AI company Zhipu (02513.HK), just eight months after listing on the Hong Kong Stock Exchange, has once again tapped capital markets for a major fundraising round. The company announced on September 13 that it has completed a new financing round of approximately $5 billion (approximately RMB 33.5 billion), coming just two months after its previous placement in July that raised approximately $4 billion. The funding comprises approximately $2 billion in share placement and approximately $3 billion in zero-coupon convertible bonds, primarily earmarked for next-generation foundation model R&D and computing power infrastructure development.
This marks Zhipu’s third equity financing since its Hong Kong Stock Exchange IPO in January this year. From the IPO to two subsequent refinancing rounds, the pace of fundraising has continued to accelerate, with capital allocation remaining highly consistent—computing power, model R&D, and infrastructure. Across the three financing rounds, Zhipu has cumulatively raised over $10 billion from capital markets.
Discounted Placement Combined with Zero-Coupon Premium Convertible Bonds
The financing was conducted in two independent tranches, with neither conditional upon the other.
For the share placement, Zhipu placed up to 21.965 million new H-shares at HK$714 per share to no fewer than six institutional investors. The placement price represented a discount of approximately 9.96% to the closing price of HK$793 on the trading day prior to the announcement (September 11), and a discount of approximately 19.95% to the five-day average price of HK$891.90. The placement shares represent approximately 4.50% of the enlarged issued share capital, with estimated net proceeds of approximately HK$15.664 billion (approximately $2.0 billion).
For the convertible bonds, Zhipu issued zero-coupon convertible bonds due September 2027 with an aggregate principal amount of RMB 20.14 billion (approximately $3.0 billion), issued at 100.5% of principal and redeemable at par upon maturity. The initial conversion price was set at HK$892.50 per share, representing a premium of approximately 12.55% to the pre-announcement closing price and a 25% premium to the placement price. Assuming full conversion, the bonds would convert into approximately 26.365 million new H-shares, representing approximately 5.36% of the enlarged issued share capital. Estimated net proceeds from the bonds are approximately $3.011 billion.
The zero-coupon structure means bondholders receive no interest payments, while the premium conversion price means conversion only becomes economically meaningful if the share price rises further. This structure signals expectations for long-term share price appreciation to the market—the company is effectively trading away short-term interest costs in exchange for investor confidence in sustained growth of the company’s equity value.
Sixty Percent of Proceeds Bet on Computing Power and Next-Generation Models
According to the announcement, Zhipu plans to allocate net proceeds as follows:
| Use of Proceeds | Percentage |
|---|---|
| R&D for next-generation GLM foundation model and “fully self-trained” system, as well as large-scale training and inference computing power deployment and upgrades | Approximately 60% |
| Business expansion, strategic investments, and potential M&A | Approximately 15% |
| Optimizing capital structure, supplementing working capital, and other general corporate purposes | Approximately 25% |
The company expects the proceeds to be fully utilized by June 30, 2028.
The announcement noted that as the company continues to advance R&D related to the next-generation GLM foundation model, fully self-trained system, and long-horizon task reinforcement learning, along with sustained growth in demand for its MaaS platform, the need for training and inference computing resources has correspondingly increased. The company stated that “the current supply and delivery conditions for high-quality computing resources in the market are relatively favorable, and computing resources require a certain lead time from contract signing to deployment and go-live. Completing the financing arrangement at this stage helps align the pace of new computing power deployment with the company’s expansion plans.”
Share Dilution and Lock-Up Commitments
Upon completion of this financing, the single largest shareholder group (comprising Beijing Lianpai Technology Development Center and other concert parties) will see its shareholding gradually diluted from the current approximately 28.58%. Assuming completion of the placement and full conversion of the bonds, the group’s aggregate shareholding would decrease to approximately 25.89%. However, the company’s public float will remain above 10% of the enlarged issued share capital, in compliance with Hong Kong Stock Exchange listing rules.
The company also committed that from the date of the placement agreement until 60 days after the closing date, it will not undertake additional equity financing arrangements without the written consent of the joint placing agents.
Accelerating Fundraising: Second Move Within Two Months
This is part of Zhipu’s intensive capital markets activity.
According to Bloomberg, in July this year, Zhipu raised approximately $4 billion through a placement of approximately 19.8 million shares at HK$1,588 to HK$1,698 per share. At that time, computing power, R&D, and talent were similarly the primary uses of proceeds.
Earlier, in January this year, Zhipu completed its Hong Kong Stock Exchange IPO, issuing approximately 37.42 million new H-shares at HK$116.20 per share, with total net proceeds from the global offering (including exercise of the over-allotment option) of approximately HK$4.896 billion (approximately $624.3 million). According to disclosures, as of August 31, 2026, IPO proceeds had been fully utilized, and approximately 34.92% of the July placement proceeds had been deployed.
From a broader perspective, Zhipu’s intensive fundraising reflects the evolving competitive logic of the large language model industry.
The first round of competition in the LLM industry eliminated companies unable to continuously iterate their models. Today, a handful of players—Zhipu, DeepSeek, Kimi, MiniMax—have proven they can remain at the model table. But what the next round of competition demands is changing—not just model capability itself, but whether companies can secure sufficient computing power, stably operate large-scale clusters, integrate domestic Chinese chips, drive down training and inference costs, and most critically, whether they have sufficient capital to sustain all of this.
Models, talent, capital, and infrastructure—none can be missing. Zhipu’s $5 billion financing is, to a large extent, positioning for the next phase of competition. With R&D budgets and funding arrangements further clarified, Zhipu’s capital foundation for continuing training experiments, advancing model iterations, and refining its computing power systems has been strengthened, adding leverage for participation in the next phase of global frontier model competition.