HomeArticle

Overseas AI4S Top 30: Who is turning scientific discoveries into business

潮涌AI2026-11-17 09:53
AI4S will eventually return to three simple and unpretentious things.

Financing news in the AI for Science sector can easily be turned into another ranking list sorted by transaction amount.

In the article "Tech Giants Are Scrambling for AI Pharmaceutical R&D, With the Core Focus on AI4S", Tide AI has elaborated on why both technology enterprises and pharmaceutical companies are willing to invest heavily in scientific discovery simultaneously.

However, three things are completely different: a model leading on benchmark tests, a candidate drug entering Phase III clinical trials, and software generating 200 million USD in revenue. A cooperation agreement worth up to 1 billion USD offered by a pharmaceutical company does not equal that the funds have already been received...

To this end, Tide AI has screened 30 representative enterprises in the overseas AI4S track (comprehensively considering technical verification, commercialization evidence, financing and industrial cooperation, selecting 10 enterprises respectively in three directions: Materials and Chemistry, AI Pharmaceuticals, and other Scientific Intelligence fields). Combining the development status and cutting-edge layout of overseas enterprises, we will jointly explore where the continuously warming AI4S wave is heading.

The following content is sorted out by the Tide Editorial Department, and the order does not represent investment rating or market ranking ——

01

Materials and Chemistry: Enter the R&D process first, then go through experimental verification

01 Schrödinger: The most transparent revenue, and the most visible input cost

Company Business:

Schrödinger integrates computational chemistry software, molecular simulation and in-house drug discovery businesses, which is one of the few companies in this group that can disassemble its business model through public financial reports. In 2025, the company's total revenue reached 255.9 million USD, including 199.5 million USD from software revenue and 56.4 million USD from drug discovery revenue; the annual contract value (ACV) of its software business was 198.5 million USD; the net loss in the same period was 103.3 million USD.

Founding Team:

The company was co-founded by computational chemistry researchers including Richard Friesner and William A. Goddard III, and the current CEO is Ramy Farid. Its accumulation started from physical modeling and medicinal chemistry, rather than adding scientific interfaces to general models after the boom of large models.

Prospect Highlights:

The software business has proved that customers are willing to pay continuously for R&D tools, while drug discovery provides higher equity flexibility and brings heavier R&D investment. When evaluating this company, we should not only focus on software renewal rate and ACV, but also check whether its pipelines can convert losses into realizable clinical and licensing value.

02 SandboxAQ: Rewrite the quantum computing narrative into enterprise scientific tools

Company Business:

After being spun off from Alphabet, SandboxAQ mainly promotes Large Quantitative Models, and serves materials, pharmaceuticals, finance and security scenarios with quantum physics, chemical simulation and AI. In April 2025, the company announced that its Series E financing exceeded 450 million USD, and the total accumulated financing since the spin-off exceeded 950 million USD.

Founding Team:

CEO Jack Hidary has long been engaged in deep tech entrepreneurship and investment, and Eric Schmidt serves as the chairman of the board. The advantage of the team lies in uniting scientific research talents, enterprise sales resources and capital under a unified narrative system.

Prospect Highlights:

What SandboxAQ really needs to prove is that large models can be continuously converted into enterprise contracts, rather than just a cooperation list covering multiple popular tracks.

03 Quantinuum: Quantum hardware companies begin to accept public market pricing

Company Business:

Quantinuum was formed by the merger of Honeywell's quantum computing business and Cambridge Quantum, covering ion trap hardware, compilers, software and quantum chemistry applications. It is not a pure AI materials company, but its computing capability targets chemical and material simulation, so it is included in this group for observation.

Founding Team:

The company merged two teams respectively responsible for hardware and quantum software, and the current CEO is Rajeeb Hazra. Compared with many quantum companies that only talk about long-term computing power, it has to face the inspection of listed companies' revenue, losses and delivery rhythm.

Prospect Highlights:

The company has completed its initial public offering after over-allotment. According to Morningstar reports, the IPO raised about 1.68 billion USD, with an issue price of 60 USD, corresponding to an issue valuation of about 15.6 billion USD. Listing brings sufficient funds, and also makes operating data beyond "quantum advantage" a continuous assessment indicator.

04 CuspAI: Organize the full-chain resources required for material discovery through industrial alliances

Company Business:

CuspAI's MIRA platform starts from target performance, and connects generative design, simulation, synthesis route planning and experimental verification. In July 2026, the company completed a 450 million USD Series B financing led by Kleiner Perkins and NEA, with a reported valuation of 2.6 billion USD.

Founding Team:

Co-founder Chad Edwards and machine learning scholar Max Welling complement each other in industrialization and model research respectively; Yann LeCun, Geoffrey Hinton and other industry elites are on the advisory list.

