Hong Kong AI Market 2026: China’s AI Listing Capital

In the second week of January 2026, Hong Kong did something no other market outside the US has managed: it floated two frontier AI labs in 48 hours. Zhipu AI (Z.ai) raised HK$4.17 billion (about US$558 million) on 8 January at a US$7.1 billion valuation, with retail investors oversubscribing the offer 1,159 times (Caixin, January 2026). A day later, MiniMax — the company behind video generator Hailuo AI — netted US$619 million and surged 109% on debut (AFP via Malay Mail, January 2026), ending its first session at a market cap of roughly US$14 billion (SCMP, January 2026).

That is the Hong Kong AI story in 2026. The city is not primarily a place where foundation models get built — it is the place where China’s AI industry gets funded, listed and valued. Hong Kong topped global IPO fundraising in 2025 with nearly HK$275 billion (US$35.4 billion) across 106 offerings, ahead of the NYSE and Nasdaq, with HKEX itself crediting “tech trends, sparked by Chinese artificial intelligence start-up DeepSeek” for pulling global capital back in (HKEX data, December 2025). Around that capital-markets engine sits a genuine local ecosystem: roughly 500 AI-related organisations, including 290 companies and 180 active investors (FSDC, 2025), a government-built supercomputing centre running near capacity, and a sovereign LLM already used across most government departments.

Here is how the market breaks down — the money, the players, and where adoption actually stands as of July 2026.

How big is Hong Kong’s AI market in 2026?

Hong Kong’s AI economy is best measured in capital rather than software spend. The Financial Services Development Council counts around 500 AI organisations in the city — 290 companies across 14 industries, 180 investors and 30 innovation hubs — with DeepTech (52 companies) and AI infrastructure (48) the largest clusters (FSDC, 2025). The same FSDC analysis flagged a 23% jump in HKEX IPO volume in Q1 2025, “largely fuelled by AI companies”.

On the investor side, the money is strikingly international: of 183 active AI investors tracked by the FSDC, 46 are US-based, 42 from Hong Kong, 17 from mainland China and 17 from Singapore. That mix is precisely why Chinese AI firms list here — Hong Kong is where dollar capital meets China’s model builders. For how the city fits into the region’s wider spend, see our Asia AI market country breakdown.

Hong Kong is now China’s AI listing capital

The January IPOs were the proof point. Zhipu AI priced at HK$116.20 per share, raised HK$4.17 billion, and committed 70% of proceeds to model R&D through 2028; its shares jumped a further 20.6% on day two (Caixin; AFP, January 2026). MiniMax, founded only in 2022 and claiming around 200 million users, doubled on debut after raising US$619 million (AFP via Malay Mail, January 2026). SCMP put the pair’s post-listing worth at roughly US$9 billion (Zhipu) and US$14 billion (MiniMax) — small next to OpenAI’s ~US$500 billion and Anthropic’s ~US$350 billion private marks, but the first time public investors anywhere could buy a pure-play frontier LLM developer (SCMP, January 2026).

The read-through matters beyond two tickers. HSBC analysts expect “2026 to be the initial year for Chinese enterprises to gradually shift investment focus from model and compute towards AI applications”, with monetisation accelerating (via SCMP, January 2026). Hong Kong is where that monetisation gets priced.

Who is next in the 2026 AI IPO pipeline?

Moonshot AI — maker of the Kimi models — plans a Hong Kong listing within six months, on the back of a funding round valuing it at just over US$30 billion (Bloomberg, July 2026). Infrastructure specialist StepFun is widely tipped as another contender (Asia Tech Review, 2026). The notable absentee is DeepSeek: it is pursuing a Shanghai STAR Market listing as early as Q2 2027 at a valuation of US$50–71 billion, after a US$7.4 billion June round backed by Tencent, JD.com and CATL (Fortune, July 2026). The emerging pattern, as Fortune frames it, is that “national champion” labs central to China’s chip-substitution effort list on the mainland, while internationally facing companies pick Hong Kong for dollar-investor access — a dynamic we unpack in our China AI strategy pillar.

What is the Hong Kong government spending on AI?

Three commitments anchor the public-money side. First, the HK$3 billion, three-year AI Subsidy Scheme funds access to Cyberport’s Artificial Intelligence Supercomputing Centre (AISC), covering an average 70% of list prices, with nearly 20 universities, institutions and companies approved so far (news.gov.hk, January 2026). The AISC itself — Hong Kong’s largest AI compute facility at 3,000 petaflops (Cyberport) — began phased deployment in December 2024 and has averaged close to 80% utilisation since, with Cyberport at points reporting more than 90% of capacity in use (news.gov.hk, January 2026; SCMP, 2025). Demand, in short, is outrunning supply.

Second, the HK$1 billion Hong Kong AI Research and Development Institute, first announced in the 2025-26 budget, is now real: former Microsoft executive Professor Harry Shum chairs its 14-member board, which held its first meeting in spring 2026 with a mandate to push “AI for all” and turn academic research into industrial deployment (SCMP, 2026). Third, infrastructure is heading north — a Sandy Ridge data facility cluster in the Northern Metropolis is slated to deliver roughly 250,000 sq m of gross floor area from 2029 (CBRE, 2026), alongside plans for a second supercomputing centre on the same site (SCMP, 2025).

