Zhipu AI's Market Cap Rollercoaster

Zhipu AI's Market Cap Rollercoaster | Generated by AI

Home 2026.07

Question: How has Zhipu AI’s (listed as “Knowledge Atlas Technology”) market capitalization evolved over the recent year?

Answer:

Zhipu went from an obscure Beijing lab to one of the wildest re-ratings in Chinese equities within about six months. Timeline:

IPO → mid-2026 trajectory

Jan 2026: IPO on HKEX under ticker 2513.HK (entity name “Knowledge Atlas Technology”), priced at around HK$116.20 per share.
Mid-June 2026: Stock at HK$489 billion market cap, roughly 4x rival MiniMax’s HK$124.2B. This was already before the big spike — driven by Anthropic being ordered to suspend Fable 5/Mythos 5 access for foreign nationals, which handed Chinese open-weight models a demand vacuum.
May 29, 2026: Briefly hit HK$1,993 intraday, pushing market cap above HK$880 billion (~$112B), a nearly 1,600% gain since the January listing. Then a sharp pullback of HK$300-400 billion in market value evaporating within a few trading days.
June 22-23, 2026: Second, bigger spike on the GLM-5.2 launch — shares jumped as much as 42% intraday to HK$2,980, closing at HK$2,410, crossing HK$1 trillion (~$128B) market cap, at which point its valuation surpassed Meituan’s. GLM-5.2 ranked second globally on Code Arena for front-end web dev, behind only Claude Fable 5, and founder Tang Jie publicly told Elon Musk matching Claude “won’t take that long.”
July 8, 2026: Closed at HK$1,825, up nearly 1,500% since IPO. Zhipu then priced a secondary placement at HK$1,588/share to raise roughly $4 billion for R&D, scaling compute, and possible M&A, timed right after its six-month post-IPO lock-up expired.
As of July 7, 2026 (PitchBook, USD-denominated): stock price $205.28, market cap $91.5B, TTM revenue $101M, EPS -$1.49.

Fundamentals underneath the rally

2025 revenue rose 132% to 724 million yuan; adjusted net loss was 3.2 billion yuan. Net loss (unadjusted) actually widened 59.5% to 4.718 billion yuan, with operating loss of 3.737 billion yuan — R&D alone was 3.18 billion yuan, over 4x revenue.
MaaS/API platform ARR hit 1.7 billion yuan, up 60x year-over-year — this is the number bulls point to as proof of real demand, not just narrative.
JPMorgan forecasts revenue of 4.6B yuan (2026), 11.4B yuan (2027), 30.9B yuan (2028) — the year they project Zhipu turns profitable, having pulled forward their prior loss-forecast for 2028.
Its price-to-sales ratio is described as exceeding 1,000x by Korean analysts, well above even OpenAI comparables — bubble-risk framing you should weigh against the bull case.
On June 1, 2026, Zhipu announced plans for a secondary listing on the Shanghai STAR (A-share) board, offering 2-8% of post-issuance share capital — a second liquidity/capital-raising leg layered on top of the HK listing.

Why the moves are this violent — a UBS analyst attributed the lofty valuations to a scarcity premium and limited tradable shares from lock-up periods rather than pure fundamentals. Only under 4% of total shares are currently tradable, with 220 million shares from the first lock-up releasing in early July — that’s the mechanical overhang risk on the stock right now, separate from model performance.

Bottom line for your context: Zhipu (GLM series — the model powering a chunk of the cheap/open-weight tooling you’re likely already testing against DeepSeek-v4 and MiMo) is now priced roughly like a frontier lab despite <100M USD trailing revenue, purely on ARR growth rate + GLM-5.2’s Code Arena ranking + scarcity of tradable Chinese AI equity. Worth tracking GLM-5.2/5.3 benchmark trajectory as your leading indicator for whether the re-rating holds or reverts — same signal that would tell you if it’s worth benchmarking GLM against DeepSeek-v4-pro for your consulting stack.

References:

Back Donate