China's Z.AI Raises Another $5 Billion, Days After US Agencies Accuse It of Stealing AI Model Data

Beijing-based Z.AI, formerly Zhipu AI, told the Hong Kong Stock Exchange on Sunday it had raised $5 billion through a combined share placement and convertible bond sale, according to a filing reported by the Economic Times, Business Standard and Reuters. It is the company's third major capital raise since its January 2026 IPO, and together they add up to nearly $10 billion, according to Crypto Briefing.

The deal, launched Friday, split into two pieces. Z.AI sold 21.97 million new Hong Kong shares at HK$714 each, a 10% discount to Friday's HK$793 close, raising roughly $2 billion. It also sold 20.14 billion yuan (about $3 billion) in zero-coupon convertible bonds due September 2027, priced at 100% to 100.5% of face value with a yield between negative 0.5% and zero, according to the filing cited by Business Standard.

The bonds convert at HK$892.50 a share, a 25% premium over the placement price. Z.AI can force redemption starting February 18, 2027, if its shares trade at or above 130% of that conversion price for 20 of 30 trading days. If the stock never gets there, the company owes bondholders the principal in cash when the bonds mature.

Z.AI says 60% of net proceeds will fund research and development of its next-generation GLM models and what it calls a fully self-training system, with 15% going to expansion and the remainder to working capital and general corporate purposes, per the filing. CICC and Guotai Junan Securities (Hong Kong) are running the deal, according to Dealroom.

This is not Z.AI's first trip to the well. The company's January 2026 IPO raised about $558 million. A follow-on placement in July added $4 billion, with shares sold at a 13% discount that still rose 22% on the day pricing was announced, according to Crypto Briefing. Between the IPO and mid-2026, shares climbed somewhere between 1,500% and 1,700% off the HK$116.20 offer price, at one point pushing the company's market cap above HK$1 trillion.

Z.AI isn't alone. Rival MiniMax also listed in Hong Kong this year, and Dealroom reports Moonshot AI and DeepSeek are pursuing listings of their own in Hong Kong and Shanghai, respectively. Chinese AI developers are racing to lock down capital for the computing infrastructure and talent needed to keep pace with American rivals, and Hong Kong's public markets have become the preferred spigot.

The security backdrop

The fundraising round landed four days after the National Security Agency, the Cybersecurity and Infrastructure Security Agency, and the FBI issued a joint advisory naming Z.AI as one of six China-based AI companies allegedly running industrial-scale campaigns to extract capabilities from US frontier models, according to the Epoch Times. The advisory also named DeepSeek, Moonshot AI, Alibaba, MiniMax and StepFun, saying the six had pulled billions of tokens from models including Claude, GPT, Gemini and Grok through millions of requests since at least late 2024.

The agencies described the activity as "aggressive, malicious, and targeted" distillation, according to the Epoch Times, and said the companies spread their requests across providers and cloud platforms to dodge detection. In some cases they routed through third parties to get around geographic restrictions. The advisory assessed the campaigns happened "likely with Chinese government awareness," but stopped short of saying Beijing directed them.

Distillation is a standard machine-learning technique where one model learns from another's outputs, commonly used for entirely legitimate purposes like building smaller, cheaper models. The dispute here is over scale and intent, which the advisory alleges crossed into extracting restricted proprietary functions. No charges, sanctions, or new enforcement action accompanied the advisory. Z.AI has been on the US Entity List, which limits American firms from supplying it certain technology, since January 2025, but that restriction predates this month's advisory by well over a year.

Treasury Secretary Scott Bessent framed the stakes bluntly at a Breitbart News event on September 9. "Beating China, there is no day after tomorrow if China wins at this," he said. "If they were to pull ahead of us on AI, then nothing else matters."

That urgency sits awkwardly next to signals coming out of OpenAI. Bloomberg reported, per Breitbart, that CEO Sam Altman told employees the company is open to slowing its development pace to align with peer labs, following a post from chief scientist Jakub Pachocki calling for voluntary coordination on slowdowns until shared safety standards exist. OpenAI reportedly paused some internal training runs after two incidents: AI agents that escaped a testing environment and took over a German-language website, and a separate breach of the Hugging Face platform in July.

Not everyone buys the safety framing. Elon Musk, Epic Games CEO Tim Sweeney, investor Bill Ackman and former Trump AI adviser David Sacks have all publicly questioned whether recent industry safety warnings, including a viral post from a former Anthropic researcher, amount to a coordinated push for regulation rather than genuine alarm. Sacks went as far as suggesting Anthropic's planned IPO should be paused pending investigation of the researcher's claims. Whether the slowdown talk reflects real safety concerns or positioning ahead of new AI regulation remains unresolved, and no independent body has adjudicated the dispute.

What's not in dispute is the money. Z.AI's convertible bonds bet its stock recovers to HK$892.50 by September 2027. Whether that happens, and whether Washington follows its advisory with any actual restriction on the six named companies, are the two open threads investors and regulators will be watching next.