From Trillion-Dollar Computing Power to the "Next-Generation Drug King": Value Divergence and Capacity Fulfillment in NVIDIA's AI4S Ecosystem

Jensen Huang once asserted that life sciences is the most profound application scenario for AI. With NVIDIA and Eli Lilly reaching an AI joint laboratory collaboration worth up to $1 billion, the consensus among top-tier capital has been established. NVIDIA BIO26 Biopharmaceutical Special Report Looking across the life sciences ecosystem companies NVIDIA has officially announced this year, the major trend of betting on life sciences is clearly visible, and the market's valuation perception of this track is also continuously upgrading: Anthropic's valuation approached $1 trillion after completing financing, and it entered the field by deploying the Claude Science R&D platform; Chai Discovery completed $400 million in financing backed by biomolecular models and de novo antibody design capabilities, and successively partnered with pharmaceutical companies such as Eli Lilly, Novartis, and Bristol Myers Squibb; Lila Sciences, with its "AI hypothesis + Siasun Robot&Automation execution" scientific factory concept, saw its valuation pushed up to $8.5 billion; XTALPI (02228) has connected the closed loop of digital reasoning and physical validation, with Agentic AI directly scheduling Siasun Robot&Automation clusters to complete the "design-synthesis-test-feedback" cycle for new molecules and new materials in the real world. This ecosystem map is evolving into a heavily fortified AI4S industry map: Anthropic provides general reasoning, Chai strengthens biomolecular models, Lila and XtalPi explore autonomous laboratories, and companies such as Dassault connect instruments with industrial scenarios. AI4S is bidding farewell to pure "model leaderboard competition" and entering a "productivity race" in the real world. In drug R&D, the cost of a single chemical hallucination is months of time and millions of dollars. Model parameters are merely an admission ticket; the closed-loop experimental results from dry and wet laboratories are the core touchstone determining industrialization market value. Beyond Project Collaboration: Exporting "Core R&D Infrastructure" to Global MNCs Among the ecosystem partners showcased by NVIDIA, XtalPi is one of the few companies from China. Its uniqueness lies not in training yet another drug model, but in having already enabled models, scientific intelligent agents, and Siasun Robot&Automation laboratories to work together. Today, XtalPi's "AI + Siasun Robot&Automation" system has crossed the proof-of-concept stage and gained continuous validation and "repeat purchases" from Eli Lilly, the world's most valuable pharmaceutical company. The collaboration path between the two sides demonstrates extremely strong strategic penetration: from an initial small molecule drug discovery deal with a potential total value of $250 million, expanding to a $345 million bispecific antibody collaboration; recently, it completed the delivery of a multi-million-compound management system for Eli Lilly's Shanghai R&D center, as well as the acceptance of the HTE (high-throughput experimentation) platform. Eli Lilly's continuously expanding repeat purchasesfirst buying small molecule discovery capabilities, then leveraging the AI antibody platform, and finally directly procuring automation infrastructuremark XtalPi's transition from single-project collaboration to an infrastructure supplier for core R&D processes backed by MNC repeat purchases. This logic has already been confirmed at the financial level. In the first half of 2026, after excluding the impact of the prior year's high-base upfront payment, XtalPi's revenue grew 73.8% year-over-year. Among this, AI4S smart solutions revenue reached RMB 193.5 million (a year-over-year surge of 136.4%), accounting for half of the total. The quantification of efficiency is XtalPi's core moat: Agentic HTE compresses the traditional 3-4 week experimental cycle to approximately 6 days; the SureRoute retrosynthesis system reduces the chemical hallucination rate to 4.6%, with Top-1 recommendation accuracy reaching 74.3% (several times that of existing general large models). At the same time, XtalPi systematically disclosed its asset pipeline: nearly 40 proprietary and jointly incubated pipelines, including more than 10 pipelines that have entered IND or IND-enabling stages, covering small molecules, antibodies, peptides, small nucleic acids, and molecular glues. At this point, XtalPi's dual commercial engines have fully taken shape: externally exporting "equipment + platform" to accumulate stable infrastructure revenue; internally generating pipeline assets at scale to pursue milestone sharing and enormous upside from asset breakthroughs. The "Next-Generation Drug King" Revelation: Valuation Leap from Opening Up Undruggable Targets In August of this year, Revolution Medicines' molecular glue drug RASONQUE (targeting RAS) received FDA approval, with Phase III clinical survival nearly doubling and death risk dropping by 60%, making it a veritable "next-generation drug king." Its core revelation is that the value of new modalities is not about grabbing share in traditional red oceans, but about reshaping the boundary of "what can become a drug." Once historically "undruggable" targets are conquered, their asset ceiling will far exceed traditional R&D outsourcing service fees. Greg Verdine, an early pioneer of the RASONQUE approach, later founded DoveTree and reached a sky-high collaboration with XtalPi with a potential total value of up to $5.99 billion. Currently, XtalPi has locked in a total of $70 million in upfront payments, and its first oncology project has advanced to the IND-enabling stage. This not only confirms the extremely high druggability conversion rate of the underlying technology, but also gives the market a clear glimpse of the exponential explosive potential of XtalPi's platform in future business development (BD). This is not only an endorsement by top experts of XtalPi's molecular discovery capabilities, but also a validation of its underlying mechanism. In some projects, XtalPi's XGlue platform optimized target protein degradation activity to picomolar (pM) levels in just a single quarter. In the context of innovative drugs, being able to identify molecules that remain highly effective at extremely low concentrations within a very short time means greatly improving the probability of later clinical success. Three Frontier Modalities and Valuation Anchors: From