HKGAI V1: the sovereign model most governments only talk about

Hong Kong is one of the few markets in Asia with a working “sovereign LLM” in production. The Hong Kong Generative AI Research and Development Center (HKGAI), an InnoHK-funded venture led by HKUST under Professor Yike Guo, launched HKGAI V1 in 2025 as the city’s first locally developed large language model. Its internal tool is already used by more than 70% of government departments (SCMP, 2025), and its consumer assistant HKChat passed 720,000 registered downloads within months of its November launch (HKUST, April 2026). At InnoEX 2026 the centre showed seven home-grown applications, from the HKPilot office suite to legal tool LexiHK — a deliberate positioning of Hong Kong as a builder, not just a banker, under the government’s “AI+” strategy.

How are Hong Kong’s banks adopting generative AI?

Finance is the adoption engine, and the HKMA has built the most structured GenAI regime in Asia. The second cohort of its GenA.I. Sandbox, announced in October 2025, selected 27 use cases from over 60 proposals, involving 20 banks and 14 technology partners, with a heavy tilt towards “AI vs AI” quality control and deepfake-defence testing; trials began in early 2026 on Cyberport’s AISC (HKMA, October 2025). In March 2026 the programme expanded into GenA.I. Sandbox++, a joint initiative of the HKMA, SFC, Insurance Authority and MPFA covering banking, securities, insurance and pensions — including free GPU access at the AISC for participants (HKMA, March 2026; Charltons Law).

The model layer is pragmatically mainland-leaning: DeepSeek’s early-2025 breakout was credited by HKEX itself with reigniting global interest in Chinese tech listed in Hong Kong, and the city’s banks and institutions have been building on Chinese open-weight models alongside Western APIs. Which models are actually usable in the city — the split-stack question — is covered in our companion piece on AI access in Hong Kong, so we won’t rehash it here.

Enterprise adoption is broad but shallow

The honest caveat in the Hong Kong story is execution. The Deloitte-HKU AI Adoption Index 2026, surveying more than 100 C-suite executives across Hong Kong and mainland China, found 56% of organisations in the “implementing” phase but only 8% at company-wide transformation — and just 23% with operational AI deployments showing measurable financial impact (Deloitte-HKU, January 2026). Some 45% of executives said AI initiatives fell short of expectations, and 9% reported negative returns. Tellingly, the top barriers were organisational and cultural (50%) rather than technical (39%). Customer service (58%) and marketing (54%) lead current use cases.

That gap — record capital and compute on one side, thin enterprise ROI on the other — is the spread to watch. If Chinese AI vendors listed in Hong Kong start converting regional enterprises the way they are moving into Southeast Asia (see our analysis of Chinese AI models expanding across the region), the adoption numbers catch up to the valuations. If not, 2026’s IPO pops will look frothy in hindsight.

What could slow Hong Kong’s AI market down?

Three risks are visible. Power first: CBRE notes Hong Kong’s data-centre growth is capped at roughly 6-8% annually by electricity constraints, not demand, while power-rich markets like Malaysia scale faster (CBRE, 2026). Second, the listed labs carry real financial risk — Zhipu reported a net loss of RMB2.36 billion in the first half of 2025 (Caixin, January 2026), and MiniMax faces a US$75 million copyright suit from Disney, Universal and Warner Bros. Discovery over Hailuo AI (AFP, January 2026). Third, open-source pressure: DeepSeek’s free-to-run models squeeze the pricing of every listed competitor — Zhipu’s coding assistant already sells at around US$3 a month to compete (SCMP, January 2026). None of these derail the listing-hub thesis; all of them will decide which tickers survive it.

FAQ: Hong Kong AI market 2026

How many AI companies are there in Hong Kong?

The FSDC counts roughly 500 AI-related organisations in Hong Kong, including 290 companies across 14 industries, 180 active investors and 30 innovation hubs (FSDC, 2025). DeepTech and AI infrastructure are the largest company clusters.

Which Chinese AI companies are listed in Hong Kong?

Zhipu AI (Z.ai) and MiniMax both listed on HKEX in January 2026, raising US$558 million and US$619 million respectively (Caixin; AFP, January 2026). Moonshot AI, valued above US$30 billion, plans to follow within six months of July 2026 (Bloomberg, July 2026).

Will DeepSeek IPO in Hong Kong?

Current reporting says no — DeepSeek is pursuing a Shanghai STAR Market listing as early as Q2 2027, at a valuation between US$50 billion and US$71 billion (Fortune, July 2026), consistent with the pattern of “national champion” labs listing on the mainland.

What is HKGAI V1?

HKGAI V1 is Hong Kong’s first locally developed large language model, built by the InnoHK-funded Hong Kong Generative AI R&D Center led by HKUST. Its tools are used by more than 70% of government departments, and its consumer app HKChat has over 720,000 registered downloads (HKUST, April 2026; SCMP, 2025).

How much is the Hong Kong government investing in AI?

Core commitments include the HK$3 billion AI Subsidy Scheme subsidising compute at Cyberport’s 3,000-petaflop AI Supercomputing Centre, and the HK$1 billion Hong Kong AI Research and Development Institute chaired by Harry Shum, whose board convened in spring 2026 (news.gov.hk, January 2026; SCMP, 2026).

Are Hong Kong banks allowed to use generative AI?

Yes, within a supervised framework. The HKMA’s GenA.I. Sandbox second cohort involves 20 banks and 27 use cases (HKMA, October 2025), and the cross-regulator GenA.I. Sandbox++ launched in March 2026 extends supervised GenAI trials to securities, insurance and pension providers.