--- title: "Zhipu Take: RMB 1 trn mkt cap — Is China's Anthropic finally here?" type: "Topics" locale: "en" url: "https://longbridge.com/en/topics/42052740.md" description: "Against a persistently weak tape, Jun was a standout for Zhipu. News just before Jun about its Stock Connect inclusion sparked heavy flows and sharp intraday swings.From Jun 10 to Jun 18, the stock doubled in roughly a week. On Jun 22, $KNOWLEDGE ATLAS(02513.HK) crossed HK$1tn in market cap.This strength beat market expectations. In our prior earnings take, we noted Zhipu’s stronger momentum vs. $MINIMAX-W(00100.HK). Our core view: capital is paying up for the scarcity of model intelligence." datetime: "2026-06-22T12:29:04.000Z" locales: - [en](https://longbridge.com/en/topics/42052740.md) - [zh-CN](https://longbridge.com/zh-CN/topics/42052740.md) - [zh-HK](https://longbridge.com/zh-HK/topics/42052740.md) author: "[Dolphin Research](https://longbridge.com/en/news/dolphin.md)" --- # Zhipu Take: RMB 1 trn mkt cap — Is China's Anthropic finally here? Zhipu’s Jun outshone a persistently weak tape. First, the pre-Jun inclusion news (index eligibility) sparked sharp flows on the day. Then from Jun 10 to Jun 18, the stock doubled in roughly a week, and on Jun 22, $KNOWLEDGE ATLAS(02513.HK) broke the HK$1 tn market-cap mark.Such strength beat the market’s expectations. In our prior earnings take, we noted Zhipu’s greater momentum vs. $MINIMAX-W(00100.HK), with the core view that capital is pricing the scarcity of model intelligence. At today’s valuation, that remains a reasonable lens, but ‘intelligence scarcity’ alone is clearly not sufficient to explain the move. **1) Is ARR the driver?** Zhipu’s valuation framework had already shifted post-CNY rally toward an overseas B2B comp set (Claude/Anthropic). Looking at Anthropic’s ARR curve, there is no clear slowdown yet; YTD it even accelerated, with Apr/May monthly growth over 55%. This has expanded the market’s imagination for Zhipu’s growth path. In our last note, we offered a reference: assume Zhipu replicates Anthropic’s cadence and takes ~1 year (implying ~12% MoM) to reach $1 bn ARR; at Anthropic’s P/ARR multiple, implied EV would be ~$60 bn. But Zhipu’s market cap is already ~$150 bn. Back-solving with the same Anthropic P/ARR suggests implied ARR of ~$4 bn and a Mar–Jun MoM of ~150%, which looks unrealistic. Taking a step back and using Anthropic’s scale-up phase as a check: Anthropic went from $1 bn to $5 bn ARR in ~half a year, running ~30% MoM. If we assume Zhipu is already at ~$1 bn ARR now (which itself implies ~60% MoM over the past three months) and then grows another six months at ~30% MoM, ARR would reach ~$5 bn. On Anthropic’s multiple, that roughly squares with a ~$150 bn market cap.In other words, today’s valuation embeds at least two layers of expectations: (1) Zhipu’s ARR is already near ~$1 bn, and (2) MoM of ~30% can be sustained for the next half year. Yet its most recent annual report disclosed official ARR of only ~$250 mn. That implies ARR rising from ~$250 mn to ~$1 bn in three months and then holding ~30% MoM for six more months, a very demanding setup. While current ARR specifics remain unclear, the observable volume-price dynamics suggest Zhipu is in a phase of rising volumes and rising price. Hence, a rapid ARR uptrend looks well anchored.**Volume:** On OpenRouter, overall platform model calls continue to rise. For Zhipu, token calls on OR grew ~40% MoM on Avg. from Jan–Jun 2026 (Dolphin Research Est.), but only ~8% since Mar, partly because usage spikes are highly synchronized with new model releases, which typically drive about a month of intense calling before normalizing. **Price:** On headline API cards, there appears to be no hike (5.2 vs. 5.1). In practice, GLM-5.2 shifted from tiered pricing to a blended rate, moving volumes that previously enjoyed lower tiers to the higher rate, effectively a small hike. Relative to peers, Zhipu’s pricing is now near the upper bound among domestic models.We see two reasons why premium pricing holds: scarcity of intelligence underpins pricing power, and a primarily enterprise-facing mix means B-end clients focus on productivity gains from higher-intelligence models and are less price-sensitive. Comparing two biz. models, C-end exemplified by MiniMax/OpenAI vs. B-end by Anthropic/Zhipu, the latter has outperformed on valuation, with the B-end monetization narrative now largely proven. Anthropic has paired stronger intelligence with the highest pricing; once it hit the intelligence ceiling, users complained yet still paid, and Zhipu’s price hike on its Coding Plan in Feb was read as confidence in its model strength.By contrast, on the C-end, attempts like Doubao’s paid plan or OpenAI’s ads risk user backlash. Netting the volume/price analysis above, usage is indeed growing fast (though clearly below our earlier back-solve), and price has inched up implicitly, yet this still struggles to underwrite the core assumption that Zhipu is already near $1 bn ARR. Thus, the sharp rally looks more like multiple expansion. **2) Where does the multiple expansion come from?