XtalPi Holdings (02228.HK) is a founder-led AI-for-science platform built by three MIT-trained physicists. It is less a drug company than a toolmaker and co-development platform that combines quantum-physics computation, AI models, cloud computing and robotic labs to speed up drug and materials discovery, then sells that capability to pharma and biotech clients. The report's verdict is Rating: Hold, serious science with blue-chip partners, but a share price that already pays for cleaner, broader earnings than the platform has so far delivered.
The company is trying to climb from fee-for-service work toward milestone-and-royalty economics. In 2025 revenue jumped 201.2% to RMB802.6 million, and drug-discovery solutions alone rose to RMB537.9 million from RMB103.7 million, proof that platform monetization can step-change when a big contract hits its milestones. But the growth was narrow: one customer was more than 45% of revenue, the reported profit leaned on RMB514.0 million of fair-value gains rather than cash, and operating cash flow was still negative at RMB-165.4 million.
The bull case rests on a string of large collaborations. The DoveTree deal brought US$51 million received, a further US$49 million potential payment, and up to US$5.89 billion in milestones plus royalties; partnerships with Lilly (up to US$345 million) and Pfizer, and a June 2026 GPCR collaboration worth over US$400 million in potential value, add to the list. If these are not one-offs, XtalPi is crossing the bridge from project vendor toward scientific infrastructure that big pharma pays to access.
On valuation the report is cautious. At about HK$6.96 the stock has already fallen back from a HK$15.12 high, but it still trades around 31x sales and discounts a future where milestone wins repeat and broaden. The report's base-case fair value is HK$6.3 to 8.5, with an optimistic HK$11.6 to 13.0 and a conservative HK$3.8 to 4.1; the ideal buy zone is HK$3.8 to 4.1, well below today's price. You are paying for proof that has not yet arrived.
The main risks are heavy customer and deal concentration, profits flattered by fair-value gains, repeated share and convertible issuance that dilutes existing holders, and the wide gap between headline deal values and cash actually collected. The report's stance is a serious platform at a demanding price, better bought much lower. This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.
Prices in the article are as of publication; see the valuation band above for the live price.
Meta
- Ticker: 02228.HK
- Company: XtalPi Holdings Limited
- Price & market cap: HK$6.96 close as of 2026-06-23; market cap about HK$29.95 billion based on 4,303.37 million shares outstanding reported on 2026-06-24
- Currency: HKD
- Report date: 2026-06-24
- Industry: Healthcare Technology
- One-line positioning: AI-for-science platform selling drug-discovery and R&D automation solutions, with 2025 revenue of RMB802.6 million and growth led by milestone-heavy drug-discovery contracts.
Assumed research scope: primary listing basis is Hong Kong 02228.HK; quote date is the 2026-06-23 close; financial statements are reported in RMB and converted into HKD where useful using the CFETS HKD/CNY midpoint of 0.86408 on 2026-06-23, equal to about HK$1.1573 per RMB.
Research summary
XtalPi is not, in the strict economic sense, a drug company. It is a toolmaker, a model builder, an experimental-services provider, and increasingly a co-development platform trying to climb the value chain from fee-for-service work toward milestone-and-royalty economics. The distinction matters because the stock usually trades on “AI drug discovery” excitement, while the financial statements still read like a capital-intensive, R&D-heavy technology-enabled services business with episodic milestone revenue. In 2025, revenue rose to RMB802.6 million from RMB266.4 million. Two features of that jump matter more than the headline. Drug-discovery solutions jumped to RMB537.9 million from RMB103.7 million, which shows that platform monetization can step-change when a large contract hits recognized milestones. But one customer contributed RMB365.1 million, more than 45% of total revenue, so the reported growth was real but far from broadly distributed.
The market is trading three narratives at once. One is the Chapter 18C pioneer story: XtalPi was the first specialist-technology company to list under Hong Kong’s Chapter 18C route, and by early 2025 it had crossed the HK$250 million commercial-company revenue threshold ahead of schedule, which let its lock-up rules reset. The second is the “AI for Science” story, in which the company sells itself as a platform that combines quantum-physics-based computation, AI models, cloud computing, and robotic labs across drug discovery and materials science. The third is the deal-value story, where investors fixate on the headline values of collaborations with DoveTree, Lilly, Pfizer, and most recently an unnamed large biopharma GPCR program. The announcements all restate one promise: that the platform is good enough to move from outsourced work into high-value scientific infrastructure big pharma will pay to access.
The share price has moved in phases that match those narratives. The June 2024 IPO priced at HK$5.28 per share and raised roughly HK$989 million gross, making XtalPi the first Chapter 18C listing. Through late 2024 and early 2025, the stock benefited from the company clearing the Chapter 18C commercial threshold, then from two early-2025 placings that replenished cash and signaled capital-market appetite. The real re-rating came when the market saw a path from “interesting platform” to “very large contract economics”: the DoveTree term sheet was announced on 2025-06-23, the definitive agreement update followed on 2025-08-05, management then issued a positive profit alert for the first half and later for the full year, and Reuters data show the stock’s 52-week high reached HK$15.12 on 2025-10-06. By June 2026 the stock had fallen back to roughly HK$7, after the market absorbed repeated capital raises, a large January 2026 convertible-bond issuance, and the fact that the 2025 profit number was flattered by fair-value gains rather than cash generation.
That points to the central bull-bear disagreement. The bulls say 2025 was the first proof that XtalPi’s platform can convert scientific capability into large commercial contracts, and that the current number understates the long-tail value of milestones, royalties, biologics, and recurring enterprise-platform work. They point to the DoveTree deal structure, which included US$51 million received, a further US$49 million potential payment, and up to US$5.89 billion in regulatory and commercial milestones plus royalties; to Ailux’s Lilly collaboration with “tens of millions” in upfront and near-term milestones and up to US$345 million total potential value; to the Pfizer platform expansion; and to the June 2026 GPCR collaboration with over US$400 million total potential value. If those are not one-offs, XtalPi has begun crossing the bridge from project vendor to infrastructure provider.
The bears reply that recognized revenue is still far too dependent on lumpy milestone accounting, while cash conversion remains weak. The objection is not theoretical. Audited 2025 operating cash outflow was still RMB165.4 million, capital expenditure was RMB74.4 million, and lease payments were about RMB37.0 million. The return to full-year profit depended in part on RMB514.0 million of net fair-value gains from financial assets and on the disappearance of the large pre-IPO convertible-redeemable-preferred-share fair-value loss that distorted 2024. The company’s technology may be improving faster than its earnings quality. So the question that actually matters is not whether XtalPi turned profitable, but how much of the profit was operating, repeatable, and cash generative.
