Z.ai Raises $5 Billion in Hong Kong

Read original on The Next Web (TNW)

#funding#foundation-models#china-aiz.aiz.aizhipu

A $5 billion raise gives Z.ai major firepower to challenge other foundation-model providers.

30-Second TL;DR

What Changed

Z.ai raised approximately $5 billion in a single financing move.

Why It Matters

The financing gives Z.ai substantial capital to expand model training, infrastructure, and commercialization. It also signals strong investor appetite for Chinese foundation-model companies despite market and geopolitical risks.

What To Do Next

Evaluate Z.ai’s API and model roadmap as an additional supplier if you need geographic or vendor diversification.

Who should care:Founders & Product Leaders

Key Points

•Z.ai raised approximately $5 billion in a single financing move.
•About $2 billion came from 21.97 million new shares priced at HK$714.
•The remaining $3 billion came from convertible bonds maturing in September 2027.

Key numbers$2 billion$714$793.10%

Deep Insight

Background and context from public sources — not the original article. 9 sources cited.

Enhanced Key Takeaways

•The $2 billion equity placement was executed at HK$714 per share, representing a 10% discount to its pre-deal closing price of HK$793.
•The $3 billion debt offering consists of zero-coupon convertible bonds maturing in September 2027, issued at 100.5% of face value with a negative yield to maturity (-0.5% to 0%) and an initial conversion price of HK$892.50.
•Around 60% of the net proceeds are allocated to R&D for next-generation foundation models, self-training systems, and non-Nvidia domestic compute infrastructure.
•This transaction marks Z.ai's third capital raise since its January 2026 Hong Kong IPO, bringing total post-IPO fundraising to approximately $9.6 billion alongside an approved dual-listing pipeline on Shanghai's STAR Market.
•Despite offering open-weights foundation models under the MIT License, Z.ai is currently pacing toward $1 billion in annual commercial sales.

Competitor Analysis

Z.ai (Zhipu AI)

Listing Status / Market: Public (HKEX: 2513.HK; STAR Market pending)
Core Foundation Model Family: GLM (General Language Model) Series
Compute / Hardware Approach: Fully domestic, non-Nvidia AI accelerator clusters

MiniMax

Listing Status / Market: Public (Hong Kong listed in 2026)
Core Foundation Model Family: MiniMax proprietary series
Compute / Hardware Approach: Mixed cloud & hybrid accelerator infrastructure

Moonshot AI

Listing Status / Market: Private (Pursuing HK/Shanghai listing)
Core Foundation Model Family: Kimi series
Compute / Hardware Approach: Commercial cloud infrastructure

DeepSeek

Listing Status / Market: Private (Pursuing HK/Shanghai listing)
Core Foundation Model Family: DeepSeek-V / DeepSeek-R series
Compute / Hardware Approach: Hybrid cluster architecture

Developer	Listing Status / Market	Core Foundation Model Family	Compute / Hardware Approach
Z.ai (Zhipu AI)	Public (HKEX: 2513.HK; STAR Market pending)	GLM (General Language Model) Series	Fully domestic, non-Nvidia AI accelerator clusters
MiniMax	Public (Hong Kong listed in 2026)	MiniMax proprietary series	Mixed cloud & hybrid accelerator infrastructure
Moonshot AI	Private (Pursuing HK/Shanghai listing)	Kimi series	Commercial cloud infrastructure
DeepSeek	Private (Pursuing HK/Shanghai listing)	DeepSeek-V / DeepSeek-R series	Hybrid cluster architecture

Technical Deep Dive

Domestic Accelerator Infrastructure: Transitioned large-scale cluster operations entirely away from Nvidia architectures to non-standard, domestic Chinese AI accelerators.
Foundation Architecture: Primary architectures stem from the GLM (General Language Model) family originally incubated at Tsinghua University in 2019, utilizing open-weights distribution under the MIT License.
Self-Training Systems: R&D capital target focuses on building out automated, end-to-end self-training systems to enhance foundation model capabilities autonomously.

Future ImplicationsAI analysis grounded in cited sources

Domestic silicon transition will drive capital expenditures above typical industry averages.

Deploying and optimizing large-scale foundation model training on non-standard domestic accelerators incurs higher initial engineering and cluster management overhead than mature Nvidia stacks.

Z.ai will execute a secondary listing on Shanghai's STAR Market.

The firm has already cleared regulatory tutoring and shareholder approval to target an additional RMB 15 billion on the mainland exchange.

Timeline

2019-11

Spin-out from Tsinghua University and initiation of the GLM model series

2026-01

Completes Hong Kong IPO, becoming China's first publicly listed pure-play LLM developer

2026-06

Share price peaks intraday at HK$2,980 on the HKEX

2026-07

Executes HK$31.4 billion (~$4 billion) post-IPO share placement

2026-09

Raises $5 billion via combined H-share placement and negative-yield convertible bonds

2019-11
Spin-out from Tsinghua University and initiation of the GLM model series
2026-01
Completes Hong Kong IPO, becoming China's first publicly listed pure-play LLM developer
2026-06
Share price peaks intraday at HK$2,980 on the HKEX
2026-07
Executes HK$31.4 billion (~$4 billion) post-IPO share placement
2026-09
Raises $5 billion via combined H-share placement and negative-yield convertible bonds