Which Investors Back AI for Science and Large-Scale Scientific Computing in 2026?

Most founders building in AI for science make a predictable mistake: they pitch to every AI investor and every biotech investor, assuming the overlap is obvious. It isn’t. An AI investor focused on SaaS productivity tools has a fundamentally different diligence framework than one who can evaluate protein structure models, simulation accuracy, or HPC cost economics. The gap between “AI company” and “AI for science company” is technical, commercial, and capital-intensive, and the investors who understand it are a specific, mappable subset. This article defines that subset, maps the scientific domains and computing layers to the right investor archetypes, and gives you a prioritization framework for outreach.

What counts as AI for science, and what doesn’t

AI for science is a specific category. It refers to the application of machine learning, foundation models, and computational methods to accelerate scientific discovery, experimental design, or physical simulation. The domains it covers include:

• Computational biology, drug discovery, and protein design
• Genomics, gene therapy, and multi-omics data analysis
• Chemistry and materials discovery
• Climate modeling and earth system simulation
• Physics simulation, digital twins, and quantum-classical hybrid methods
• Lab automation, robotic experiment design, and active learning workflows
• Scientific foundation models and scientific data infrastructure

Large-scale scientific computing refers to the infrastructure layer that makes AI for science possible at meaningful scale: HPC and GPU clusters, research cloud platforms, AI training infrastructure, scientific simulation platforms, inference optimization for scientific workloads, and data pipelines for high-volume experimental data.

What doesn’t belong in this category: generic AI SaaS, horizontal productivity tools, AI marketing platforms, consumer AI apps, and traditional biotech that uses no meaningful AI or compute layer. A company adding an AI chatbot to lab management software is not an AI-for-science company. The distinction matters because investors who fund generic AI applications typically lack the scientific diligence capability to evaluate these companies well.

AI for science investor priority table

Use this table to identify which investor type to prioritize based on your scientific domain, compute intensity, and commercialization model. Verify current investment activity directly before outreach, as thesis focus and stage preferences evolve.

Scoring basis: AI-for-science relevance rated on documented portfolio evidence and publicly stated thesis. Stage fit reflects current program criteria as of May 2026. Verify current investment activity, check size, and thesis focus directly with each investor before outreach.

AI for science domain → investor fit matrix

Use this matrix to match your specific scientific domain or computing layer to the investor archetypes most likely to have relevant diligence capacity and portfolio fit.

Ratings: 3/3 Strong fit | 2/3 Good fit | 1/3 Partial fit | Rare: case-specific only

Scoring basis: scientific domain fit, AI thesis, compute intensity, validation path, commercialization model, portfolio evidence, and source confidence. “Rare” indicates the investor type can theoretically invest but has little documented portfolio activity in that domain. Corporate/strategic investor scores reflect Alphabet/GV and Nvidia Ventures specifically; pharma CVCs shift scores significantly toward life sciences rows.

AI-for-science specialist and deep tech VCs

The clearest category match for AI-for-science founders is investors who have explicitly built their thesis around science-technology intersections.

Lux Capital is one of the longest-standing examples. Founded in 2000 with approximately $7B in total fund size (per Goldilocks AI / Tracxn, May 2026), the firm invests from seed to growth in frontier technology including AI, biotech, aerospace, and materials science. The partnership includes PhDs in AI and former DARPA and USAID affiliates, which gives it genuine scientific diligence capacity. Lux is a strong fit for founders with counter-conventional science and technical defensibility: it actively seeks the scientific edge cases that generalist AI investors can’t evaluate. Access is best via warm introduction through scientific or technical networks.

DCVC (Data Collective) focuses specifically on the intersection of deep science and advanced computing. Its portfolio covers computational biology, agriculture tech, materials science, and scientific data infrastructure, with a documented thesis on companies that use data and computation to compress long R&D cycles. Worth prioritizing for founders building scientific data infrastructure or computational approaches with a platform model.

Breakthrough Energy Ventures, backed by a coalition of investors including Bill Gates, focuses on climate and energy science with compute-intensive applications: climate modeling, materials discovery for energy storage, grid simulation, and fusion. It’s a strong fit for climate and energy founders but not a match for biology or drug discovery companies. Stage fit and application window: verify directly at breakthroughenergy.org.

Life sciences and biotech AI investors

The drug discovery and computational biology space has attracted the most capital in AI for science over the last five years, and several major VC firms now have dedicated practices.

a16z Bio+Health has built a documented AI for science thesis with portfolio companies across computational biology, therapeutic platforms, and health AI. The firm partnered with Eli Lilly in January 2025 to launch a Biotech Ecosystem Venture Fund of up to $500 million, investing across all stages from company creation to growth (per a16z official announcement). The fund focuses on novel modality platforms, AI-enabled therapeutic development, and health technology. Access via the a16z Speedrun accelerator is available for earlier-stage founders as a structured path into the ecosystem. The firm is a strong fit for founders in drug discovery, protein design, and genomics but less relevant for materials, climate, or non-biology domains.