XtalPi now sits in an awkward but interesting middle ground. It is stronger than a pure concept stock because it has real revenue, blue-chip counterparties, a very large cash cushion, and some evidence that platform quality is good enough to win serious work. It is weaker than the most enthusiastic narrative implies because revenue concentration is high, recurring software-style revenue is still not the dominant economic engine, and the business keeps consuming operating cash. At 2025 year-end the company had RMB2.57 billion of cash and cash equivalents, RMB1.91 billion of bank deposits, RMB2.57 billion of current financial assets at fair value through profit or loss, and only a modest amount of borrowings relative to liquid resources. By 2025-06-30, management’s “cash balance” definition had reached RMB5.31 billion, which it said exceeded bank borrowings by RMB5.03 billion. For a pre-consensus platform business that is a genuine balance-sheet advantage, and it explains why capital markets have stayed open to XtalPi.
Horizontally, the closest public parallels are imperfect. Schrödinger is the cleanest software-and-physics analog and Recursion the closest “TechBio operating system” analog, while Certara is the nearest model-informed software-and-services platform in regulated drug development. WuXi AppTec is the best listed reference for what customers pay for when speed, reliability, and embeddedness in pharma workflows matter more than a pure software label. Set against those names, XtalPi reads as a hybrid: more commercial than many AI-drug peers, more technologically ambitious than most CRO-like peers, yet earlier and less proven in monetization than the established winners in either camp.
The right qualitative portrait label is company in transition. XtalPi is moving from proof-of-concept science platform to capital-markets-backed commercial platform, and from earning mostly service revenue to trying to capture option value from milestones, royalties, and platform licensing. Two things follow. Over the next 12 months, the stock will probably trade more on contract composition, cash conversion, and whether another large collaboration turns into recognized revenue than on any abstract debate about AI. Over three to five years, the question becomes whether XtalPi can make its economics look less like a laboratory-enabled services company and more like a scaled scientific infrastructure company with repeat business and durable pricing power. The stock is no longer a pure hope story. It is also not yet a high-quality compounder.
Company vertical history
XtalPi was founded in 2015 by Wen Shuhao, Ma Jian, and Lai Lipeng, three MIT-trained physicists who built the company around a specific scientific bet: that drug and materials discovery could be improved by a tighter combination of first-principles physics, machine learning, cloud computing, and automated experiments, not by statistical AI alone. The company’s own materials still describe that origin in those terms, and outside reporting from MIT and Reuters-backed early funding coverage showed that the market noticed the unusual founder profile early. In 2018, Sequoia China, Google, and Tencent joined a US$15 million round that brought total funding to US$20 million; by the IPO, legal advisers and market materials described a much broader roster of backers including HongShan, Mirae Asset, Tencent, Google, China Life Chengda, and 5Y Capital, with total private financing of roughly US$732 million by the end of 2023. That mix of science-first founders and unusually deep early capital shaped almost everything that followed.
The origin problem XtalPi tried to solve was real. Classical drug discovery is slow, expensive, and information-poor at the molecule-selection stage. AI-only approaches promised speed but often leaned on historical data that did not fully capture molecular physics. XtalPi’s answer was a closed loop: physics-based simulation proposes, AI prioritizes, robotic labs test, new data feed the models, and the cycle repeats. Early commercial traction came through solid-state and computational chemistry work, including collaborations with pharma. Over time the company broadened into automated chemistry, biologics support, and materials-science applications such as energy and advanced materials. That broadening worked as a hedge against the long monetization cycle of therapeutics. Materials programs have shorter paths to revenue; drug-discovery programs offer larger upside but slower and less predictable cash conversion.
The listing path matters because it explains the stock’s investor base and the way the company is judged. XtalPi listed on the HKEX Main Board on 2024-06-13 via Chapter 18C, the new specialist-technology route designed for companies that may be large enough and technologically significant enough to list before traditional profitability or revenue maturity. The IPO sold 187.373 million shares at HK$5.28, with an over-allotment of 8.796 million additional shares later exercised at the same price. Net proceeds from the global offering, including the over-allotment, were about HK$915.2 million. Market commentary around the transaction emphasized both the symbolism of the first Chapter 18C listing and the company’s blue-chip shareholder base.
XtalPi’s history is easiest to understand in four stages.
The first stage was platform formation. The goal was proof, not scale: that the company’s physics-and-AI stack could solve narrow but painful problems for pharmaceutical customers. The business model was service-led, science-heavy, and cash-consuming. What that period left behind was credibility with sophisticated customers and investors, without which later equity funding would have been much harder.
The second stage was platform broadening. By 2022 to 2024, XtalPi was no longer only a computational chemistry specialist. It had built intelligent robotics capabilities, started speaking more explicitly about “AI + robotics,” and pushed into materials science, energy, chemicals, and related fields. In 2024 annual results, the then-named intelligent robotics solutions business was already larger than drug-discovery solutions, at RMB162.8 million versus RMB103.7 million. That stage told the market something important: the company could monetize the automation side earlier than the big drug-discovery promise. The cost was that the story became harder to categorize. Investors who wanted “pure AI drug discovery” had to accept that a substantial part of reported revenue came from automation and R&D solutions.
The third stage was public-market validation and threshold crossing. After the June 2024 IPO, 2024 revenue reached RMB266.4 million, enough for the company to state that it had reached the HK$250 million commercial-company threshold under Chapter 18C earlier than expected. At the same time, management highlighted a declining monthly cash burn rate, from RMB62.2 million in 2023 to RMB48.1 million in 2024. The exchange’s commercial-company reclassification also mattered for insider lock-ups. In May 2025, the co-founders voluntarily extended their own lock-up by another year even though the updated mandatory expiry date had moved to 2025-06-12 after the Chapter 18C status change. That helped the narrative because it signaled founder confidence. It did not change the underlying economics, but it lowered one capital-markets overhang.
The fourth stage is the current one: monetization by large collaboration. The DoveTree term sheet in June 2025, the definitive agreement in August 2025, the Lilly biologics collaboration in November 2025, the Pfizer platform expansion in June 2025, and the June 2026 GPCR collaboration are more than new logos. They are management’s attempt to teach the market a new way to value the company, as scientific infrastructure that can earn upfront cash now and option-like economics later rather than as a vendor of narrow services. The 2025 interim results were the first clean proof that this can move the P&L quickly; revenue rose 403.8% year on year in the first half, and the company booked the US$51 million DoveTree first-phase payment as revenue. By the full year, drug-discovery revenue had overtaken the renamed AI for Science intelligent-solutions line by a wide margin. That changed the shape of the company without yet settling the question of repeatability.
