Zhipu Secures $5 Billion in Two-Part Financing, Doubling Down on Compute and Next-Gen Models
TREE NEWS reports: Chinese AI developer Zhipu announced on September 13 that it has completed a new financing round of approximately $5 billion (roughly RMB 33.5 billion), combining a $2 billion share placement with a $3 billion zero-coupon convertible bond. The move comes just two months after a $4 billion placement in July, marking the company’s third equity fundraising since its Hong Kong IPO in January.
The financing is split into two independent, non-conditional tranches. In the placement, Zhipu will issue up to 21.965 million new H-shares at HK$714 per share to no fewer than six institutional investors — a 9.96% discount to the September 11 close of HK$793 and a 19.95% discount to the five-day average of HK$891.90. The placement represents about 4.50% of enlarged share capital and is expected to net approximately HK$15.664 billion.
The convertible bond carries a principal of RMB 20.14 billion, matures in September 2027, and pays zero interest. Issued at 100.5% of par and redeemable at par, it has an initial conversion price of HK$892.50 per share — a 12.55% premium to the pre-announcement close. If fully converted, it would create about 26.365 million new H-shares, or 5.36% of enlarged capital, netting roughly $3.011 billion.
Structure Signals Long-Term Confidence
The zero-coupon, premium-conversion structure is notable: bondholders receive no interest and only benefit economically if the stock rises above the conversion price. This implies management expects the share price to appreciate meaningfully over the next two years and reduces near-term cash interest burden — a structure typically used by high-growth firms with strong equity narratives but heavy capital needs.
Where the Money Goes
- 60% — R&D for next-generation GLM foundation models and a “Fully Self Training” system, plus deployment and upgrades of large-scale training, production inference, and compute infrastructure.
- 15% — Business expansion, strategic investments, and potential M&A, including minority or controlling stakes in AI-complementary companies or assets.
- 25% — Capital structure optimization, working capital, and general corporate purposes.
The company expects to deploy all proceeds by June 30, 2028. It cites rising demand for training and inference compute from its GLM roadmap and its MaaS platform, and notes that favorable supply and delivery conditions for high-quality compute make now an opportune time to lock in capacity, given the lead time from contract to deployment.
Ownership and Lock-Up
After the financing, the largest single shareholder group — acting in concert through entities including Beijing Lianpai Technology Development Center — will see its stake diluted from about 28.58% to approximately 25.89% if the placement completes and the bond is fully converted. Public float will remain above the 10% minimum required by exchange listing rules. Zhipu has also committed not to conduct additional equity fundraising for 60 days after closing without underwriter consent.
Market Implications
For AI and semiconductor supply chains, the deal is a clear signal that top-tier Chinese model developers are entering a capital-intensive phase. The 60% allocation to compute and infrastructure reinforces demand for GPUs, domestic AI chips, data center capacity, and power — a positive read-through for hardware and cloud providers, though also a reminder of the sector’s cash burn.
For Hong Kong equities, the discounted placement may create short-term pressure on Zhipu shares and peers, as new supply hits the market at a discount. However, the zero-coupon convertible — with its premium conversion price — signals that institutional investors are willing to accept no yield in exchange for equity upside, a vote of confidence in the company’s long-term trajectory.
For the broader AI funding landscape, Zhipu’s rapid-fire raises — IPO in January, $4 billion in July, $5 billion in September — underscore a widening capital gap between a handful of surviving foundation-model players and the rest of the field. Competition is shifting from model benchmarks to compute access, cluster reliability, domestic chip integration, and cost per token. Scale of capital is becoming a moat in itself.
Key Takeaways for Investors
- Compute is the new battleground. Zhipu’s 60% allocation to infrastructure confirms that compute — not just algorithms — is the binding constraint for frontier AI.
- Convertible structure is a confidence signal. Zero coupon plus premium conversion means investors are betting on equity upside, not yield.
- Watch dilution and lock-ups. The placement discounts and 60-day fundraising freeze are near-term technical factors for Zhipu shares.
- Supply chain beneficiaries. GPU vendors, domestic AI chip makers, data center operators, and power suppliers stand to gain from this wave of AI capex.
- Consolidation ahead. With over $10 billion raised cumulatively, Zhipu is positioning for a shakeout that leaves fewer, better-capitalized players.
