Introduction: AI pharma is moving from "proof of concept" to "value realization." This article breaks down how AIDD is reconstructing the entire value chain of new drug R&D, and through Nvidia's investment map, reveals the ecosystem landscape of this track.
Part 1 Foreword: When "Alchemy" Meets "Code," the "Impossible Triangle" Is Being Broken
On the investment map of biomedicine, we are accustomed to using "ten years, one billion dollars, one-tenth" to describe the brutal law of new drug R&D: a new drug takes an average of 10 years and $1 billion in investment, and the probability that a candidate drug entering clinical trials will ultimately be approved is less than one-tenth. This law is like a curse, locking in the industry's efficiency ceiling and causing countless capitals to shrink back amid long waits and high sunk costs.
However, when AlphaGo's algorithmic logic began to reconstruct predictive models of molecular structures, and when the Transformer architecture unearthed hidden correlations in massive biomedical data that human scientists could hardly perceive, a silent revolution was taking place deep in laboratories. This is not merely an iteration of tools, but a reconstruction of underlying logic.
We are at a critical turning point: artificial intelligence drug design (AIDD) is crossing the critical threshold from "proof of concept" to "value realization." Globally, multinational pharmaceutical companies are casting votes of confidence in AI upstarts through multi-billion-dollar BD deals; domestically, with falling computing costs and the accumulation of high-quality biological datasets, AIDD is no longer merely a simple extension of CADD (computer-aided drug design), but has begun to reshape the entire value chain from target discovery to clinical trial design.
Although the FDA has not yet approved any drug fully discovered and designed by AI to be marketed, multiple AI pharma companies have significantly accelerated the pace of advancing clinical candidates. In July 2026, Insilico Medicine announced that its AI-driven oral TNIK inhibitor Rentosertib officially launched Phase III clinical trials (GENESIS-IPF-3), becoming the world's first AI-driven innovative drug to enter Phase III. The target of this drug was identified by PandaOmics from multi-omics data, and the molecular structure was generated and optimized by Chemistry42, relying entirely on Insilico Medicine's generative AI platform Pharma.AI. Figure 1 summarizes the current clinical stage distribution of some AI pharma pipelines.
But AIDD remains shrouded in fog. Algorithms are powerful, but can they truly understand the complex pathophysiological environment of the human body? How will "data silos" and the scarcity of high-quality annotated data be overcome? For investors, the core question is: in this wave of technological revolution, who has a true "moat," and who is merely an "castle in the air" packaged with AI concepts?
We believe that the ultimate significance of AIDD is not merely to shorten several years of R&D cycles or save tens of millions of dollars in costs, but to transform new drug R&D from a trial-and-error art with a "gambling" nature into a predictable, repeatable, high-probability precision science through computational simulation and prediction.
In the following articles, we will analyze the technological core of AIDD and, through Nvidia's investment map, gain insight into how it screens and builds the future ecosystem of AIDD.
Part 2 How AI Reconstructs the Entire Value Chain of New Drug R&D
2.1 Protein Structure Prediction: AI Pries Open the Bottleneck of Target Discovery
Protein structure prediction is a key step in target development. However, traditional protein structure determination methods, such as X-ray crystallography, nuclear magnetic resonance spectroscopy, and cryo-electron microscopy, have limitations in equipment requirements, protein structure complexity, and time consumption. To address these challenges, scientists have turned to artificial intelligence algorithms to predict protein structures, bringing major advances such as AlphaFold 3 and RoseTTAFold.
In June 2024, the DeepMind team published research results on AlphaFold 3 in Nature (Abramson et al., Nature, 2024), marking a new era in protein structure prediction. Unlike previous generations that could only predict single protein chains, AlphaFold 3 introduced a new Pairformer architecture and diffusion model, capable of uniformly predicting the complex structures of all life molecules—including proteins, DNA, RNA, small molecule ligands, and ions; in benchmark tests for protein-ligand complex structure prediction, its accuracy improved by about 50% compared with traditional physics-based scoring methods. This means that prediction of the key "target-drug" binding mode in drug discovery can yield candidate conformations close to experimental precision within minutes, significantly alleviating the bottleneck constraint of structural biology on new drug R&D, though wet-lab validation is still required.