Several key nodes still affect the company today. The first is the 2024 Chapter 18C IPO itself. It gave XtalPi a public-market currency and a larger investor audience, but it also exposed the business to a market that can overreact to narrative and underreact to revenue quality. The second is the pair of January and February 2025 placings, which together raised about HK$3.21 billion net. Those transactions materially extended runway and made later acquisitions possible, but they also taught investors that management will issue equity when market demand exists. The third is the August 2025 placing of up to 285.92 million shares at HK$9.28, which raised another estimated HK$2.63 billion net. The fourth is the January 2026 HK$2.866 billion zero-coupon convertible bond due 2027, convertible at HK$13.85. Taken together, those capital actions say two things at once: XtalPi has exceptional financing access for a still-early platform company, and existing shareholders face a real dilution regime whenever market conditions are favorable.
A fifth node is the strengthening of the finance function. In March 2026, Zhou Feiran, previously a J.P. Morgan healthcare investment banker, was appointed CFO. That looks like the right hire for the stage XtalPi has reached, because the company no longer only needs scientific storytellers; it needs a finance team that can translate business quality into cleaner capital allocation, cleaner disclosure, and eventually cleaner valuation support.
The history reads cleanly. XtalPi is a science platform slowly learning to speak the language of capital markets. It has already proven it can raise money, attract elite partners, and generate very rapid bursts of reported revenue. What it has not yet fully proven is that those bursts add up to a smooth, cash-generative business.
Financial vertical review
The published financial history since listing is short, but long enough to show the skeleton of the business. Revenue rose from RMB174.4 million in 2023 to RMB266.4 million in 2024 and then to RMB802.6 million in 2025. The underlying mix changed sharply. In 2024, intelligent robotics solutions were the larger business at RMB162.8 million. In 2025, drug-discovery solutions became the lead engine at RMB537.9 million, while AI for Science intelligent solutions produced RMB264.7 million. Geography shifted too: 2025 revenue from mainland China was RMB226.1 million, from the United States RMB458.3 million, and from other regions RMB118.2 million. The company did not just get bigger; its commercial center of gravity moved toward higher-value biopharma work in the U.S. market.
The gross-margin story looks excellent on the surface and requires caution underneath. Using the cost-of-revenue figures disclosed in the 2025 results announcement, gross margin rose from roughly 46% in 2024 to about 70% in 2025. In the first half of 2025, the implied gross margin was even higher because the first US$51 million DoveTree payment was recognized as revenue while the associated delivery cost was not proportionate to that payment’s size. That is exactly why a platform business can look suddenly software-like before it is actually software-like. High milestone revenue is attractive. It is also inherently lumpy. Investors should treat 2025 gross margin as proof of earnings power, not as proof of a stable new baseline.
Earnings quality is the key fault line. The group reported full-year profit in 2025 for the first time, and management’s March 2026 positive-profit alert had already flagged a turnaround to at least RMB100 million of profit attributable to shareholders. But the audited drivers show why that number cannot be read as pure operating strength. Other net gains jumped to RMB514.0 million, mainly because of fair-value gains on financial assets at FVTPL. In addition, the 2024 fair-value loss on convertible redeemable preferred shares disappeared after listing because the instruments converted into ordinary shares. Those are real accounting facts. They are not recurring operating cash flows. That is why operating profit of RMB55.2 million is more informative than net profit alone, and why operating cash flow matters even more.
Cash conversion improved, but it is still weak. Net cash used in operating activities was RMB478.7 million in 2024 and RMB165.4 million in 2025. That is a major improvement, yet it remains negative. Capital expenditure rose from RMB46.7 million in 2024 to RMB74.4 million in 2025, and lease payments were substantial in both periods. On a rough owner-earnings basis using operating cash flow less capital expenditure and lease payments, the business was still materially negative in 2025. That means the operating leverage story is real, but the self-funding story is not here yet.
The balance sheet is strong by any reasonable near-term standard. At 2025 year-end, XtalPi held RMB2.57 billion of cash and cash equivalents, RMB1.91 billion of bank deposits, RMB2.57 billion of current financial assets at FVTPL, and RMB18.0 million of restricted cash. Total liabilities were only RMB786.6 million. By 2025-06-30, management said “cash balance” had reached RMB5.31 billion and exceeded bank borrowings by RMB5.03 billion. In practical terms, that gives XtalPi time. It can absorb uneven contract timing, keep investing in models and labs, and still remain a going concern without needing immediate rescue financing. For an early-stage platform business, time is the rarest asset. XtalPi has bought a lot of it.
There are two balance-sheet caveats. One is receivables concentration. Trade receivables and notes rose to about RMB151.1 million at year-end 2025, up from RMB98.7 million a year earlier, which is not alarming on its own but is worth watching given milestone concentration. The second is that a large share of liquid resources sits in FVTPL financial assets and treasury products rather than idle cash. Management explicitly said these investments were made for treasury-management purposes, and in 2024 highlighted specific note investments and related unrealized gains. That is not the same as a distressed company reaching for yield, but it does make the P&L noisier and blurs the clean line between operating performance and treasury results.
Returns on capital are not yet the right primary lens. The company’s financial history is still dominated by platform build-out, listing-related accounting changes, and milestone timing. The better question is whether incremental revenue is becoming less expensive to generate. On that score, the evidence is encouraging. Management’s own cash-burn metric fell in 2024 and again in 1H25, to RMB49.7 million per month in the first half of 2025 from RMB62.1 million a year earlier, even as revenue accelerated sharply. That says the model has operating leverage. It does not yet say the model has matured.
Price and valuation history
Because XtalPi listed only in June 2024, its price history is short but unusually eventful. The first phase ran from the IPO at HK$5.28 through the early post-listing window, when the market treated the stock as a narrative-heavy Chapter 18C experiment. The second phase was the commercial-threshold re-rating: as 2024 revenue reached the Chapter 18C commercial-company threshold and management showed lower cash-burn intensity, the market became more willing to pay for scale potential rather than just scientific novelty. The third phase was the collaboration boom, when the DoveTree announcement, positive profit alerts, and the broader AI-healthcare trade pushed the stock to its 52-week high of HK$15.12. The fourth phase is the normalization phase, with dilution absorption, more sober reading of 2025 earnings quality, and a June 2026 price near HK$7.
The valuation center moved for reasons that were partly deserved and partly thematic. Deserved: 2025 brought proof that XtalPi could sign and recognize large-value collaborations, and that the drug-discovery segment could become the growth engine. Thematic: the broader pharma sector’s growing willingness to spend on AI made every credible platform name more interesting. Reuters summarized that industry shift in May 2026, noting that major pharmaceutical companies were increasingly turning to AI tools and automated labs in the hope of compressing timelines and cost. That backdrop amplified XtalPi’s re-rating.
At the current price area, the stock is far below its peak but still not plainly cheap against what is proven in cash terms. Reuters metrics around 2026-06-24 showed price-to-sales a little above 31x, while the 2026-06-24 Reuters quote page showed a similar valuation stack and a 52-week range of HK$5.27 to HK$15.12. That means the market has de-rated the name from exuberant levels, but it still prices in a large amount of future execution. The stock is not trading like a distressed science experiment. It is trading like a company the market still expects to become strategically important.
