By Chen Guangjing
Do China’s AI drug discovery champions need to worry about survival?
For Wen Shuhao, co-founder and chairman of XtalPi (2228.HK), the question is less about existential dread and more about strategic pragmatism. “We must survive first,” Wen said during an exclusive interview with Huxiu.
His comment comes against the backdrop of a strong 2025 performance. The Shenzhen-based pioneer in AI-driven drug discovery (AIDD) reported a 200% surge in revenue to 800 million yuan ($110 million) and turned its first-ever annual net profit of 258 million yuan. It also holds more than 7 billion yuan in cash — a position that makes talk of “survival” sound almost paradoxical.
Yet the comment reflects a shifting industry reality. As Nvidia sells computing power directly to pharmaceutical groups and drugmakers build their own bio-supercomputing labs, AIDD has become a crowded and contested field. Long development cycles mean that proving — or disproving — AI-designed drugs takes years. To date, no such drug has reached the market, leaving investors divided over the technology’s promise.
That caution is visible in the stock market’s subdued response to XtalPi’s move into profitability. Analysts point out that 45% of XtalPi’s 2025 revenue came from a single contract with U.S.-based DoveTree Pharmaceuticals. In biotech, such windfalls are rare and often non-recurring; peers like RemeGen and Akeso have famously slipped back into the red after initial bursts of profit.
Unlike traditional biotech groups, however, XtalPi has positioned itself as an infrastructure provider rather than a drug developer.
From algorithms to infrastructure
Founded in 2015 by three postdoctoral researchers from Massachusetts Institute of Technology, XtalPi set out to apply quantum physics and artificial intelligence to drug discovery. The idea — using first-principles calculations to model molecular interactions — was novel at the time and quickly attracted capital. Backers have included Tencent, Sequoia Capital and Google.
But investor expectations are evolving. Early enthusiasm for algorithmic breakthroughs has given way to a demand for scalable, repeatable outputs. In practice, this is pushing companies like XtalPi towards heavier investments in physical infrastructure.
The shift is evident across the industry. In January, Eli Lilly agreed a $1bn project with Nvidia to build a biological supercomputing platform, while Roche soon followed with an even larger deployment of advanced GPUs.
For AI drug developers, the economics are daunting. Clinical trials are expensive and risky, and even well-funded technology groups have struggled to sustain the capital required. Nvidia itself exited its stake in Recursion Pharmaceuticals last year, underscoring the challenges.
XtalPi has chosen a more cautious path. Rather than advancing its own drug pipelines into costly clinical stages, it has focused on refining its algorithmic platform while investing heavily in automated laboratories. These facilities, integrated with AI systems, generate high-quality experimental data and provide services to pharmaceutical partners.
This model is gaining traction. Revenue from XtalPi’s automated lab business has grown rapidly, becoming a core pillar of income once large one-off deals are excluded. The trade-off, however, is lower margins compared with pure software companies.
For Wen, the rationale is clear. “Among algorithms, computing power and data, it is data that ultimately limits large-scale application in biomedicine,” he says. High-quality, reproducible datasets — rather than raw computing capacity — are emerging as the next competitive frontier.
Spending the war chest
XtalPi’s cash pile is central to its strategy. Beyond continued investment in labs and computing infrastructure, the company is pursuing acquisitions to expand its commercial footprint and technological capabilities.
The company also aims to broaden its scope beyond pharmaceuticals into new materials, where it sees a larger long-term market. Its physics-based algorithms are particularly suited to predicting material properties, offering advantages over purely statistical AI models.
At the same time, XtalPi is building an ecosystem of partnerships with biotech start-ups and research groups. By participating in multiple drug development programs — often in exchange for milestone payments and future royalties — it seeks to diversify risk and capture upside without bearing the full cost of clinical trials.
Opportunities and risks
The broader outlook remains uncertain. In new materials, weak intellectual property protection and fragmented data pose challenges, while competition is intensifying as technology groups and industrial players enter the field.
In pharmaceuticals, long timelines and low success rates continue to weigh on expectations. Whether XtalPi’s collaborations will yield blockbuster drugs remains to be seen.
What is less in doubt is the direction of travel. AI is reshaping scientific research and industrial innovation, with the potential to turn drug discovery from a series of high-stakes bets into a more systematic, scalable process.
For companies such as XtalPi, the priority is survival — and the careful deployment of resources — until that transformation fully takes hold
Source:
AGI Channel