Prospect Highlights:

The AI Materials Foundry launched at the same time has more than 45 founding members, with NVIDIA providing computing power, Meta FAIR and other institutions providing models and research capabilities. These members are ecological collaborators, and should not be directly regarded as paying customers. The next deliverable for CuspAI is synthetic, scalable materials that can enter industrial processes, rather than the size of the alliance itself.

05 Periodic Labs: 300 million USD seed round, betting on AI scientists that can perform experiments automatically

Company Business:

Periodic Labs combines models, robots and wet experiments, with the goal of enabling AI to put forward hypotheses, operate experiments and continue learning from the results. In 2025, the company completed a 300 million USD seed round financing.

Founding Team:

Liam Fedus once participated in the R&D of models such as ChatGPT, and Ekin Dogus Cubuk came from Google DeepMind's material research team. The two represent the two paths of cutting-edge models and computational materials respectively.

Prospect Highlights:

The experimental results disclosed by the Neon system belong to the company's internal benchmarks. The truly critical points are whether external laboratories can replicate the results and whether the automated closed loop can run stably.

06 Citrine Informatics: Do not chase the largest model, sort out the enterprise's own material data first

Company Business:

Citrine's platform unifies formula, process, experiment and performance data, and then carries out performance prediction, formula optimization and experiment recommendation. It serves the existing R&D processes of material enterprises, rather than requiring customers to accept a fully automatic scientific laboratory first.

Founding Team:

Greg Mulholland and Bryce Meredig founded the company in 2013, bringing backgrounds in material engineering, commercialization and material machine learning respectively. More than ten years of industry accumulation makes it more like an enterprise R&D data infrastructure supplier.

Prospect Highlights:

PitchBook currently records that its total accumulated financing is about 73.7 million USD. The value of this type of company depends on whether customers continue to import historical data into the platform, and whether model suggestions can be applied to real formula and mass production decisions.

07 Uncountable: Hide AI in the R&D data system, which is closer to cash flow instead

Company Business:

Uncountable provides a cloud-based R&D information management platform, covering electronic experiment records, experimental data integration, formula and AI optimization. In June 2025, the company completed a 27 million USD Series A financing led by Sageview Capital.

Founding Team:

Noel Hollingsworth, Jason Hirshman and Will Tashman founded the company together. The team does not focus on a single material model, but solves the widespread data silo problem in chemical, food, cosmetics and pharmaceutical enterprises.

Prospect Highlights:

The company's official announcement shows that it has more than 120 customers and about 90 employees, and claims to maintain profit growth in the past 9 years. Its enlightenment is very straightforward: Scientific AI does not always have to charge for "discovering new substances", it can also charge for data and decision-making of R&D organizations first.

08 Radical AI: From knowledge retrieval to experimental closed loop, cases still need third-party reproduction

Company Business:

Radical AI connects scientific literature, material models and automated experiments, targeting alloys, energy and industrial materials. Different from simply generating candidate structures, it emphasizes the complete process from knowledge extraction to experimental feedback.

Founding Team:

Founder and CEO Joseph Krause positions the company as a builder of material R&D infrastructure. The team needs to deal with literature copyright, data quality, experiment execution and industrial verification at the same time, which is more difficult to organize than a single model product.

Prospect Highlights:

Wiley's cooperation case discloses that the project used 195,359 papers, generated more than 1,000 candidate materials, and completed phased results in 16 weeks. Follow-up attention should be paid to whether external customers can replicate the efficiency improvement.

09 Aether Biomachines: Bring protein design out of the pharmaceutical sector and sell it to industrial material scenarios

Company Business:

Aether uses the Protein Function Model to design proteins with specific functions, with applications including 3D printing additive materials, rare earth extraction, plastic recycling and carbon capture. In December 2025, the company added 15 million USD in new financing, bringing the total accumulated financing to 64 million USD.

Founding Team:

Founder and CEO Pavle Jeremić leads the team to build its own high-throughput screening and experimental data system. Compared with platforms that only do sequence generation, it pays more attention to functional data and industrial applications.

Prospect Highlights:

The company discloses that products such as RapidPrint and Ultra outperform its selected industry benchmarks in printing speed and strength. The next step is to focus on customer certification, bulk procurement and stability under different working conditions, rather than directly equating experimental results with scale revenue.

10 Discovered Materials: Cut in from chip heat dissipation, cross the synthesis threshold first

Company Business:

Discovered Materials uses AI research agents to handle the simulation, synthesis and experimental verification of semiconductor materials, and first solves specific problems such as chip heat dissipation. In August 2026, the company completed a 9 million USD seed round financing led by Lightspeed India Partners.