Business model and moat
XtalPi’s revenue structure now splits into two lines. Drug-discovery solutions generated RMB537.9 million in 2025 and AI for Science intelligent solutions generated RMB264.7 million. In 2024 those numbers were reversed in hierarchy: intelligent robotics solutions, as then named, contributed RMB162.8 million while drug discovery contributed RMB103.7 million. That swing matters because it changes what kind of business XtalPi is becoming. Automation and R&D solutions still matter, but the value story now rides on drug discovery. Within that, revenue can come from standalone services, platform-enabled projects, and milestone-based collaboration economics. The 2025 annual report and 1H25 interim results make clear that management is pushing the business toward a mix with more milestone value and higher customer lifetime value, especially in antibodies and platform co-development.
The cost structure combines heavy fixed investment with meaningful variable delivery costs. R&D expense was RMB418.2 million in 2024 and RMB569.2 million in 2025. Employee benefits were RMB602.4 million in 2024 and RMB735.4 million in 2025. These are the economics of a science platform, not a lightly staffed software company: it must keep paying for researchers, model development, automation, and facilities before it can recognize the most attractive revenue. So operating leverage can be very strong on the way up and painful on the way down. When a large milestone lands, margins jump. When revenue timing slips, staff and R&D do not shrink quickly enough to protect earnings.
The first real moat is technical integration. XtalPi’s pitch rests on connecting first-principles simulation, AI, high-performance computing, and automated experimental systems in a usable loop, not on owning one good model. Customers buy time compression and tighter iteration, not just prediction. The DoveTree, Pfizer, Lilly, and GPCR announcements all point to that same integrated offer. It is a real moat because it is hard to replicate quickly; few companies have serious dry-lab physics, AI infrastructure, and wet-lab automation under one roof. It is not unbreakable, because customers can still multi-source pieces of that stack from software vendors, CROs, and internal teams.
The second real moat is reference-driven customer trust. By late 2025 the company said it had cumulatively covered 17 of the world’s top 20 pharmaceutical companies, and earlier materials around the IPO said it had already served more than 300 biotech and pharma companies and research institutes. In scientific outsourcing, reference quality matters more than a consumer brand. Once a company becomes trusted for hard problems inside multiple top pharma organizations, conversations begin at a higher level. That does not eliminate competition, but it improves sales efficiency and partner willingness to test broader platform relationships.
The third moat is capital access. XtalPi has repeatedly raised money at scale: IPO proceeds, two early-2025 placings, an August 2025 placing, and a January 2026 convertible. For a business that still runs negative operating cash flow, access to capital is part of the moat. It lets the company keep building while weaker peers are forced to cut. The catch is obvious. Capital access helps the company more than it helps per-share economics unless the new capital produces returns above dilution cost.
Some alleged moats are weaker than they sound. “AI for Science” is not a moat by itself; it is a marketing category. XtalPi’s data asset is valuable, but the market for models and scientific AI is moving fast, and customers can increasingly license or build specific capabilities elsewhere. Switching costs also look lower than the bull case sometimes implies. The fact that one customer accounted for over 45% of 2025 revenue is evidence of value, but it is also evidence that the business is still won contract by contract rather than locked in through broad, sticky enterprise subscriptions.
Management and governance are better than average for an early platform name, but not yet premium. The company remains founder-led, with Wen Shuhao, Ma Jian, and Lai Lipeng deeply involved and together personally locked up on 591.5 million shares, or 14.71% of the company, through 2026-06-12 under their voluntary extension. The board includes three independent non-executive directors. There is no obvious dual-class structure in the public documents reviewed, which reduces one common governance discount. The March 2026 CFO appointment was sensible. I do not see a major governance red flag in the reviewed disclosures. The real governance question is capital allocation discipline: how much of the large cash pile ends up funding durable platform advantage, and how much merely funds breadth for breadth’s sake.
Industry and cycle
XtalPi sits across three adjacent markets: AI-enabled drug discovery, computational chemistry and biosimulation, and laboratory automation / tech-enabled R&D execution. That is why no single industry frame is fully satisfactory. The pharmaceutical buyer increasingly wants all three. Reuters reported in May 2026 that major pharma companies were leaning further into AI to accelerate R&D, with the promise of lower cost and shorter timelines. That trend helps the whole category. It also raises the bar. As more buyers take AI seriously, they demand results that connect to actual development decisions. XtalPi benefits if the market rewards integrated platforms. It suffers if AI budgets fragment into many point solutions.
The industry’s profit pool is uneven. Pure software vendors such as Schrödinger and Certara capture high gross margins and cleaner recurring economics when they become embedded in customer workflows. Service-heavy or development-heavy models can grow faster in bursts but generally have noisier margins. Tech-enabled CRO and CRDMO models like WuXi AppTec capture value through scale, reliability, and regulatory execution rather than through software-style price multiples. XtalPi is trying to occupy a narrow band between those profit pools. It wants software-like valuation and collaboration upside without giving up the experimental infrastructure that makes its science credible. That is strategically ambitious, but hard to execute.
The company is exposed to several kinds of cycle at once. There is a technology-iteration cycle, because model quality can become obsolete quickly, and a capital-markets cycle, because early platform companies rely on favorable financing windows. A pharma R&D spending cycle also bears on it, though that runs milder than a classic semiconductor or commodity cycle. XtalPi is less exposed to a pure macroeconomic cycle than a cyclical manufacturer, but it is not defensive. If interest rates stay high and the market stops paying premium multiples for narrative-heavy healthcare tech, the share price can compress even while the business keeps growing. The 2025-26 path already showed that.
Policy and regulation matter in two specific ways. Chapter 18C gave the company access to the public market before it looked like a mature commercial company, which clearly helped. More broadly, drug discovery and biopharma remain highly regulated end markets. XtalPi is not directly running pivotal clinical trials on its own balance sheet, so it avoids much of the binary clinical risk that pipeline biotechs face. But its customers’ enthusiasm can still be constrained by data-security concerns, cross-border scientific collaboration rules, and the growing scrutiny over how AI tools are validated in high-stakes settings. None of those risks looks like an immediate existential constraint in the reviewed materials. They do argue against giving the company a frictionless long-term multiple.
Horizontal competitor analysis
There is no perfect direct comparable for XtalPi. The right frame is a few useful public peers, none identical. Schrödinger, Recursion, and Certara are the three best public “AI / modeling / software-for-drug-development” references; WuXi AppTec is the best public operational reference for how pharma pays for speed, quality, and embedded scientific execution at scale. Investors often lump XtalPi with AI-drug-discovery names, but the intelligent automation business shifts the useful question away from which peer looks identical and toward which public models illuminate the monetization choices XtalPi still has to make.