Single-Point Breakthroughs to "Underlying Capacity Dominance" In the three major new modality tracks of antibodies, peptides, and small nucleic acids, breakthroughs in a single technology platform have all received extremely high pricing from the capital market. This is precisely the best reference framework for re-examining XtalPi's value: Antibody Network: AI design is crossing from technical narrative into the clinical realization phase. XtalPi, relying on the dual-core drive of "dry-wet closed loop + top veteran talent," is accelerating the realization of high commercial premiums from its proprietary pipelines. Chai, valued at $3.8 billion, and Generate Biomedicines, with 5 clinical pipelines in hand, prove the broad prospects of AI antibodies. XtalPi's Ailux platform has not only connected computational design with wet-lab developability assessment, but has also been validated across more than 100 projects. Backed by a bispecific antibody collaboration with Eli Lilly worth up to $345 million, its 3 autoimmune pipelines are directly targeting Phase I clinical trials in 2027. At this critical juncture, Ailux has also recruited Dr. Maria G. Belvisi, former core management member and CEO-3 of AstraZeneca (AZ), as its Chief Scientific Officer. As one of the most significant cases of international big pharma executives joining a Chinese startup, her top-tier industrial operational vision will strongly advance pipeline progression, and XtalPi is truly achieving a value leap from "platform empowerment" to "proprietary rights." Peptide Engine: The hundred-billion metabolic market has ignited M&A frenzy for peptide assets. XtalPi directly addresses R&D pain points with automated synthesis and launches a dimensional strike into the vast consumer functional molecule market. Eli Lilly's peptide blockbuster matrix (generating approximately $36.5 billion in revenue in 2025) and Roche's $5.3 billion heavy bet on Zealand Pharma confirm the powerful cash-generating ability of peptide assets. The core pain point of peptide development has always been the extremely high synthesis and screening threshold. XtalPi's PepiX platform, through AI directly connected to automated synthesis, can lock onto hit compounds in just two months for some oral cyclic peptide projects, directly positioning itself at the pinnacle of upstream production efficiency for peptide assets, while simultaneously laying out in consumer products. Currently, XtalPi has one anti-hair-loss peptide and one food-ingredient-grade peptide that inhibits carbohydrate absorption, each having passed INCI and FDA filings and been approved for market entry, reshaping the consumer functional molecule market with the underlying capabilities of AI drug discovery. Small Nucleic Acids (siRNA): The validation of precise targeting technology has spawned a wave of hundred-billion-level M&A by MNCs. XtalPi, with its high-hit-rate generalized models, has proven the scalable expansion capability of its underlying infrastructure across modalities. Novartis's $12 billion acquisition of Avidity and Alnylam's market value exceeding $33 billion have established extremely high valuation anchors in the small nucleic acid field. Benchmarking against global giants, XtalPi has efficiently laid out 6 siRNA pipelines, more than half of which have completed in vivo efficacy evaluation. Its siDiff model's hit rate in tests with unseen target genes jumped by more than 30%, and its IgA nephropathy project obtained superior non-human primate (NHP) efficacy data validation in just 9 months, a process that traditionally takes 12 to 18 months. This is not only a breakthrough in a single pipeline and a tremendous improvement in efficiency, but also demonstrates to the market that its "AI + experimentation" system can still deliver crushing R&D efficiency when facing entirely new modalities. AI cannot completely eliminate clinical risk, but it can front-load druggability assessment, allowing obviously wrong molecules to be eliminated as early as possible. When established companies have validated the commercial ceiling of new modalities, what XtalPi is competing for is the underlying universal mass-production capability of these all-modality assets. XtalPi's true scarcity lies in the fact that antibodies, peptides, small nucleic acids, and molecular glues are all sharing the same underlying "AI large model + Siasun Robot&Automation" foundational capabilities and infrastructure. When established companies have validated the commercial ceiling of new modalities, what XtalPi is competing for is the underlying universal mass-production capability of these all-modality assets. Reshaping the Valuation Model: From "Project Waves" to "All-Modality Asset Factory" It is already difficult to value XtalPi using a single comparable company. For such a composite infrastructure enterprise, the market needs a brand-new "SOTP (Sum of the Parts)" valuation perspective: its AI4S automation infrastructure benchmarks against Lila Sciences, valued at $8.5 billion; its antibody platform can reference the premium of Chai or Generate; and its molecular glue, peptide, and small nucleic acid pipelines can be valued on a risk-adjusted basis according to R&D stage, collaboration rights, and market space. Stacked on top of equipment sales, R&D services, substantial milestone payments, and a safety cash cushion of nearly RMB 8.7 billion, this constructs a moat with extremely high safety margin and counter-cyclical capability. Revaluation is not about pricing all early-stage pipelines at successful terminal value, but about confirming the underlying switch in its value creation engine: in the past, look at project closings; now, look at the compound growth of experimental equipment, data assets, and multi-modality pipelines. The logic of AI4S industry evolution is already incomparably clear: the first stage competes on model parameters, the second stage competes on closed-loop experimentation, and the final decisive factor is who can solidify experimental capability into an industrial-grade system that "continuously mass-produces high-value assets." NVIDIA's computing power ecosystem has no shortage of stars, but companies that simultaneously hold "continuous repeat purchases from top-tier MNCs," "large-scale automated laboratories," and "an all-modality pipeline reservoir" are exceedingly rare. The success of a single pipeline can only define the price of one transaction, but an AI industrial foundation that continuously mass-produces pipelines will define the platform value of an era.