Open-Source AI Startup Zhipu Surges 33% After US Bans Anthropic’s Models

Founders and VCs watching the AI space just got a masterclass in geopolitical timing. Zhipu’s immediate release of an open-source model with no restrictions—right as the US locks down rival Anthropic—turned a regulatory move into a catalyst for a 33% stock surge, underscoring how agile startups can seize windows created by policy shocks.

Shares of Chinese AI model developer Zhipu, traded through Knowledge Atlas Technology, surged as much as 48% on Monday before closing 33% higher at HK$1,461 ($186), as Wall Street banks moved swiftly to upgrade the stock following a US government order that restricted foreign access to rival Anthropic’s most advanced AI models. The rally highlights a dramatic pivot in global AI leadership, where geopolitical tensions are now directly reordering the competitive landscape for foundation models. On Friday, the Trump administration ordered Anthropic to suspend all foreign national access—including its own non-citizen employees—to its powerful Fable 5 and Mythos 5 models, citing national security concerns. Within hours, Zhipu countered by announcing the open-source release of GLM-5.2, its most capable large language model, with no usage restrictions and a clear message that cutting-edge AI should be universally available.

“

Bank of America initiated coverage on both Zhipu and MiniMax with buy ratings, setting targets of HK$1,250 and HK$500 respectively—signaling a clear preference for Zhipu’s open-source strategy.

The timing triggered a cascade of analyst actions. JPMorgan, according to reports, maintained its overweight rating on Zhipu and hiked its target price to HK$1,400 from HK$950, citing the company’s model visibility and emerging pricing power in a contested market. The bank simultaneously downgraded Zhipu’s domestic rival MiniMax, underscoring a divergence in outlook for China’s AI leaders. Bank of America initiated coverage on both Zhipu and MiniMax with buy ratings, setting targets of HK$1,250 and HK$500 respectively—signaling a clear preference for Zhipu’s open-source strategy. The dual upgrades reflect a broader Wall Street reappraisal of which AI firms can thrive as the US tightens its grip on proprietary models.

Zhipu’s strategic positioning is reinforced by early technical feedback. Ellie Jiang, head of Asia internet and media research at Macquarie Capital, noted that preliminary community evaluations show GLM-5.2 performing comparably to Anthropic’s Claude Opus 4.7 in coding and long-horizon agentic tasks. This parity, achieved by an open-source model released without restriction, directly challenges the notion that the world’s most capable AI must be locked behind corporate or national firewalls. For enterprises and developers globally, GLM-5.2 offers a credible alternative that can be freely deployed, modified, and integrated—potentially accelerating adoption in markets wary of US dependencies.

The market impact extends beyond the immediate stock pop. Zhipu’s surge signals that investors now see Chinese AI firms not merely as regional players but as serious contenders for global model supremacy if they can deliver competitive open-source tools. The fact that JPMorgan raised its target by 47% (from HK$950 to HK$1,400) indicates conviction that the company can convert this window into sustained pricing power. MiniMax’s downgrade, meanwhile, suggests that not all Chinese AI companies will benefit equally; open-source ethos and model quality are becoming key differentiators.

The US has been incrementally tightening access to advanced AI technology for foreign entities, particularly Chinese firms, through chip export controls and model restrictions. Friday’s order against Anthropic marks one of the most aggressive moves yet, directly curbing the foreign developer base that contributes to and consumes these models. For Zhipu, which has historically positioned itself as an open-source champion with models like GLM-4 and GLM-5, the opportunity is clear. By offering unrestricted access just as the US clamps down, the company is not just selling a model—it’s making a statement about the future of AI governance. The rally in Knowledge Atlas Technology shares reflects investor belief that this stance will attract both developers and enterprises seeking independence from US technology.