Schrödinger has become what many investors once wanted XtalPi to become: a company whose software and physics engine are themselves commercially valuable, with 2025 total revenue of US$256 million, including US$200 million of software revenue. The market still treats Schrödinger as risky, but the business mix is clearer. Customers buy computational capability directly, and the company can tell a more recognizable software story. XtalPi’s advantage against Schrödinger is broader wet-lab and automation integration. Schrödinger’s advantage is cleaner revenue identity. Customers who want a modeling layer may pick Schrödinger. Customers who want modeling tied tightly to automated execution may find XtalPi more compelling.
Recursion is the closest narrative rival. It also sells an AI-native operating-system idea for biology, but its economic profile remains much more biotech-like. In 2025, Recursion generated only US$74.7 million of revenue, used US$371.8 million of operating cash, and posted a US$644.8 million net loss, even though it ended the year with US$753.9 million of cash and had received more than US$500 million of milestone payments from partners to date. Compared with Recursion, XtalPi looks more commercial today and less pipeline-binary. Compared with XtalPi, Recursion arguably has deeper ownership of therapeutic upside through its own programs. Customers do not pick between them in a simple beauty contest. They pick based on whether they want a partner, a platform, or a pipeline owner.
Certara is a different kind of comparator but still useful. It is a provider of biosimulation technology and services used throughout drug development, and by 2026 it still had a sub-US$1 billion market cap despite a much cleaner, more established business model than XtalPi. That contrast matters. Markets will pay a premium for XtalPi only if they believe its scientific platform can outrun Certara-like software economics and avoid Recursion-like cash burn. If XtalPi remains primarily a work-for-hire plus occasional milestone company, it should not structurally deserve a far richer multiple than the most embedded scientific-software platforms.
WuXi AppTec is not an AI-for-science peer in the narrow sense, but it is a very important commercial peer. In 2025, WuXi AppTec delivered RMB45.46 billion of revenue and RMB58.0 billion of backlog in continuing operations, which shows what industrialized scientific infrastructure looks like at scale. The market values WuXi for execution and embeddedness, not for being fashionable. That is the benchmark XtalPi eventually has to satisfy to move from premium-narrative asset to premium-quality asset. Today it is far earlier, far smaller, and far more dependent on the market’s willingness to price optionality.
The niche XtalPi occupies is that of a challenger platform company. It leads no single public-market-defined segment, but it is the rare listed company trying to integrate molecule design, simulation, robotics, and cross-domain materials work into a commercial scientific platform, filling a gap between pure software and pure CRO execution. The danger is that this niche pays off only if XtalPi turns technical integration into commercial standardization. Otherwise it stays impressive and hard to categorize, which is great for conference stages and less great for per-share compounding.
Peer metrics snapshot
| Dimension | XtalPi | Schrödinger | Recursion | WuXi AppTec |
|---|---|---|---|---|
| Latest reported full-year revenue growth | +201.2% in 2025 | 2025 revenue US$256m | 2025 revenue US$74.7m | 2025 revenue +15.8% |
| Latest reported operating cash flow | RMB-165.4m in 2025 | strong balance sheet, path to positive adj. EBITDA by 2028 | US$-371.8m in 2025 | record cash-flow growth in 2025 |
| Balance-sheet signal | 1H25 cash balance RMB5.31bn | strong balance sheet | cash US$753.9m, runway into early 2028 | backlog RMB58.0bn |
| Current valuation signal | P/S around 31x | market cap about US$1.14bn | market cap about US$1.68bn | market cap about HK$365bn |
| What customers buy | integrated AI + physics + robotics | modeling software and discovery platform | TechBio platform plus pipeline | scaled CRDMO execution |
The business reason behind the differences is more important than the numbers. Schrödinger and Certara are closer to software economics. Recursion is closer to platform-plus-pipeline economics. WuXi is large-scale scientific manufacturing and development infrastructure. XtalPi remains between those models. That hybrid position is the source of its upside and the reason its multiple is hard to defend cleanly.
Current fundamentals and bull-bear divergence
The latest public fundamentals still revolve around 2025 and the post-year-end announcements. Reported 2025 revenue was RMB802.6 million, up 201.2%, with drug-discovery solutions growing 418.9% and AI for Science intelligent solutions growing 62.6%. Operating profit turned positive at RMB55.2 million, but net fair-value gains of RMB514.0 million were a large contributor to bottom-line optics. The more cash-relevant signal is that operating cash outflow improved to RMB165.4 million from RMB478.7 million in 2024. The business has improved dramatically; the accounting quality of the improvement is still mixed.
The last four quarters, to the extent public reporting allows reconstruction, tell a specific story. The first half of 2025 was explosive because the company recognized the US$51 million initial DoveTree payment, driving 403.8% revenue growth and the first positive half-year adjusted profit. The second half of 2025, inferred from the difference between full-year and interim figures, was much slower sequentially, which is exactly what one should expect once a major upfront milestone has already been booked. The market should not read that slowdown as operational collapse. It should read it as evidence that milestone revenue can pull growth forward.
What the market is trading right now is not the 2025 audited number by itself. It is trading the probability that 2025 was the start of a repeatable pattern. The June 2026 GPCR deal with over US$400 million total potential value, the June 2026 US$100 million buyback authorization, and the continued public emphasis on Pfizer, Lilly, and other platform relationships all aim at the same conclusion: XtalPi wants the market to believe the company has crossed from scientific possibility to commercial inevitability. I think that narrative is ahead of the evidence, but not detached from it.
The bull case rests on four hard facts. First, the company has now shown that global customers will sign contracts large enough to change the income statement. Second, 2025 drug-discovery revenue and the 1H25 results suggest that the most valuable business line is no longer merely aspirational. Third, the cash position is strong enough that management does not need to shrink the ambition to survive. Fourth, the customer roster and collaboration flow suggest that platform quality is being validated externally, not only by company marketing.
The bear case also rests on four hard facts. First, 2025 revenue concentration was extreme, with one customer above 45% of sales. Second, 2025 earnings quality was flattered by fair-value gains and cleaner post-IPO accounting rather than only by core operations. Third, operating cash flow is still negative, meaning recognized revenue has not yet become owner earnings. Fourth, the capital-markets history since listing shows repeated willingness to issue equity and convertibles, which is rational corporate behavior but dilutive shareholder behavior if returns on new capital disappoint.
Valuation analysis
Historical valuation is difficult because the listed history is short and the fundamentals have changed quickly. What can be said with confidence is that the current price is far below the 2025 euphoria peak yet still trades at a premium that assumes substantial future success. Reuters showed price-to-sales slightly above 31x around 2026-06-24, and the stock had peaked at HK$15.12 within the prior year. That combination means the market has already partially deflated the narrative premium without abandoning it.