** **1) GLM-5.2 is the first domestic model to crack the global top-3 intelligence ranks.** On Jun 13, via its Coding Plan, Zhipu released GLM-5.2 and opened the API on Jun 17. GLM-5.2 is a 744 bn-parameter MoE model with 40 bn active params and a 1 mn-token context window, showing strong capabilities in coding and long-horizon agent workflows, with a sizable step-up vs. the prior gen, and it open-sourced weights under the MIT license.With GLM-5.2, Zhipu briefly ranked No.3 globally and No.1 in China on Artificial Analysis’s intelligence index, behind only Anthropic and OpenAI; its coding/agent scores were No.4/No.2 globally and No.1/No.1 in China (global ranks already edged down by Jun 22). On Arena.ai, GLM-5.2’s coding ranked No.2 globally, ahead of Opus 4.8 and behind only Fable 5. As open models narrow the gap with closed ones, founder Tang Jie publicly said Zhipu could surpass Anthropic within the year; given domestic pricing far below Anthropic’s (Zhipu API ~1/4 of Opus, ~1/10 of Fable 5), this materially boosted confidence in import substitution. **2) The U.S. shuts; China opens.** As Washington ordered top U.S. models to pause services overseas and Anthropic cited export controls to suspend Fable 5, Zhipu almost simultaneously released an open-source version. The contrast writes itself: while the U.S. tightens access to frontier tech, China in the same week released MIT-licensed, region-unrestricted weights.This contrast can lift multiples near term, but commercially GLM-5.2 does not directly benefit from the ban. First, Fable 5’s curbs are not permanent and reportedly are already being re-opened in stages. Second, with a short window and a market view that GLM-5.2 still trails Opus 4.8 in overall UX, this demand is unlikely to translate into meaningful revenue upside. Over a longer horizon, if Zhipu delivers a model on par with or surpassing Anthropic (notably, Zhipu was absent from Anthropic’s Feb list alleging model distillation by Chinese vendors), then coupled with recent moves by the U.S. and Anthropic to tighten access, frontier models may be viewed as strategic assets in a geopolitical contest. The U.S. has two champions, OpenAI and Anthropic, while China currently has only Zhipu, making a valuation premium reasonable. **3) Ultra-thin free float.** The truly free float is very low (on day 1, under 3% of total shares), and inclusion-driven passive demand compounded scarcity in the short term. As various lock-ups roll off in 2H, float should expand, so today’s scarcity is unlikely to be the norm. Putting the three factors above together, a premium multiple is justified. But against the ARR framework in Section 1, even in an optimistic case where ARR tops $1 bn by Sep (implying ~30% MoM), and benchmarking Anthropic’s Avg. MoM of ~20% from $1 bn to $9 bn, Zhipu’s ARR would be ~$1.73 bn by end-2026. The implied P/ARR would still be ~80x, nearly 2x Anthropic’s contemporaneous multiple even under bullish assumptions. **Risk disclosure and statements:**[**Dolphin Research Disclaimer and General Disclosure**](https://support.longbridge.global/topics/misc/dolphin-disclaimer) ### Related Stocks - [02513.HK](https://longbridge.com/en/quote/02513.HK.md) - [00100.HK](https://longbridge.com/en/quote/00100.HK.md) ## Comments (5) - **Trustworthy Horse · 2026-06-22T14:46:06.000Z**: Loss-making stocks, willing buyers take the bait, emotions pay the bill! Can't stand Hong Kong + A-shares, can play with US stocks for a bit Hong Kong + A-shares - **桑不起666 · 2026-06-22T13:19:30.000Z**: Continue to observe, there is no trustworthy A-share market. - **为什么不抄底 · 2026-06-22T12:57:02.000Z**: It will keep rising until you believe it, then you'll get hooked. It's probably not trapping people yet. - **打工为了加仓 · 2026-06-22T12:36:47.000Z · 👍 6**: The issue with Hong Kong stocks is not whether to rush in or not, it's the same as the A-share market, it's a question of trust. - **用戶_EtjDls · 2026-06-22T12:35:57.000Z · 👍 3**: Pump and dump, the fish that bites is willing.