Representative companies have already emerged in this direction, such as Isomorphic Labs, incubated by DeepMind, dedicated to using AI to "solve all diseases." Public information shows that its core platform IsoDDE has about twice the accuracy of AlphaFold 3 in protein-ligand complex prediction and is regarded by the industry as a breakthrough at the "AlphaFold 4" level. The company completed a $2.1 billion Series B financing in May 2026, and according to Bloomberg reports, it is negotiating a new round of financing at a valuation of $40–50 billion, making it one of the highest-valued companies in the global AI pharma field, even though no drug has yet entered clinical trials.
2.2 AI Helps Drug Molecule Design Break Through
The acquisition of drug molecules can be divided into screening and design. The goal of screening is to quickly identify effective molecules that can bind to target proteins from massive compound libraries. Traditional high-throughput screening is costly and time-consuming, and is limited in efficiency when facing enormous compound libraries; the rise of computer-aided drug design has improved screening efficiency. Compared with efficiency improvements in the screening stage, AI technology has completely changed drug molecule design itself, especially in de novo molecular design and molecular generative modeling, marking a paradigm shift in this field.
In generative molecular design, deep neural networks (DNNs) learn the latent distribution reflected by molecular descriptors to generate new molecules with desired properties. Because model performance depends heavily on the effective transformation among input representations, latent representations, and output representations, molecular representation methods play a key role in model construction. Currently, they are mainly divided into four categories: one-hot encoding, SMILES strings, molecular graphs, and three-dimensional structures.
Between 2024 and 2025, Nature series journals published multiple frontier studies on generative AI molecular design:
▪ SynGFN (Nature Computational Science, 2025) embedded synthetic accessibility constraints into generative flow networks, achieving for the first time "naturally synthesizable" molecule generation—generated molecules not only have ideal biological activity but also come with executable synthetic routes;
▪ ECloudGen (Nature Communications, 2025) used electron cloud representation methods to introduce three-dimensional electron distribution information into molecule generation, significantly improving matching accuracy with target binding pockets;
▪ DrugGEN (Nature Computational Science, 2025) adopted a dual-generator adversarial architecture and generated multiple novel molecular scaffolds with nanomolar-level activity against tumor targets such as AKT1 and CDK2;
▪ RFdiffusion (Science, 2025) successfully designed de novo mini-proteins that can specifically bind intrinsically disordered regions (IDRs), achieving precise targeting of IDR targets long regarded as "undruggable" for the first time.
In the same period, in the antibody and mRNA fields, AI also demonstrated efficient empowerment. In 2025, Science reported research on AI generating GPCR-targeting antibodies within months, while traditional antibody discovery usually takes 12–18 months; in August of the same year, Science published the GEMORNA model, a deep generative model that can design mRNA sequences de novo, optimizing translation efficiency and stability while increasing target protein expression by 41-fold, providing significant commercial value for mRNA vaccine and drug R&D. In addition, the "creativity" of protein large language models has also been verified: ESM3, released by EvolutionaryScale in June 2024, generated green fluorescent protein with entirely new sequences by simulating more than 500 million years of natural evolution, marking a fundamental shift in the molecular design paradigm from "search" to "generation."
In terms of industrial translation, some distinctive projects have also emerged, such as Xaira Therapeutics, Chai Discovery, and Terray Therapeutics.
Xaira Therapeutics was jointly incubated by ARCH Venture Partners and Foresite Labs. Its technology platform X-Suite covers three stages of the entire drug discovery chain: finding targets, making drugs, and selecting patients. Among them, X-Cell is the largest virtual cell model in terms of parameter scale to date (4.9 billion parameters, released in March 2026); X-Design for making drugs is a generative antibody design engine whose core capability is to directly generate antibodies "from scratch" without any experimental data for that target. The company has disclosed cumulative financing of about $1 billion.
Chai Discovery was founded by a team from OpenAI, Meta FAIR, and Absci, with institutions such as OpenAI participating in investment. It is an AI molecular design company that "only does models, not pipelines," with the core positioning of transforming biomolecular design from "experimental trial and error" to "engineered generation," achieving a paradigm-level leap from "reading" to "writing." The company was founded in 2024, completed a $400 million Series C financing in July 2026 at a valuation of $3.8 billion, and has reached cooperation with multinational pharmaceutical companies such as Eli Lilly, Pfizer, and Novartis.