Peer valuation does not offer a free lunch. XtalPi’s premium versus more mature scientific-software names would be easier to defend if its revenue were more recurring and its cash conversion cleaner. Its premium versus more biotech-like AI platform peers can be defended because XtalPi is already more commercial. Against WuXi AppTec, the premium is plainly about optionality rather than current quality. The conclusion from peers is not that XtalPi is obviously overvalued because other names are cheaper; it is that XtalPi still needs unusually good execution to justify where it trades.
Cash-flow passthrough is the discipline that matters most here. Over the available public operating history in 2024-2025, operating cash flow was negative in both years despite large improvement in 2025. In 2025, the ratio of operating cash flow to net income was negative because operating cash flow was RMB-165.4 million while net profit was positive. Maintenance-versus-growth capex is not broken out, but management says capital expenditure is used to expand operations and upgrade facilities, which implies at least some of the 2025 RMB74.4 million capex was growth-oriented. Even granting that, owner earnings remained negative after capex and lease payments. The gap from headline profit is well above 30%. On valuation discipline, that forces the analysis away from P/E and toward sales multiples with an explicit adjustment for net cash.
Valuation scenario framework
This is valuation-scenario analysis within a research framework, not investment advice.
| Dimension | Conservative | Base | Optimistic |
|---|---|---|---|
| Revenue / margin assumptions | 2026 revenue RMB1.0bn; gross margin falls back toward low-60s as milestone mix normalizes | 2026 revenue RMB1.25bn; gross margin mid-60s as drug-discovery keeps scaling | 2026 revenue RMB1.55bn; another large platform deal plus stronger biologics and automation growth |
| Cash-flow assumptions | OCF still negative but improving; cash shields downside | OCF approaches breakeven in 2027 | OCF turns positive sooner as milestone revenue broadens into repeat business |
| Multiple assumptions | 14x EV/Sales on 2026 revenue | 18x EV/Sales on 2026 revenue | 22x EV/Sales on 2026 revenue |
| Key catalysts | No second mega-deal; steady execution | Better mix quality, reduced concentration, more recognized platform revenue | Another DoveTree-sized proof point, better cash conversion, deeper biologics licensing |
| Key risks | Revenue lumpiness persists | Mix normalizes slower than hoped | Narrative outruns delivery again |
| Implied upside | value about HK$5.1 per share; downside from current about 26% | value about HK$7.4 per share; upside from current about 6% | value about HK$10.5 per share; upside from current about 51% |
| Permanent-loss risk | trigger: collaboration revenue proves one-off and multiple compresses into low-teens sales | trigger: cash conversion stalls and market no longer pays premium for optionality | trigger: customer concentration worsens and dilution resumes before cash self-funding appears |
These values use the 2025-06-30 net-cash framework disclosed by management, with “cash balance” exceeding bank borrowings by RMB5.03 billion, and build from audited 2025 revenue plus the collaboration trajectory visible in 2025-26 announcements. The math is deliberately simple because precision would be false precision here; the key variable is not discount-rate fine tuning but whether the company can turn milestone-heavy growth into broader, repeatable platform revenue.
Expectation-gap analysis points to three variables. The first is the gap between announced deal value and near-term recognized revenue. The second is the gap between accounting profit and operating cash generation. The third is the gap between customer count / customer quality and concentration in any single reported period. The next results event that genuinely matters will be evidence that recognized revenue is becoming broader and cleaner, not another press release about a theoretical total deal value.
Margin of safety, checked independently, is weak. The current price is above the value implied by the conservative scenario and therefore offers no margin of safety against a slower monetization path. The most fragile base-case assumption is revenue breadth. If I cut the base revenue assumption to 70% of the modeled level and keep the multiple unchanged, the base-case value falls into the mid-HK$5 range, below the current price. If earnings were flat for three years in accounting terms, the owner-earnings yield would still be negative because owner earnings are currently negative. This is a good-company-bad-price type of setup more than a broken-company-cheap-price setup. Margin-of-safety sufficiency verdict: none.
Risk analysis
The largest business risk is milestone concentration. Probability: medium to high. Impact: high. Observable indicator: the share of revenue from the largest customer and the split between upfront / milestone / recurring service revenue. The transmission path is direct. If one or two major programs slip, reported revenue can disappoint sharply, gross margin can compress, and the market can decide the platform is less repeatable than previously thought. The 2025 customer-concentration disclosure is strong evidence that this is not a theoretical risk.
The most important financial risk is that the business still does not convert accounting wins into owner earnings. Probability: high. Impact: high. Observable indicator: operating cash flow, capex, lease payments, and treasury-related fair-value gains. The transmission path runs through valuation multiples. As long as investors believe the cash gap is temporary, they can look through it. If the gap persists after another year of headline collaborations, the market will lower the multiple first and ask questions about “AI for Science” later.
The third risk is dilution by success. Probability: medium. Impact: medium to high. Observable indicator: new-share placements, convertibles, and share-based awards. The company has already shown that it will raise money whenever demand exists. That is rational from management’s perspective because it extends runway and funds acquisitions. For minority holders, the risk is that the business grows but the per-share claim on that growth grows more slowly.
The fourth risk is competition and partial commoditization. Probability: medium. Impact: medium. Observable indicator: deal size, deal mix, and whether partners are paying for broad platform license / co-development or only discrete project work. Reuters’ industry survey in 2026 makes clear that pharma is adopting AI more broadly. That is good for demand. It also means more competitors will claim some version of the same workflow improvement. If XtalPi cannot keep technical differentiation visible in customer economics, the market could start valuing it more like an outsourced-science supplier than like proprietary infrastructure.
The fifth risk is capital-allocation drift. Probability: medium. Impact: medium. Observable indicator: acquisitions, treasury-product subscriptions, and the ratio of investment gains to operating profit. The Shanghai Siwei acquisition for RMB250 million may prove useful, but it adds integration work and goodwill/intangible complexity at a moment when investors want cleaner evidence of organic operating leverage. When a young company with a premium multiple begins to look more like a holding platform with treasury gains and bolt-on deals, valuation discipline usually tightens.
Catalysts and tracking indicators
Positive catalysts are easy to identify. Another large collaboration with meaningful near-term cash, especially if it comes with recognized revenue rather than only large theoretical milestones, would help. So would a full-year period in which operating cash outflow narrows substantially again or turns positive. The June 2026 US$100 million buyback plan is also a potential support signal, though its real importance depends on execution rather than authorization.
Negative catalysts are equally clear. A results period showing revenue growth but weak cash conversion would damage the current narrative. So would evidence that 2025’s largest contract was less repeatable than investors assumed, a renewed wave of dilution, or any quarter in which management leans too heavily on total potential deal values without matching progress in recognized revenue.
