Zhipu AI Completes $5 Billion Funding Round with Zero-Coupon Convertible Bonds at Premium Conversion
nashnova research
Zhipu announced on September 13 a roughly $5 billion fundraise — split between a share placement and zero-coupon convertible bonds — marking its third equity raise since listing in January and pushing its cumulative capital-market haul past $10 billion, a pace that underscores how large-model competition is becoming a capital-intensity contest.
How is the $5 billion structured?
The raise has two independent tranches: roughly $2 billion in share placement and roughly $3 billion in zero-coupon convertible bonds. Neither is conditional on the other.
The placement price is HK$714 per share — a 9.96% discount to the last closing price of HK$793 and a 19.95% discount to the five-day average.
The convertible bonds carry a total principal of RMB 20.14 billion, mature in September 2027, are issued at 100.5% of par, and are redeemed at par on maturity.
Zero coupon plus a conversion premium — what does this structure signal?
The bonds — zero-coupon convertible bonds (debt that pays no interest but can be swapped for shares later) — have an initial conversion price of HK$892.50, a 12.55% premium to the pre-announcement close.
This means → bondholders only profit from converting if the stock rises above HK$892.50; otherwise they simply get their principal back at maturity.
In plain terms = the company is telling the market: "We expect the share price to climb well past this level." In exchange for paying no interest, Zhipu accepts potential dilution down the road — a low-cost bet on its own future valuation.
Where will the money go?
About 60% goes to R&D on the next-generation GLM foundation model and a "Fully Self Training" system, plus large-scale training, inference compute, and infrastructure upgrades.
About 15% is earmarked for business expansion, strategic investments, and potential M&A; roughly 25% for capital-structure optimization and working capital.
Zhipu noted that supply conditions for premium compute resources are currently favorable, and that compute takes time from contract to deployment. This means → Zhipu is racing to lock in a procurement window before conditions shift.
How much dilution do existing shareholders face?
The placement shares represent about 4.50% of the enlarged share capital; full bond conversion would add another 5.36%.
The largest shareholder group (Beijing Lianpai Technology and concert parties) will see its stake drop from roughly 28.58% to about 25.89%, but public float stays above 10%, meeting Hong Kong listing rules.
Zhipu has committed to no further equity fundraising for 60 days after settlement without the placing agents' consent.
Three raises in eight months — what does the pace reveal?
The January IPO raised about HK$4.896 billion, now fully spent. The July placement raised roughly $4 billion, of which 34.92% has been used. This round adds another $5 billion. The three-round total exceeds $10 billion.
This reflects an evolving competitive logic in large models: securing enough compute, integrating domestic chips, and driving down training and inference costs are now the decisive variables.
In plain terms = the large-model race is shifting from "who has the best technology" to "who can burn capital fast enough — and accurately enough." Whether Zhipu can convert funding into a model-capability edge within its compute-deployment cycle is the test that will validate this round of capital.
市场有风险，内容仅供研究参考，不构成投资建议。