2.3 Compound Synthesis
AI can map molecular structures into forms processable by machine learning algorithms, form multiple synthetic routes based on the structures of known compounds, and recommend the best route. Compared with other links in the pharmaceutical industry, the data volume in compound synthesis is abundant, with chemical reaction data reaching tens of millions of records and relatively high quality. By using AI algorithms to recommend synthetic routes, optimize reaction conditions, and combine them with gradually automated experimental solutions, the time of the synthesis "manufacturing" stage can be shortened, comprehensively accelerating the drug R&D process. At present, databases such as Reaxys and SciFinder, as well as companies such as XtalPi and Yunshen Zhiyao, are all focusing on this link.
2.4 Novel Drug Delivery
Emerging treatment modalities such as gene therapy and cell therapy provide entirely new treatment concepts and means for tumors, rare diseases, chronic diseases, and other refractory diseases, possessing long-term and curative efficacy that general drugs may not achieve, with the hope of completely curing diseases. At present, BDgene, based on virus-like particle (VLP) technology, has achieved in vivo gene editing and mRNA delivery, supporting treatment research for multiple diseases; Dyno Therapeutics uses in vivo experimental data and machine learning to discover and optimize adeno-associated virus capsids, whose performance is significantly superior to existing in vivo gene delivery methods, expanding the range of diseases that gene therapy can treat; METiS Pharmaceuticals uses AI to develop non-viral lipid nanoparticles for nucleic acid drug delivery and sequence design. The depth of AI intervention in this link is currently still limited, and publicly verifiable evidence of clinical benefit is relatively scarce.
2.5 AI-Empowered Clinical Development and Translational Medicine
Because living systems are highly complex and patients are heterogeneous, the same disease may present different pathological states in different patients, and the effects of the same drug also vary. AI can analyze multidimensional data such as electronic health records, whole-genome profiles, and imaging, identify biomarkers associated with disease progression, and accordingly design stratified clinical trials and precisely recruit subjects, thereby reducing clinical development risk.
A typical example is the AMARANTH reanalysis study published in 2025: researchers used an AI prognostic model to re-stratify the failed BACE1 inhibitor lanabecestat trial, observing a 46% slowing of cognitive decline in the slow-progression subgroup (Vaghari et al., Nature Communications, 2025).
Three conclusions: ① AI has evolved from a single-point tool into infrastructure running through the entire value chain of "target discovery—molecular design—clinical optimization—drug repurposing"; ② generative AI is replacing discriminative models as the mainstream paradigm; AI no longer merely "predicts" molecular properties but "creates" entirely new molecules; ③ AI-driven drugs have moved from proof of concept into the period of clinical value realization, and Rentosertib entering Phase III clinical trials is a milestone. This is precisely the underlying logic of "breaking the Double Ten Law."
The maturity of technology means that AI is no longer merely a "supporting tool" for scientists; it is becoming "necessary infrastructure" in the new drug R&D process. When infrastructure tends to perfection, the influx of capital follows naturally—this is precisely the underlying logic of Nvidia's investment map that we will analyze next.
Part 3 More Than Venture Capital: How Nvidia Uses Capital to Define the Commercial Boundaries of AIDD
When studying the industrial ecosystem of AI pharma, BD deals of traditional pharmaceutical companies are certainly important, but what truly defines industry infrastructure is often the upstream "shovel sellers." To judge whether a track is mature, the key is to look at the concentration and irreplaceability of upstream infrastructure. In the AI pharma field, Nvidia is not only the world's highest-market-cap computing power supplier, but also a sharp-eyed "weather vane": its investment portfolio represents not only capital flows but also the commercialization path of technological architecture. By analyzing the biopharmaceutical companies it has invested in, we can glimpse the main commercial battleground of AI pharma in the next five years.
At the same time, it can be seen that unlike traditional VCs that focus only on financial returns, Nvidia's investment logic always revolves around one core: can the technology generate enormous computing power consumption and commercial value within its GPU ecosystem? This unique screening mechanism makes its investment portfolio the industry's most valuable "selected collection." We can clearly see that Nvidia is building a complete commercial loop from "data generation" to "model training" and then to "application landing." Below are the core commercial dimensions of Nvidia's AIDD map that we have sorted out.
3.1 Industrialized Mass Production Faction: The "Best Spokespersons" for Nvidia's Computing Power
For Nvidia, the strategic significance of this type of company lies in "verifying the necessity of computing power." Companies like Recursion and Terray possess massive biological experimental data and are the most typical downstream consumers of Nvidia GPU computing power. Investing in these companies is essentially Nvidia building a "computing power moat" in biopharma: by proving that processing massive biological data cannot do without its GPUs, it establishes itself as infrastructure that cannot be bypassed in the industrialized AI pharma process.