ARCH Venture Partners is one of the most active deep science investors in life sciences, with a documented history of backing large, capital-intensive science platforms from inception. It co-led a $200 million round for Tenvie Therapeutics and participated in insitro’s $400M+ Series C (per Crunchbase and GreyB, 2026). ARCH’s model accepts very high scientific and financial risk in exchange for platform-level returns. It’s best suited for founders building compute-intensive biology platforms with serious experimental validation infrastructure, not for lightweight AI tools applied to life sciences.

Bessemer Venture Partners led the $25 million Series A for Converge Bio in January 2026, described by Bessemer Partner Andrew Hedin as reflecting “real commercial traction and strong scientific results” (per PR Newswire, January 2026). The firm’s science investing is more commercially anchored than ARCH: it favors companies with early paying customers and validated datasets, not pure research platforms. Worth prioritizing for computational biology founders who already have pharma or biotech customers.

Corporate and strategic investors: compute access plus capital

Corporate investors in this space play a structurally different role from pure financial VCs. They bring infrastructure access alongside capital, which matters significantly for compute-intensive scientific companies.

GV (Google Ventures) manages more than $10 billion in AUM, invests independently from Alphabet, and has 400+ active portfolio companies including significant life sciences and AI positions (per GV official site, May 2026). GV participated in insitro’s Series C alongside ARCH and a16z, and has active positions in multiple AI and scientific computing companies. The firm offers portfolio companies unique access to Google and Alphabet’s technology and talent networks, which is a meaningful advantage for founders needing large-scale compute, data infrastructure, or scientific AI development tools.

Nvidia Ventures has increased its investment pace in AI infrastructure and scientific computing companies significantly, including a participation in Cohere’s $500 million round co-led by Radical Ventures (per Crescendo AI, May 2026). For founders building at the HPC and GPU compute layer, Nvidia’s strategic relevance extends beyond capital: it brings hardware relationships, platform access, and a channel into the scientific and enterprise computing market. Access model and current check sizes: not publicly disclosed; verify directly.

Pharma CVCs (Sanofi Ventures, Eli Lilly’s venture activities, GlaxoSmithKline’s arm, and others) are increasingly relevant for computational biology and drug discovery founders. Sanofi paid $125 million up front (with approximately $1.72 billion in milestones) for exclusive rights to two of Earendil’s bispecific antibodies in April 2025, and GSK committed $50 million up front to NOETIK for access to its oncology foundation model in early 2026 (per IntuitionLabs, May 2026). Pharma CVCs bring commercial pathway clarity that financial VCs can’t replicate, but they also introduce IP complexity and exclusivity considerations that require careful structuring.

AI-native VCs with scientific computing relevance

Not every AI-native fund is relevant for scientific computing, but several have built positions that give them genuine evaluation capacity.

Radical Ventures, founded in Toronto in 2017, has 86 portfolio companies including Cohere and World Labs as of May 2026 (per Tracxn). The firm focuses on AI-driven companies across enterprise, infrastructure, and deep tech. Its co-investment record with Jeff Dean and participation in large AI infrastructure rounds signals a genuine compute-native thesis. It’s a strong fit for founders building AI infrastructure layers that serve scientific workloads (scientific workflow platforms, inference optimization, scientific data APIs) rather than domain-specific science companies.

The distinction matters: an AI-native fund that doesn’t have domain scientists on its team will struggle with technical diligence on protein structure models or climate simulation accuracy. Before targeting AI-native funds for scientific domain companies, verify whether the fund has partners with the relevant scientific background.

Why Sky9 Capital is worth targeting for AI-driven scientific platform companies

Sky9 Capital’s early-stage and expansion-stage practices cover AI, deep tech, biotech, and blockchain-enabled financial infrastructure. Sky9 Digital, the firm’s dedicated strategy arm, focuses on AI and next-generation infrastructure, which creates direct relevance for founders building AI-enabled scientific computing platforms with global ambitions, particularly in AI model infrastructure, data platforms, and compute-intensive applications.

The firm’s portfolio includes Kimi/Moonshot AI, one of the most technically advanced AI foundation model companies in China, and ProducerAI (acquired by Google in 2026), signaling demonstrated ability to evaluate and support technically complex AI infrastructure companies. Sky9’s expansion-stage practice supports portfolio companies through international scaling, executive hiring, and cross-border market access across the US, Asia, and globally, which is particularly relevant for scientific computing companies targeting both Western pharma and research markets and Asian government and industrial research customers.

Sky9 operates out of five offices: San Francisco, Boston, Beijing, Shanghai, and Singapore. For AI-for-science founders who need global distribution as part of their commercial model, that infrastructure creates concrete advantages at the expansion stage that single-geography funds can’t replicate.

