XtalPi Holdings Turns to Loss in Half-Year: AI Drug Discovery Enters the Deep Water Zone of High Investment and Strong Realization

On the evening of July 24, XtalPi announced a profit warning for the first half of 2026, forecasting revenue between RMB 380 million and 400 million, with a net loss attributable to shareholders ranging from RMB 215 million to 275 million; in the same period last year, revenue was RMB 517.1 million, with a net profit attributable to shareholders of RMB 82.8 million. The reversal in financial statements is primarily due to the rhythm of revenue recognition: in the first half of 2025, the DoveTree pipeline licensing cooperation contributed an initial payment of USD 51 million, approximately RMB 365.1 million; in the first half of this year, the same project recognized a second payment of USD 19 million, approximately RMB 129.4 million. Excluding this project, XtalPi expects endogenous revenue to be no less than RMB 250 million, representing year-on-year growth exceeding 65%, among which the revenue from AI for Science smart solutions is expected to be no less than RMB 180 million, with year-on-year growth exceeding 120%. R&D expenses for the first half of the year are expected to be no less than RMB 340 million, with year-on-year growth exceeding 50%.

The decline in profits coincides with accelerated business operations. In June, XtalPi just signed a collaboration agreement for GPCR oral small molecule drugs with an international biopharmaceutical company, with a potential total value exceeding USD 400 million; in February, its robotics laboratory delivered an automated workstation for formulation stability testing to BASF. Quantum physics, AI models, automated experiments, and drug pipelines are being integrated into a unified R&D system. Short-term financial statements bear higher investment costs, while long-term returns depend on endogenous revenue, customer repurchase rates, pipeline milestones, and the replication speed of experimental platforms.

After the ebb of one-time licensing, endogenous business begins to shoulder the main role in growth

In 2025, XtalPi achieved revenue of RMB 802.6 million, a year-on-year increase of 201.2%; annual profit was RMB 134.6 million, and adjusted profit was RMB 258.2 million, marking the first time the company achieved annual profitability. Among these, revenue from drug discovery solutions was RMB 537.9 million, with a single client contributing RMB 365.1 million; revenue from AI for Science smart solutions was RMB 264.7 million, a year-on-year increase of 62.6%. The annual profitability is of milestone significance, and the pull effect of large upfront payments on revenue and profit also needs to be acknowledged.

AI pharmaceutical companies rarely enjoy the smooth curves typical of traditional software firms. Technical services recognize revenue as projects progress, and pipeline collaborations are influenced by upfront payments, R&D milestones, clinical progress, and commercialization outcomes. A single collaboration can quickly boost current-period revenue, but even if R&D proceeds normally, book growth rates may subsequently decline. The shift to a net loss in the first half of this year mainly reflects the disappearance of a high base and front-loaded R&D expenditures. Whether customer demand has weakened still requires judgment based on recurring business metrics.

After stripping out the DoveTree project, revenue growth exceeded 65%, and AI for Science business growth exceeded 120%, making the operational profile clearer. The company attributes growth to technological upgrades, new customer acquisition, existing customer repurchases, and the scaling up of robotics laboratories and intelligent services. In 2025, the repurchase rate for AI for Science intelligent services exceeded 75%, with algorithms, experimental workflows, data management, and automated equipment beginning to enter sustained cooperative relationships.

Drug discovery business provides milestone-based returns, while robotics laboratories and intelligent services provide relatively continuous income. The former amplifies the economic returns of successful pipelines, while the latter accumulates customers, experimental data, and project cash flows. XtalPi still needs to reduce the impact of single licensing deals on annual profits, allowing recurring business to cover more fixed expenses.

After the official release of interim results, gross margin, accounts receivable, contract liabilities, and operating cash flow will provide more information than the net profit range. If rapid expansion of endogenous revenue is accompanied by improved gross margins, R&D investments will have the opportunity to be diluted by a larger revenue scale; however, if revenue increases continue to rely on equipment customization and staff expansion, profit recovery will still face pressure.

Algorithms enter wet labs, competition begins to extend to heavier R&D infrastructure

In 2025, XtalPi's R&D expenditure was RMB 569.2 million, a year-on-year increase of 36.1%; for the first half of 2026, R&D expenses are expected to be no less than RMB 340 million, with year-on-year growth exceeding 50%. Calculated based on the forecasted revenue range, R&D expenses account for 85% to 89% of the revenue for the same period. Investments are concentrated in autonomous laboratories, agent systems, in-development pipelines, and multimodal technology platforms. The company also launched the open intelligent R&D platform, XtalPi Science.