Tracking dashboard
| Indicator | Normal range | Alert threshold |
|---|---|---|
| Revenue growth | above 25% YoY | below 10% YoY |
| Drug-discovery revenue share | above 55% | below 50% |
| Largest-customer revenue share | below 30% | above 40% |
| Gross margin | 60%–70% | below 55% |
| Operating cash flow / revenue | better than -10% | worse than -20% |
| Net cash balance over annualized R&D | above 4x | below 2x |
| Annual share-count growth | below 5% | above 10% |
| Fair-value gains as share of pre-tax profit | below 20% | above 40% |
Each indicator earns its place. Revenue growth shows whether the platform is still winning work, and drug-discovery share shows whether the higher-value business is truly taking over. Largest-customer share is the read on whether the revenue base is broadening, while gross margin signals how much of the mix is milestone-heavy. Cash-flow intensity is the test of whether the P&L is becoming real, and net cash versus R&D measures how much time the company still has. Share-count growth reveals what management thinks the dilution cost of capital is. Fair-value gains versus pre-tax profit separate the part of the earnings story that belongs to operations from the part that is treasury noise. The source documents to monitor are annual and interim reports, HKEX business-update announcements, monthly returns, and any disclosure around placements, buybacks, or convertibles.
Cross-synthesis summary
Vertically, what XtalPi has genuinely proven is not that AI can “solve” drug discovery. That is far too grand and far too vague. It has proven something narrower and more valuable: a founder-led science platform built around physics, AI, and automation can win meaningful commercial work from sophisticated customers, can compress certain parts of the R&D loop enough to get paid for it, and can persuade capital markets to keep funding the build-out while the business model is still maturing. Those are real capabilities. They are not trivial. They also do not yet amount to a durable proof that XtalPi has become a cash-compounding software or infrastructure franchise.
Past success came from a mix of era tailwinds and authentic company-specific strength. The era tailwind was obvious: the market became far more willing between 2024 and 2026 to believe that AI could matter in pharma R&D. The company-specific strength was the integrated nature of XtalPi’s platform and the founders’ willingness to build actual experimental capability rather than sell only model demos. That combination gave the company a more credible commercial story than many AI-for-science aspirants. The problem is that the same tailwind that helped XtalPi also helped competitors and inflated valuation standards. What matters now is whether company-specific strength can carry the stock after the easy thematic multiple has already been spent.
Horizontally, XtalPi’s real edge versus peers is the attempt to connect both sides rather than to be the single best model builder or the best lab operator. Against software peers, it can say the science gets tested in-house. Against execution peers, it can say the work is intelligence wrapped around automation, not just labor. That is a meaningful strategic position. Its weakness is that the market has not yet seen enough proof that this hybrid position produces hybrid economics in the good sense. Right now the economics still look like a volatile compromise: better than a pure service business on margin opportunity, worse than a true software platform on repeatability, better than a clinical-stage biotech on current commerciality, worse than an established scientific-infrastructure leader on cash discipline.
The current valuation is rewarding future success more than past success. Past success deserves some premium. The company crossed the Chapter 18C commercial threshold early, grew fast, and won impressive partners. The current multiple still assumes much more: that large collaborations will continue, that more of them will convert into recognized revenue, that customer concentration will fall rather than rise, and that negative owner earnings are temporary. The market’s most likely misjudgment is not about whether XtalPi has strong science. The evidence increasingly says it does. The likely misjudgment is about how fast strong science becomes broad, recurring, high-quality earnings.
Over the next year, the decisive variables are revenue composition, cash conversion, and contract breadth. Over the next three years, the decisive variable is whether XtalPi becomes a trusted platform embedded in customer workflows rather than a sequence of large announcements. Over five years, the decisive variable is whether the company can graduate from equity-funded transition to self-financed compounding. The company becomes a much better investment if two things happen together: recognized revenue broadens beyond one or two major contracts, and operating cash flow approaches breakeven without another heavy round of dilution. The original judgment should be re-examined if the opposite happens: if another year of marquee partnerships still leaves cash conversion weak and the share count meaningfully higher.
Bull and bear reasons
Bull reasons:
- The 2025 results proved that XtalPi’s drug-discovery platform can generate revenue on a scale that changes the P&L, with drug-discovery revenue rising to RMB537.9 million from RMB103.7 million in one year.
- The collaboration pipeline is not hypothetical: DoveTree, Lilly, Pfizer, and a new June 2026 GPCR collaboration all show that top-tier counterparties are willing to pay for access to the platform.
- The balance sheet is unusually strong for an early platform company, with a 1H25 cash-balance definition of RMB5.31 billion and net cash still well above borrowings.
- Founder alignment is meaningful, with the three co-founders voluntarily extending lock-ups on 14.71% of the company through June 2026.
Bear reasons:
- One customer contributed RMB365.1 million in 2025, which means the year that changed the story was also a year of severe customer concentration.
- 2025 profit quality was mixed because RMB514.0 million of fair-value gains and post-IPO accounting effects materially flattered the bottom line.
- Operating cash flow remained negative in 2025, and owner earnings remained materially below reported profit after capex and lease payments.
- Since listing, management has repeatedly raised capital through placings and convertibles, which strengthens the company but weakens the margin of safety for existing shareholders.
Pre-mortem
The most plausible 50% drawdown script is commercial normalization plus multiple compression, not scientific failure. Imagine that in 2027 no second DoveTree-sized collaboration produces similar recognized revenue, the largest customer still contributes over 35% of sales, revenue growth drops below 20%, and operating cash flow remains negative. In that case the market could stop treating XtalPi as premium infrastructure and instead value it closer to a volatile scientific-services company. A move from a low-30s sales multiple toward the low teens on slower growth could cut the share price roughly in half even without a balance-sheet crisis.
A second script is dilution layered onto disappointment. Suppose management continues investing aggressively, recognized revenue stays lumpy, and another funding event arrives before self-funding is visible. If the share count rises again just as the market becomes skeptical of AI-drug-discovery optionality, the stock could fall sharply even while the company remains solvent and technologically relevant. That is the classic risk in narrative-rich growth stories with open capital-market access.
Final research conclusion
XtalPi is a serious company with serious science, and the market is right to distinguish it from pure concept stocks. The platform has commercial traction, the partner list is credible, the balance sheet is strong, and 2025 showed that the company can produce numbers large enough to matter. But the current ownership case still asks investors to pay for the company XtalPi may become rather than the cash machine it already is. The reason for restraint is not skepticism about AI; it is skepticism about revenue quality and per-share economics.
