Home/AI for Science/XtalPi (晶泰科技)
EN↔中文
Company

XtalPi (晶泰科技)

Shenzhen AI drug-discovery company founded on crystal-structure prediction — the first AI4S drug-discovery firm to list in Hong Kong.

1. Core Product / Service

XtalPi combines quantum physics, AI, and cloud automation to predict molecular structures and properties, anchored on its origin in crystal-structure prediction (CSP) — determining how a molecule packs into a crystal lattice, a step that is critical for drug solid-form selection. It has broadened into an "intelligent digital drug discovery" platform (ID4) spanning hit discovery through lead optimization, plus antibody discovery (XupremAb) and structure prediction (XtalFold) (Internal research notes).

The other half of the business is robotic wet labs: XtalPi runs AI-driven autonomous labs that close the "design → execute → learn" loop, and also sells lab-automation solutions to outside customers. In its 2026 interim presentation it regrouped the business into drug-discovery solutions and AI4S intelligent solutions, with the robotic lab framed as the shared infrastructure and the company repositioned as an "open scientific-intelligence infrastructure platform" (Internal research notes)[gelonghui 2026-10-12].

Lab scale (per its listing prospectus): 200+ robotic workstations with a company-estimated peak throughput of ~1,500 experiments per day; >10,000 m² of lab space, of which >4,500 m² is robotic-workstation area. All premises are leased — 17 sites totalling ~37,200 m², including a ~11,200 m² Shenzhen HQ and a ~15,500 m² Shanghai lab, plus a robotic lab in Somerville, MA (Internal research notes).

2. Target Users & Pain Points

  • Pharma and biotech — the company reports serving 17 of the top-20 global pharma companies, including Pfizer, Eli Lilly, Roche, J&J/Janssen, Merck and Bayer (Internal research notes) — needing solid-form screening, polymorphism prediction, and AI-assisted drug R&D.
  • Pain solved: experimental crystal-form screening is slow and material-hungry; CSP shifts it to computation, and downstream AI tooling compresses discovery timelines.

3. Competitive Landscape

Player Positioning Vs. XtalPi
insilico-medicine Generative-AI biotech (Pharma.AI) Insilico is AI-first biology; XtalPi is physics-first CSP origin
isomorphic-labs Alphabet AI drug discovery US big-tech-backed; XtalPi is independent + China-listed
dp-technology AI molecular simulation (China) Closer to simulation infra; overlaps on drug design
Schrödinger Physics-based computational chemistry US-listed incumbent with a longer track record
megarobo Lab-automation integrator (China, private) MegaRobo sells custom automation projects (larger lab-automation revenue, but 2/3 of total revenue is industrial); XtalPi sells outcomes and has a capital-market channel
药明康德 (WuXi AppTec) Full-stack traditional CRO WuXi is services-at-scale across chemistry/bio; XtalPi is the AI-driven, physics-first design layer

4. Unique Observations

  • CSP as a wedge: crystal-structure prediction is a niche where physics-based accuracy is verifiable in the lab, giving XtalPi a defensible early moat before it broadened into full drug discovery.

  • Pfizer validation: XtalPi's early high-profile collaboration was with Pfizer on crystal-form prediction — an early external validation of its physics+AI approach (Internal research notes).

  • China AI4S capital-market benchmark: as the most prominent listed Chinese AI4S drug-discovery company, XtalPi is the public-market reference point for the sector's "platform revenue real, blockbuster drug still absent" reality (Internal research notes).

  • AI-CRO vs full-service CRO (analysis): compared with WuXi AppTec by customer type and revenue mix, XtalPi competes as a differentiated AI-driven design layer (services-skewed revenue) rather than a full-stack chemistry/biology CRO, the same "new AI-native CRO vs incumbent" split seen elsewhere in the AI-bio stack (Internal research notes).

  • Capital, not operations, is the edge (Oct 2026): compared with domestic lab-automation peers (megarobo, 奔曜/Benyao, 玄刃/Xuanren), XtalPi looks stronger mainly because it has a public-market funding channel; stripping out one-off licensing income, its operating loss is in the same range as MegaRobo's (Internal research notes).