Founding Team:

Advaith Sridhar and Akash Ramdas met at IIT Madras; Ramdas has a PhD in materials science from Stanford, and Sridhar has experience in AI model and agent R&D. The combination of the two corresponds to the two core capabilities of materials science and model engineering.

Prospect Highlights:

The company has disclosed candidate materials and the Material Discovery Bench, but in the path presented on its research page, more than 500 computational candidates finally converge to 1 relatively credible synthesis route. This proportion is not humiliating, on the contrary, it illustrates the difficulty of material discovery: Generating candidates is only the starting point, and being synthetic, testable and accessible to fabs is where the subsequent costs lie.

02

AI Pharmaceuticals: Platform stories begin to give way to specific candidate drugs and cash terms

01 Isomorphic Labs: One of the most well-funded players, human clinical evidence is still ahead

Company Business:

Isomorphic Labs starts with AlphaFold related capabilities, builds an AI-first drug design platform, and carries out discovery cooperation with pharmaceutical companies such as Eli Lilly and Novartis. In May 2026, the company completed a 2.1 billion USD Series B financing, and the total of two rounds of external financing is about 2.7 billion USD.

Founding Team:

The company was founded by Demis Hassabis, and its core team comes from Google DeepMind and the drug R&D field. It tries to expand structural prediction into a complete molecular design and R&D decision-making system, rather than staying at the protein structure database level.

Prospect Highlights:

As of the public progress, many of its projects are still in preclinical stages such as IND-enabling, and it cannot be concluded that the candidate drugs have obtained human verification. The most important node in the future is not the model ranking list, but the first IND and the first human trial.

02 Xaira Therapeutics: Starting with nearly 1 billion USD, betting on a complete AI biological R&D organization

Company Business:

From its establishment, Xaira has laid out models, data, experiments and drug pipelines at the same time, covering platform modules such as X-Cell, X-Design, X-Patient, X-Scientist and X-Automate. The startup financing is close to 1 billion USD, which gives it the financial conditions to build its own data and experimental system from the very beginning.

Founding Team:

The company is incubated by ARCH Venture Partners and Foresite Labs, Marc Tessier-Lavigne serves as CEO, and the team brings together researchers from Stanford University, Genentech and cutting-edge AI fields. It is not a single algorithm team, but a research organization configured as a large biotech.

Prospect Highlights:

High startup capital allows Xaira to build its own data and experimental system, but also raises the requirements for clinical delivery. To judge its progress, we should check the candidate drugs, indications and trial stages one by one, instead of replacing R&D evidence with the label of "AI-native".

03 Chai Discovery: From structural foundation models to pharmaceutical enterprise workflows

Company Business:

Chai Discovery develops generative models for protein, antibody and molecular interactions, and extends model capabilities to pharmaceutical R&D. In 2026, the company completed a 400 million USD Series C financing, with total accumulated financing exceeding 600 million USD and a reported valuation of 3.8 billion USD.

Founding Team:

Joshua Meier, Jack Dent and other partners founded the company, and the team has experience in both biological foundation models and engineering products. Compared with companies that build a large number of pipelines on their own, Chai is more inclined to turn models into general tools for external R&D teams.

Prospect Highlights:

The antibody design hit rate disclosed by the company is about 16%, which comes from preprints and enterprise test results, and cannot be directly extrapolated to the clinical success rate. The cooperation information of Eli Lilly, Novartis, Pfizer and other enterprises shows that the model has entered the industrial process, but the payment scale and R&D contribution still need to be verified separately.

04 Generate Biomedicines: Phase III pipeline brings platform companies into the hard verification stage

Company Business:

Generate Biomedicines takes generative protein design as the core, and promotes both self-developed pipelines and external cooperations. GB-0895 has entered Phase III for severe asthma, and has launched a Phase Ib study for COPD.

Founding Team:

The company is incubated by Flagship Pioneering, co-founded by Gevorg Grigoryan and other partners, and the current CEO is Mike Nally. Generative models, protein engineering and clinical development are integrated in the same organization, which means it has to bear the full capital cycle of a biotech.

Prospect Highlights:

Regulatory documents show that the company's revenue in the first quarter of 2026 was 7.2 million USD, and cash and marketable securities at the end of the period were 516.6 million USD. These figures are more informative than labels like "the world's first": the subsequent value depends on the clinical readout of GB-0895, cash consumption and whether other candidate drugs can follow up.

05 Recursion: From automated experiments to capital expenditure, large-scale operation has a clear bill

Company Business:

Recursion combines cell images, automated experiments, computational models and clinical pipelines, and expands its capabilities through mergers, acquisitions and transactions. In October 2026, the company announced a 200 million USD oversubscribed follow-on offering to supplement funds for R&D and operation.

Founding Team:

Chris Gibson and other