At the current price area, the stock is ownable only for investors who already accept platform volatility and can tolerate a long wait for cleaner cash conversion. What worries me most is the gap between headline collaboration value and owner earnings. Two things would change my mind: a materially lower price, or another year showing that recognized revenue is broadening, the largest-customer share is falling, and operating cash outflow is shrinking enough to make 2025 look like the beginning of a durable pattern rather than a milestone spike.
【Company-profile scores】
- Fundamental quality: medium
- Growth: high
- Moat: medium
- Financial soundness: strong
- Management credibility: medium
- Valuation attractiveness: low
- Risk level: high
- Suitable investor type: long-term growth
【Investment rating】
- Rating: Hold
- One-line thesis: Strong platform validation is now real, but the current price still discounts cleaner and broader earnings than the business has yet delivered.
- Three price signals:
- 【Ideal Buy Price】3.8–4.1 HKD Basis: about 20% below the conservative scenario value of roughly HK$5.1 per share.
- Acceptable hold price: 6.3–8.5 HKD
- Clearly overvalued price: above 11.6 HKD
- Current-price classification: acceptable hold
- Whether to wait for a better price: yes. A buy case becomes materially better below about HK$4.1, or at a higher price only if the next major results period shows broader revenue and much better cash conversion. The opportunity cost of waiting is missing another narrative-led re-rating.
- Target holding horizon: 3–5 years
- Expected annualized return: conservative about -26%; base about +6%; optimistic about +51%
- Max-loss risk: roughly 50% if milestone-heavy growth fades, cash conversion stays weak, and the multiple compresses toward the low-teens sales range
- Reassessment-trigger signals:
- if the largest customer remains above 40% of annual revenue for another full year
- if operating cash outflow worsens again after 2025’s improvement
- if gross margin falls below 55% for a full reporting period without evidence of healthier recurring revenue
- if annual share-count growth rises above 10% again through new financings or heavy equity issuance
- if large collaboration announcements stop translating into recognized revenue within 12–18 months
【Valuation Range】
- current: 6.96 (close as of 2026-06-23)
- bear (conservative · ideal buy zone): [3.8, 4.1]
- base (fair · acceptable hold zone): [6.3, 8.5]
- bull (optimistic · above the clearly-overvalued line): [11.6, 13.0]
Key data tables
Revenue and cash profile
| Dimension | 2024 | 2025 | 1H25 |
|---|---|---|---|
| Revenue | RMB266.4m | RMB802.6m | RMB517.1m |
| Drug-discovery solutions revenue | RMB103.7m | RMB537.9m | RMB435.2m |
| Intelligent solutions revenue | RMB162.8m | RMB264.7m | RMB81.9m |
| Net cash used in operating activities | RMB-478.7m | RMB-165.4m | company cash-burn metric only |
| Capital expenditure | RMB46.7m | RMB74.4m | RMB45.2m |
| Cash balance / liquid resources | RMB3.12bn† | RMB7.07bn‡ | RMB5.31bn§ |
† 2024 year-end “sum of cash and cash equivalents, term deposits, current portion of FVTPL financial assets and restricted cash.” ‡ 2025 year-end sum of cash and cash equivalents, bank deposits, current FVTPL financial assets and restricted cash, reconstructed from disclosed balance-sheet lines. § Company-defined “Cash Balance” at 2025-06-30.
The business meaning of this table is that XtalPi’s growth is now real enough to matter, while the funding model remains balance-sheet-supported rather than internally financed. Revenue acceleration has outrun cash conversion, but not by enough to threaten solvency because the cash cushion is large.
Capital-markets timeline
| Event | Date | Terms |
|---|---|---|
| IPO price | 2024-06-13 | HK$5.28 per share |
| January placing | completed 2025-01-24 | net proceeds about HK$1.125bn |
| February placing | completed 2025-02-25 | net proceeds about HK$2.080bn |
| August placing | announced 2025-08-29 | estimated net proceeds about HK$2.630bn |
| Convertible bond | completed 2026-01-28 | HK$2.866bn zero-coupon CB due 2027, conversion price HK$13.85 |
The pattern is clear: XtalPi has had no trouble raising money. That is a competitive advantage at the company level and a dilution risk at the shareholder level.
Collaboration economics snapshot
| Partner | Announcement date | Near-term economics | Headline total potential value |
|---|---|---|---|
| DoveTree | 2025-06-23 / 2025-08-05 | US$51m received; further US$49m potential payment | up to US$5.89bn plus royalties |
| Lilly via Ailux | 2025-11-05 | upfront and near-term milestones in tens of millions of USD | up to US$345m |
| Unnamed large biopharma GPCR partner | 2026-06-09 | upfront plus fully funded early R&D | over US$400m |
| Pfizer | 2025-06-29 | platform expansion, no headline financial amount disclosed | not disclosed |
This table is the core of the valuation debate. Bulls see validation and optionality. Bears see theoretical totals that can sit far above near-term cash. Both readings are supported by the disclosed deal structures.
Research uncertainties
The public record is rich on announcements and half-year / full-year results, but limited on standalone quarterly detail. That means “last four quarters” analysis is necessarily reconstructed from half-year and full-year disclosures rather than from four separate reported quarters.
The exact repeatability of 2025’s large collaboration revenue is still uncertain because the company has not yet published enough periods showing how quickly deal announcements convert into recognized revenue across multiple customers.
The split between maintenance capex and growth capex is not disclosed, so owner-earnings analysis necessarily uses a conservative approximation by deducting all disclosed capex and lease payments.
The public disclosures do not yet allow a clean separation between “platform subscription / repeatable software-like revenue” and “project / milestone / services revenue” at the granularity investors would ideally want.
Sources
Primary sources were HKEX and company investor-relations disclosures: 2024 annual results announcement, 2025 interim results announcement, 2025 annual results announcement, Chapter 18C-related materials, placements, convertibles, buyback announcement, CFO appointment, and collaboration announcements with DoveTree, Lilly, and the June 2026 GPCR partner.
Secondary sources used for market structure and context included HKEX’s Chapter 18C guidance, HKEX Tech 100 materials, Reuters reporting on pharma AI adoption, Reuters and market-data quote pages, and peer-company earnings disclosures from Schrödinger, Recursion, Certara, and WuXi AppTec.
Other tickers mentioned
- RXRX.US: closest public TechBio platform comparison for collaboration-driven monetization and cash-burn tradeoffs
- SDGR.US: closest public software-and-physics comparator for scientific modeling economics
- CERT.US: useful benchmark for cleaner biosimulation software valuation and customer embedment
- 02359.HK: large-scale tech-enabled CRDMO reference for what mature scientific infrastructure looks like
- PFE.US: strategic platform partner in AI-driven molecular modeling
- LLY.US: counterpart to the Ailux bispecific-antibody collaboration
This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.
Full report
Sign in to read the full report
Sign up free to unlock the full text, the Baillie growth scorecard, and full-text search.
Log in / Sign up free