  • Clients may become competitors: top AI-drug companies such as insilico-medicine build their own robotic labs, which caps how much "lab-as-a-product" revenue integrators can capture — XtalPi hedges by keeping the lab in-house and selling results.

5. Financials / Funding

  • Pre-IPO: eight private rounds totalling US$732M — reported as the largest pre-IPO total among AI drug-discovery companies; e.g. Series C US$288M (Sep 2020, led by SoftBank Vision Fund 2). Major pre-IPO holders included Tencent (13.7%), Sequoia China (8.3%) and 5Y Capital (7.9%) (Internal research notes).
  • Listed: HKEX (2228.HK), June 2024 under Chapter 18C (specialist technology) — the first AI drug-discovery company to list in Hong Kong. (Earlier versions of this page said Nov 2023; corrected.)
  • Post-IPO capital: roughly RMB 7.6B raised within about a year of listing through three placings plus a HK$2.87B convertible bond; cash about RMB 7B (Internal research notes).
  • FY2025: revenue RMB 803M (+201%), first reported profit; but ~45% of revenue was a one-off DoveTree upfront (US$51M, on an end-to-end collaboration with total potential value up to ~US$5.99B) and ~RMB 0.5B fair-value gains — ex-items, operating loss ≈ RMB 0.4B (Internal research notes). Robotics-solution revenue ≈ RMB 265M (+63%).
  • 2026H1 (reported 2026-07-24): revenue ≈ RMB 394M (−24% YoY on a high DoveTree base; ex-upfront growth ≈ +65–74%); AI4S intelligent-solutions revenue > RMB 180M (+120–136%), ≈49% of revenue; net loss to shareholders ≈ RMB 252M; R&D expense +50%+, directed at autonomous labs, agent systems and pipeline [gelonghui 2026-10-12].
  • Revenue model: drug-discovery services + licensing/milestones + lab-automation solutions; recurring profitability not yet demonstrated.

6. People & Relationships

  • Founders: three MIT physicists — 温书豪 (Shuhao Wen), 马健 (Jian Ma), 赖力鹏 (Lipeng Lai).
  • Partners: Pfizer (early); Eli Lilly — AI + robotics collaboration announced 2023-05-30 to de novo design candidates for an undisclosed target, worth up to US$250M in upfront and milestone payments (upfront/milestone split not disclosed) [BioSpace; Yicai Global, 2026-10-14]; DoveTree (2025, end-to-end collaboration) (Internal research notes).
  • Peers: insilico-medicine, isomorphic-labs, Schrödinger, dp-technology.
Data profile · AI for Science

晶泰控股 XtalPi

量子物理加 AI 加机器人实验的药物与材料研发平台,港股 2228

做什么:量子物理 + AI + 机器人实验;2026 中期材料强调"Physical AI / Agentic HTE / Agentic Synthesis"与"自进化闭环" [2]。港股代码 2228;成立年份 2015(二手 ⚠️)。

财务口径:2025 年收入 8.026 亿元(+201.2%),药物发现解决方案约 5.379 亿元(+418.9%),AI for Science 智能方案约 2.647 亿元(+62.6%);净利 1.346 亿元,含公允价值收益,2024 年净亏约 15.1 亿元;现金约 70.7 亿元 [1]。另有搜索摘要称净利 2.58 亿元、部分媒体称 1.24 亿元(归母口径),口径不同,引用须说明(二手);一家转述稿把收入误写为 80.3 亿元(单位错)。覆盖全球前 20 药企中 17 家(公司口径)。2026 年发行可转债约 25.4 亿元。

争议:盈利含大额一次性项目与首付款,不等于纯经营盈利;"市值约 358 亿港元"未回源,不采用。

  1. https://www.itiger.com/news/2622747110 (2026-10-15)
  2. https://ir.xtalpi.com/media/bunjanst/2026-中期业绩发布会演示材料-cn.pdf (2026-10-15)
Last compiled: 2026-10-14