September. The clampdown coincided with tighter access curbs reported by users in mainland China and Hong Kong in recent weeks.

Surging growth amid massive spending

Despite commercial headwinds from China, Anthropic reiterated confidence in its growth trajectory, citing its dominance of market segments that demanded “the highest intelligence”.

Financial disclosures in the filing showed revenue reached US$16.3 billion in the first half of the year – a 14-fold surge compared with the same period last year.

The company held around US$132 billion in cash, cash equivalents and liquid investments as of June 30, a war chest surpassing those of hyperscalers such as Amazon.com and Meta Platforms.

While Anthropic posted a net loss of US$80.7 billion for the six-month period, US$72.4 billion of that stemmed from a one-off accounting adjustment, the company said.

Gross margins stood at about 85 per cent in the first half, significantly outpacing Hong Kong-listed Chinese competitors Z.ai – also known as Zhipu AI – at 26 per cent and MiniMax at 18 per cent during the same period.

“The companies do not measure these costs in exactly the same way, but the gap is very large,” Hello China Tech’s Zhao said.

Felix Wang, managing director and global tech sector head at Hedgeye Risk Management, said there was general relief among investors that Anthropic’s IPO was likely to happen soon.

“A lot of investors I talk to just want the IPO to happen now,” he said, while noting that Anthropic was unlikely to fetch the US$2 trillion valuation it had reportedly been seeking.

“This is why markets were up yesterday – a sigh of relief that the IPO process is moving forward, though nothing is guaranteed.”