Recursion Pharmaceuticals was founded in 2013 and listed on Nasdaq on April 16, 2021, industrializing drug discovery by decoding biology. In July 2023, Nvidia made a strategic investment of $50 million. Its core platform Recursion OS is an AI-native end-to-end drug discovery and development system that integrates biology, chemistry, and clinical development into a unified intelligent system. Its architecture includes a physical layer (automated wet labs generating standardized data at high throughput) and a digital layer (deep learning foundation models, computer vision, generative AI, causal AI, etc.) in an iterative cycle, generating enormous proprietary data. Its pipeline focuses on oncology and rare diseases, among which the MEK1/2 inhibitor REC-4881 is the first clinical proof-of-concept product of the AI platform, with an indication of familial adenomatous polyposis. Phase II data showed that after 12 weeks of treatment, patients' polyp burden decreased by a median of 43%–47%, and improvement was still maintained 12 weeks after discontinuation.
Terray Therapeutics was founded in 2018 and is headquartered in Pasadena, California. Nvidia participated in its Series B financing in 2024. Since its establishment, Terray has been committed to improving the speed and success rate of small molecule drug discovery and development through large-scale experiments and generative AI. Its core platform EMMI (Experimentation Meets Machine Intelligence) integrates three core layers: ultra-high-throughput microarray chips, large-scale proprietary datasets (more than 17 billion target-molecule interactions), and a full-stack AI model suite (such as the chemical foundation model COATI and the potency prediction model TerraBind). Its core logic is to deeply couple AI and high-throughput experiments in the iterative cycle of "generation—prediction—selection," thereby rapidly discovering and optimizing entirely new molecules on "undruggable" targets lacking known binders.
3.2 Disruptive Technological Innovation Faction: Nvidia's Future "Toolbox"
The existence of this type of company is to "define future algorithm standards." Hardware needs software to drive it. Generate, Evozyne, and EvolutionaryScale represent the forefront of generative protein AI and biological large models. Among them, Evozyne collaborated with Nvidia to develop the ProT-VAE generative model for therapeutic protein design. Nvidia invests in them not only for financial returns, but also to optimize and integrate these cutting-edge algorithm models (such as Chroma and ESM3) into its BioNeMo platform. This is a "technology positioning": when future pharmaceutical companies all use these generative models to design proteins, these models will naturally adapt to and depend on Nvidia's computing architecture, thereby locking in future market entry points.
Generate Biomedicines was founded in 2018 and is a clinical-stage generative biology company. Its technology platform is the self-developed protein generative model Chroma, with the core positioning of learning the "universal grammar" of proteins through AI to realize "on-demand programming" of protein drugs. It was incubated by top venture capital firm Flagship Pioneering, and Nvidia participated in its Series C investment in September 2023. Its core asset GB-0895 is a long-acting anti-TSLP monoclonal antibody, the world's first AI-designed antibody to enter Phase III, for the treatment of severe asthma. Based on the marketed antibody Tezspire (tezepelumab), this drug systematically rewrote its CDR region through the Chroma model, increasing affinity by about 20-fold to 106 fM and supporting dosing once every six months. It took only four years from molecular design to entering Phase III clinical trials. In December 2025, the company announced the launch of two global Phase III clinical trials, SOLAIRIA-1/2; in January 2026, the first patient was dosed; and in February 2026, it listed on Nasdaq.
EvolutionaryScale was founded in 2023, and its founding team comes from Meta's FAIR (Fundamental AI Research) department. It is a pioneer in AI for Biology. Nvidia participated in its seed round investment. In June 2024, the company released the protein language model ESM3 (see Section 2.2), which can generate functional proteins with entirely new sequences by simulating the natural evolution process and is regarded as an important breakthrough in the protein design field.
3.3 Deep Vertical Integration Faction: The "Full-Stack Verifiers" of Nvidia's Technology
The core value of this section lies in "completing the commercial loop." If the first two types of companies focus on computing power consumption at the "R&D end," then Iambic, CytoReason, and Inceptive represent the "application end" and "understanding end" of AIDD. By investing in these companies, Nvidia aims to verify that its technology stack can be used not only for upstream model training but also for downstream inference deployment (such as Iambic's NeuralPLexer NIM) and complex mechanistic understanding (such as CytoReason's disease models). This is a key piece of the puzzle in Nvidia's transformation from "selling graphics cards" to "providing full-stack AI medical solutions," proving that its ecosystem can cover the entire life cycle of drug R&D.
