Beijing-based AI company Z.AI is raising approximately $5 billion to fund next-generation model development and computing infrastructure, underscoring the mounting financial pressure on firms competing in frontier artificial intelligence. The fundraising plan, disclosed in a Sunday regulatory filing, includes roughly $2 billion through a Hong Kong share placement and another $3 billion via convertible bonds, according to Channel Insider. The company, previously called Zhipu AI, says the capital will largely support building more advanced models and expanding computing capacity as the cost of creating larger, more capable systems continues to climb.
Z.AI will issue 21.97 million new shares priced at HK$714 apiece, representing nearly a 10% discount from Friday's HK$793 closing price. Separately, the company will issue 20.14 billion yuan ($3 billion) in zero-coupon convertible bonds maturing in September 2027, carrying an initial conversion price of HK$892.50—roughly 12.6% above Friday's close. The bonds were issued at 100.5% of face value and will settle in U.S. dollars. This fundraising follows another $4 billion share placement completed less than two months earlier in July. First-half revenue jumped 400% to 953.89 million yuan (US$142 million), while annual recurring revenue reached $1.6 billion by the end of August. However, the company posted a 2.07 billion yuan first-half loss, and research and development spending climbed one-third to 2.13 billion yuan. Cash reserves stood at 3.99 billion yuan at the end of June. Shares closed at HK$793 after hitting an intraday peak of HK$2,980 on June 22, marking a roughly 73% decline from that high.
About 60% of net proceeds will fund research and development of next-generation models and a fully self-training system, according to Z.AI. Another 15% will support business expansion, with the balance earmarked for capital structure improvements, working capital, and general corporate purposes. The company released its GLM-5.3 flagship model in August, followed by the more affordable GLM-5.3-Flash multimodal model, and management has begun developing a successor with additional parameters, a longer context window, and native multimodal capabilities. The report notes that rising demand for Z.AI's services and ongoing work on new models have increased the company's computing capacity requirements.
The massive capital injection reflects the escalating economics confronting the broader AI sector, where rapid revenue expansion doesn't necessarily cover the expense of training and operating increasingly sophisticated models. Macquarie estimates Z.AI and rival MiniMax could remain unprofitable through 2030 as computing costs continue rising. The fresh funding gives Z.AI additional runway to absorb mounting training and operational expenses while competing against OpenAI and Anthropic globally, and DeepSeek, Moonshot AI, and MiniMax domestically. For enterprise technology providers and channel partners, the report emphasizes that companies able to sustain these infrastructure expenses are more likely to keep expanding model availability, pricing options, and partner ecosystems. Z.AI is also pursuing a Shanghai STAR Market listing, though no formal application has been publicly accepted yet. Separately, Tencent-backed Enflame nearly tripled in its Shanghai market debut after raising about $912 million, offering Asia-Pacific partners another potential domestic alternative to Nvidia for AI infrastructure.
The latest raise positions Z.AI to remain competitive in a race where simply staying current demands unprecedented financial commitment, but the company's widening losses illustrate how frontier AI development has become a capital endurance test few can sustain at scale. Companies that successfully navigate this transition will likely reshape how enterprises access and deploy advanced AI capabilities, while those that can't secure ongoing funding face the risk of falling behind as model capabilities and infrastructure demands accelerate. The willingness of investors to continue backing unprofitable AI ventures through this decade suggests the market views leadership in foundation models as strategically valuable enough to justify years of negative cash flow.