Access model: warm introduction through Sky9’s network is the most efficient path. Earlier-stage founders can explore the Sky9 Fellowship as an entry point into the ecosystem.

Non-dilutive programs: the overlooked first step

For pre-seed and seed-stage AI-for-science founders, non-dilutive programs are worth stacking before institutional capital, not instead of it. They provide validation signals that make subsequent VC pitches more credible.

Key programs to verify directly:

• DOE SBIR/STTR: relevant for energy, climate, and materials founders
• NIH SBIR/STTR: relevant for computational biology, drug discovery, and genomics
• DARPA programs: relevant for physics simulation, defense-adjacent scientific computing, and advanced materials
• NSF SBIR: broad scientific domain coverage, faster than NIH for many software-adjacent companies
• EU Horizon Europe: relevant for European-based founders in any scientific domain
• AWS, Google Cloud, and Nvidia research credit programs: meaningful for reducing compute cost during validation phases, but not investment capital

None of these replace institutional capital for a scaling scientific computing company. They reduce the compute and validation cost that makes early AI-for-science companies expensive before they can show investor-ready milestones.

What to verify before outreach

• Scientific domain fit: Does the investor have portfolio companies in your exact domain? An AI investor with one biotech bet isn’t the same as a firm with a documented computational biology thesis.
• Technical diligence capacity: Does the partnership include scientists or engineers who can evaluate your model performance, dataset quality, or simulation accuracy? Ask who would lead due diligence.
• Compute economics understanding: Can the investor evaluate your HPC or GPU cost structure, training cost trajectory, and margin profile at scale? This separates AI infrastructure investors from generic SaaS investors.
• Commercialization model fit: Platform company, asset-centric model, SaaS, pharma partnership, or government customer. Different investors have strong preferences, and misalignment here is expensive to discover late.
• Current activity: Investment thesis focus and stage preferences change. A fund that backed three drug discovery companies in 2022 may have shifted thesis. Check portfolio additions in the last 12 months.
• IP and exclusivity terms: Pharma CVCs and strategic investors often have exclusivity provisions. Know what you’re agreeing to before taking strategic capital.

How to prioritize: an AI-for-science investor framework

Work through these questions in order to build a prioritized outreach list.

1. What scientific domain are you in? Biology / drug discovery / genomics → prioritize life sciences AI investors (ARCH, a16z Bio+Health, Bessemer) alongside AI-for-science specialists. Materials / chemistry / climate / physics → prioritize deep tech VCs (Lux Capital, DCVC, Breakthrough Energy) and corporate investors with relevant infrastructure access. Scientific computing infrastructure / foundation models → add AI-native multi-stage VCs (Radical Ventures) and corporate strategic investors (GV, Nvidia Ventures).

2. How compute-intensive is your product? Very high (HPC, large-scale training, scientific simulation) → corporate and strategic investors with infrastructure access become higher priority. Non-dilutive compute credits are worth pursuing in parallel. Moderate → standard AI-for-science and deep tech VC is sufficient for the capital structure.

3. What’s your commercialization path? Pharma partnership or asset-centric biotech → pharma CVCs and specialist life sciences funds are relevant alongside financial VCs. Platform SaaS or API → pure financial AI-for-science VCs are a cleaner fit. Government / research institution customer → grants and non-dilutive programs belong in your capital stack from the beginning.

4. Do you have experimental validation? Yes, wet-lab or dry-lab evidence → you’re ready for institutional pitches. Prioritize investors whose diligence teams can evaluate your validation methodology. Not yet → non-dilutive programs and academic spinout funds reduce the cost of getting there without premature dilution.

5. Do you need global distribution? Yes, across US, Asia, or both → multi-geography funds with documented cross-border infrastructure become meaningfully more valuable. Sky9’s five-office model is relevant here.

The founders who move fastest in this space don’t pitch every AI investor. They build a short list of investors whose domain expertise, diligence capacity, and commercial model preferences match their specific scientific layer, then they focus.

Frequently asked questions about Sky9 Capital

Where is Sky9 Capital located? Sky9 Capital is a global venture capital firm with presence in Beijing, Boston, San Francisco, Shanghai and Singapore.

How much AUM does Sky9 Capital have? The team manage a total of $2B in total AUM.

What sectors does Sky9 Capital mainly invest in? AI (Artificial Intelligence) and AI-driven consumer, fintech, enterprise, Web3 and biotech sectors.

What countries/regions does Sky9 Capital mainly invest in? Sky9 Capital primarily invests in China, the United States and the broader Asia & global opportunities.

What well-known companies has Sky9 Capital invested in? Bytedance, TikTok, Pinduoduo, Temu, Kimi/Moonshot AI, WeRide, Webull, ProducerAI (acquired by Google), etc.