US artificial intelligence giant Anthropic’s initial public offering prospectus has exposed a vast profitability gap between it and rivals Z.ai and MiniMax, highlighting the challenges facing China’s AI model developers in turning technological advances into sustainable businesses.

The Claude developer’s filing with the US Securities and Exchange Commission (SEC) last week revealed a significant financial advantage over its most direct Chinese counterparts, despite the latter’s continued advances in model capabilities, analysts said.

Anthropic reported a gross profit margin of 85.4 per cent in the first half of 2026, compared with 26.4 per cent for Z.ai, also known as Zhipu AI, and 17.9 per cent for MiniMax, according to Anthropic’s S-1 prospectus and the two Hong Kong-listed companies’ latest financial disclosures.

Anthropic generated US$16.3 billion in revenue during the first six months of the year, compared with US$141 million for Z.ai and US$116.6 million for MiniMax.

The gap was also evident in the companies’ ability to monetise computing resources, a major expense for developers training and running advanced AI models, HSBC analysts Ritchie Sun, Charlene Liu and Wang Peishan wrote in a research note published on October 27.

For every US dollar spent on computing resources, Anthropic generated US$2.08 in revenue in the second quarter, compared with 46 US cents for Z.ai.

The HSBC analysts linked Anthropic’s stronger returns on computing investment partly to its progress in recursive self-improvement, in which AI models increasingly help build more powerful versions of themselves, while noting that Z.ai had also begun to see similar gains.

Z.ai’s revenue generated per dollar of compute increased 23-fold from the first half of 2025, compared with a 4.2-fold increase for Anthropic over the same period.

The test is whether a Chinese lab can launch its next top model at a higher price, and whether customers pay itPoe Zhao, tech analyst“Z.ai also mentioned that GLM started to support inference engine optimisation and training of [the] next generation of GLM,” the analysts wrote, rating its stock a “Hold” with a target price of HK$1,375.

Z.ai and MiniMax have seen their Hong Kong shares slide around 13 per cent and 2 per cent, respectively, since Anthropic’s prospectus filing.

Poe Zhao, a Beijing-based tech analyst and founder of Hello China Tech, said the financial disparity between Anthropic and the two Chinese AI model companies reflected differences in pricing and scale.

“The simplest test is whether gross profit can pay for research,” said Zhao. “In the first half of 2026, Anthropic’s gross profit covered its R&D spending about 1.4 times. Zhipu’s covered about one-eighth, and MiniMax’s about one-fourteenth.”Zhao said gross margins were largely determined by pricing, while the ability to cover research and development costs depended more heavily on scale.

“All three companies must pay to stay near the frontier, and only Anthropic has the revenue to spread that cost,” he said.

Chinese AI developers have increasingly competed on price to attract users, but Zhao questioned whether that strategy could sustain the enormous investment required to develop frontier models.

“Low prices can win usage, but at today’s margins they cannot fund research,” he said. “For now, the capital market pays for the gap.”He added that technological improvements for Chinese AI companies did not necessarily translate into greater pricing power, given fierce competition both at home and abroad.

Unlike Anthropic, which committed to computing capacity 12 to 18 months in advance, the Chinese developers relied more heavily on purchasing computing resources as a service, leaving them with less profit to reinvest, he added.

While Anthropic also pays external providers for computing power, its longer-term commitments help keep costs predictable, and its much larger revenue allows it to absorb those expenses more easily. Meanwhile, Chinese AI developers such as Z.ai and MiniMax face tighter profit margins as they pay cloud providers to run their models.

Nevertheless, Anthropic also faces significant financial pressures despite its greater scale and pricing power, based on its performance in the latest quarter.

In an amended IPO filing submitted to the SEC on Monday, the Claude developer reported about US$17 billion in third-quarter revenue, up 47 per cent from the previous three months.

However, it recorded an operating loss of US$567 million, reversing a US$974 million operating profit in the preceding quarter, as rising computing expenses and slowing revenue growth weighed on its bottom line.

Preparing for what could be the largest IPO in history, the company faces mounting scrutiny over its balance sheet and heavy spending, with investors questioning whether rapid AI growth could justify its staggering capital demands.

Institutional investors were urging Anthropic to target an IPO valuation of US$1.5 trillion to US$1.6 trillion, down from the US$2 trillion the start-up had previously reportedly sought, the Wall Street Journal reported on Tuesday.