Iambic Therapeutics was founded in 2019 and is a company dedicated to developing new therapies using an AI drug discovery platform. In October 2023, Nvidia participated in its Series B investment. Its core technology platform is called "Molecular Superintelligence," which is essentially an end-to-end drug discovery engine driven by dual AI models and validated through an automated experimental closed loop.
CytoReason is a computational disease model company that "only sells technology and does not make drugs." Its core positioning is to use AI to simulate human diseases at the cellular level, helping pharmaceutical companies transform decisions in the three key links that determine drug success or failure—"selecting targets, selecting indications, and selecting patients"—from "trial and error" to "predictable." It clearly defines itself as a technology company rather than a biotech company. It received strategic investment from Nvidia in 2024 and has now cooperated with multiple multinational pharmaceutical companies such as Pfizer, Sanofi, Roche, Merck, and Fresenius, covering inflammation, immuno-oncology, metabolism, and central nervous system diseases.
Inceptive Nucleics was founded in 2021 and is a biotech company committed to building AI foundation models that can learn the underlying laws of life and applying them to the rational design of nucleic acid drugs. The company focuses on the RNAi field, using AI to guide systematic exploration of sequences and chemical modification schemes to improve drug screening efficiency. Nvidia participated in its Series A financing in September 2023. In June 2026, Inceptive reached a strategic cooperation with the global RNAi leader Alnylam worth up to $2 billion, deeply integrating its generative AI models with the industry's top R&D engine, marking that AI-driven small nucleic acid drug design has officially entered the mainstream pharmaceutical landscape.
3.4 The New Era of 2025–2026: From Investment Portfolio to Strategic Cooperation Ecosystem
If Nvidia's investment in AI pharma in 2023–2024 was still in the "broad casting" layout stage, then in 2025–2026 its strategy clearly upgraded to "deep binding + ecosystem building." Nvidia is no longer satisfied with being a financial investor, but through deep strategic cooperation, joint laboratory construction, and co-building of foundation models, it deeply couples its GPU computing ecosystem with the biopharma industry chain. The most notable feature of this stage is that the cooperation model has upgraded from "investment + technology empowerment" to "jointly defining R&D infrastructure."
3.4.1 Lila Sciences: Building an AI "Science Factory"
In October 2025, Nvidia's NVentures led Lila Sciences' $115 million Series A extension financing, with a valuation exceeding $1.3 billion. Lila was incubated by Flagship Pioneering and is committed to building a "scientific superintelligence" platform—combining specialized AI models with fully automated robotic laboratories (AI Science Factories) to realize a fully automated scientific research loop of "proposing hypotheses—designing experiments—executing experiments—analyzing results—iterative optimization."
3.4.2 Nvidia–Eli Lilly $1 Billion Joint Innovation Laboratory
In January 2026, at the J.P. Morgan Healthcare Conference, Nvidia and Eli Lilly announced a five-year strategic cooperation with total investment of up to $1 billion to jointly build an AI drug R&D joint innovation laboratory in the San Francisco Bay Area, the largest inter-company joint R&D cooperation in the AI pharma field to date. The laboratory is based on the BioNeMo platform and the next-generation Vera Rubin GPU architecture, integrating Lilly's biology and pharmaceutical expertise with Nvidia's AI model construction capabilities. The goal is to build a "continuous learning system" that closely connects Lilly's automated wet labs with computational dry labs, realizing a 7×24-hour AI-assisted experimental loop. Nvidia has thus upgraded from "investing in AI pharma companies" to "directly participating in defining the pharmaceutical industry's R&D infrastructure."
3.4.3 Nvidia–CZI Virtual Biology Program
Nvidia is cooperating with the Chan Zuckerberg Initiative (CZI), founded by Mark Zuckerberg and his wife, in the virtual cell direction. In April 2026, CZI's Biohub announced a five-year $500 million launch of the "Virtual Biology Program," jointly with institutions such as the Broad Institute, Wellcome Sanger Institute, Allen Institute, and Arc Institute, to build human cell prediction models based on a SuperPOD composed of 1,024 H100 GPUs; Nvidia serves as a technical supporter. The goal of the program is to enable scientists to conduct "digital experiments" on complete cells in computers just as AlphaFold predicts protein structures—this is regarded as the next "Holy Grail" of AI biology.