This round of investment addresses a long-standing challenge in AI pharmaceuticals. Models can quickly generate candidate molecules, but synthesis, screening, analysis, and feedback still require wet lab completion. If experimental capacity cannot expand synchronously, computational speed is difficult to fully transmit to the R&D cycle. XtalPi has built a collaborative system integrating Scientific AI, Physical AI, and Agentic Systems, enabling models to propose solutions, robots to conduct experiments, agents to schedule tasks, and experimental results to feed back into the models.

In the field of AI pharmaceuticals, "readiness" has expanded to include high-quality data, automated experimental capabilities, and reusable processes. Algorithms are responsible for finding possibilities, while experimental systems are responsible for eliminating wrong answers. Stable wet-lab feedback determines whether models can move from generating candidate molecules to delivering clinical assets.

XtalPi disclosed in 2025 that it possesses over 200 full-chain AI models and upgraded its AI for Science business into an R&D system covering supply chain, molecular design, synthesis, screening, and process optimization. The workstations delivered to BASF cover sample management, testing analysis, and data management, expanding the business scope from drug discovery to chemical and materials R&D.

However, there is also an engineering challenge. Robotics laboratories possess attributes of equipment, integration, and customization. The greater the difference in customer scenarios, the harder it is to dilute the costs of engineers, hardware, and implementation. XtalPi needs to increase the proportion of standardized modules, allowing the same set of equipment, software, and agents to enter more customers.

The efficiency formed by R&D investment ultimately falls on delivery cycles, repurchase rates, gross margins, and unit project costs. The number of models can demonstrate technical reserves, but project replication proves platform attributes. Future leaders in the AI pharmaceutical industry will likely need to manage both algorithmic efficiency and laboratory capacity simultaneously, blurring the boundaries between software companies and R&D service providers.

Platform revenue and pipeline rights expand simultaneously, but cash flow remains a long-term constraint

On June 9, XtalPi announced a collaboration with an international biopharmaceutical company to develop GPCR target oral small molecule drugs. The partner pays an upfront fee and bears all early-stage R&D costs. XtalPi can also receive pre-clinical, clinical, and commercialization milestone payments and sales royalties, with the project's potential total value exceeding USD 400 million. The transition from pilot to formal development shows that international pharmaceutical companies are willing to entrust the early-stage R&D of complex targets to the XtalPi platform.

This structure balances near-term R&D revenue with long-term pipeline rights and alleviates capital pressure on individual pipelines. The potential total value cannot be directly equated with orders or future revenue, as pre-clinical studies, clinical trials, and commercialization all carry uncertainties. Subsequent observations should focus on the size of upfront payments, delivery of candidate molecules, IND advancement, and the receipt of milestone payments.

XtalPi has already established five technical platforms for small molecules, macromolecules, molecular glues, peptides, and small nucleic acids. Service business provides customer entry points, proprietary or shared equity pipelines raise the revenue ceiling, and AI for Science solutions migrate technological capabilities beyond drugs. Expanding business boundaries also increases the difficulty of capital investment, project selection, and organizational management.

At the end of 2025, XtalPi's cash and cash equivalents were RMB 2.5731 billion, with a net outflow from operating activities of RMB 165.4 million, narrowing compared to RMB 478.7 million in 2024. Cash reserves can support investments in models, laboratories, and pipelines. The company still needs to avoid long-term reliance on single licensing deals and external financing. Whether the AI for Science business can maintain growth and gradually improve operating cash flow will test the quality of R&D investment.

XtalPi currently has the financial conditions to bear short-term losses but faces higher fulfillment requirements. Continuous growth in R&D expenses can expand technological depth, but too many R&D projects may also disperse resources. Platform enterprises need to make choices between breadth and focus: which pipelines retain long-term rights, which projects are primarily service-income driven, and which robotic capabilities are suitable for standardized output will all affect the quality of cash flow in the coming years.

Industry competition will increasingly compare experimental closed loops, proprietary data, delivery efficiency, and pipeline outcomes. Whether customer repurchases can translate into stable revenue, whether robotics laboratories can replicate across industries, whether collaborative pipelines can continuously enter clinical stages, and whether operating cash flow can improve with scale expansion will jointly influence XtalPi's long-term position.

The most scarce capability in AI pharmaceuticals is the continuous elimination of wrong answers and the delivery of the few correct ones into clinical trials; generating more molecules is merely the starting point. By choosing to bear heavier experimental and asset responsibilities, XtalPi has opened up its growth ceiling, but operational difficulties have risen in tandem. Short-term losses can be understood, but repeatable R&D achievements and revenue returns are the source of long-term trust.