Looking at Nvidia's latest layout, its investment logic has evolved from early "financial investment + technology empowerment" to "strategic co-building + ecosystem lock-in," presenting three core features:
First, from investing in startups to jointly defining R&D infrastructure with industry giants such as Eli Lilly, with investment scale jumping from tens of millions to billions;
Second, from single investment to ecosystem building with the trinity of "investment + platform + model"—through the BioNeMo platform, foundation models such as Evo 2 (a genome-scale DNA language model developed by institutions such as Arc Institute), and the NVentures investment portfolio, building a complete loop from data generation to model training and then to application landing;
Third, the investment direction extends bidirectionally from the "tool layer" to the "application layer + infrastructure layer"—there are both infrastructure such as Lila Sciences' "AI Science Factory" and application-layer layouts such as Proxima (a new Nvidia investment in 2026 focusing on specific drug modalities).
For investors, Nvidia's investment map is not only a weather vane for capital flows but also a "barometer" of AI pharma technology maturity—when Nvidia upgrades from "shovel seller" to "mine builder," it means that AI pharma has moved from the exploration period into the industrialization landing period.
Part 4 Summary and Outlook
AI's transformation of the pharmaceutical industry has crossed the primary stage of "single-point tool replacement" and is reshaping the underlying logic of new drug R&D in the manner of "infrastructure reconstruction." This reconstruction is reflected in three progressive levels:
At the methodological level: AIDD is no longer merely an extension of CADD, but through generative models (diffusion models, large language models, generative flow networks) transforms drug design from "mass screening" to "directed creation"—from AlphaFold 3's precise prediction of complex structures, to SynGFN's realization of "naturally synthesizable" molecule generation, to ESM3's simulation of 500 million years of evolution to generate entirely new fluorescent proteins, AI has proven that it can not only "calculate" but also "create."
At the industry chain level: AI penetration has extended from target discovery and molecular design to clinical development and translational medicine—the AMARANTH reanalysis study published in Nature Communications in 2025 (see Section 2.5) shows that AI-assisted patient stratification can improve clinical trial detection efficiency and subject matching precision; however, it should be noted that its improvement in overall clinical success rates still lacks systematic empirical evidence and should not be directly extrapolated.
At the commercial ecosystem level: Nvidia's investment map, from "computing power throughput machines" such as Recursion and Terray to "algorithm definers" such as Generate and EvolutionaryScale, clearly indicates a trend: competition in AI pharma has evolved from an algorithm contest of "who can make a good molecule" to an ecosystem war of "who can define R&D infrastructure."
But we must also soberly recognize that this paradigm shift is not smooth sailing. Data quality and standardization remain the biggest bottlenecks—the scarcity of high-quality annotated data and the "data silo" phenomenon have not been fundamentally resolved; the gap between algorithm interpretability and regulatory review frameworks still exists, reminding us that AI-generated candidate molecules still need to pass rigorous wet-lab and clinical trial validation; in terms of business models, how AI pharma companies can find a sustainable profit path among "selling software," "selling pipelines," and "self-development advancement" still awaits market verification.
However, the long-term trend is irreversible. When a computing power overlord like Nvidia upgrades from "shovel seller" to "mine builder," when MNCs such as Eli Lilly and Novartis cast votes of confidence with billions of dollars in real money, and when foundation models such as AlphaFold 3, ESM3, and Evo 2 refresh capability boundaries at a pace of iteration every few months, we have reason to believe that the next five years will be a key window for AI pharma to move from "value proof" to "value realization."
For investors, the biggest opportunity is not to chase the conceptual heat of individual AI pharma targets, but to identify long-term builders capable of simultaneously mastering the three capabilities of "high-quality data closed loop," "differentiated algorithm moat," and "clear clinical landing path"—because in this reconstruction, the ultimate winner is not the fastest runner, but the one who can continuously iterate in the "dry-wet loop" and ultimately translate computational advantages into clinical benefit.
[Disclaimer]
The content of this article represents only the views of the company based on current public information and materials provided by invested companies, and does not constitute any investment advice, medical advice, or guarantee of the performance of specific companies. Investment involves risks; market entry should be cautious. Readers should make independent judgments based on public information. Unauthorized reproduction or commercial use is prohibited.