Startup Funding Watch: Venture Capital Moves Deeper Into the AI Stack & Biggest Funding Deals
Venture Capital Moves Deeper Into the AI Stack as Bessemer Raises $5.75B and Snorkel AI Triples to $3.5B
Capital Is Moving Deeper Into the AI Stack — Not Just Into AI Applications
The latest funding activity confirms a shift that has been building for months: venture capital is moving deeper into the AI stack — into training data, enterprise AI platforms, scientific datasets, physical systems, and the funds that will finance the next cycle. Bessemer Venture Partners announced $5.75 billion across two new funds dedicated to investing across the AI stack, explicitly spanning compute, infrastructure, foundation models, developer tools, applications, and agents. Snorkel AI raised $350 million at a $3.5 billion valuation as demand for AI training data accelerates. Basecamp Research raised $140 million at an ~$800 million valuation for AI-designed therapeutics, with Nvidia and Anthropic's Anthology Fund participating.
The pattern is not simply that AI is attracting capital. It is that the market is repricing entire layers of the AI stack — data, training infrastructure, enterprise agents, and scientific applications — as first-class investment categories rather than features of a model company. For founders, the implication is straightforward: the most durable opportunities are no longer limited to building applications on top of AI models. They increasingly sit underneath the application layer — in data, infrastructure, platforms, and the physical systems AI is beginning to touch.
Funding at a Glance
Company Funding Rounds
| Company | Amount | Stage | Sector |
|---|---|---|---|
| Enveda | $311M | Series B | AI Biotech |
| Snorkel AI | $350M | Series E | AI Training Data |
| Hubble Network | $200M | Growth | Space/Connectivity |
| Basecamp Research | $140M | Series A | AI Biotech |
| Vantora | $100M | Series A | Physical AI |
| Ema | $77M | Series A | Enterprise AI |
VC Fund Formation
| Fund | Amount | Focus |
|---|---|---|
| Bessemer Venture Partners | $5.75B | AI stack — seed through growth |
Editorial note: Company rounds and VC fund formation are presented separately. Bessemer's $5.75B is capital formation — a pool to be deployed over multiple years — not a single startup financing.
Biggest Funding Deals
Bessemer Raises $5.75 Billion for the Next AI Investment Cycle — and Says It Will Fund the Whole Stack
Firm: Bessemer Venture Partners | Fund size: $5.75 billion across two funds | Split: $1.75 billion for seed and early-stage; $4 billion for growth-stage | AI portfolio: More than 260 AI-native companies backed since 2022, according to the firm | Stated focus: Compute, infrastructure, foundation models, developer tools, applications, and AI agents. Why this matters beyond the headline number: The important story is not the size of the fund. It is where the capital is intended to go. Bessemer's explicit framing of AI as a multi-layer technology market — rather than a single application category — is a signal that major venture firms are now structuring their entire strategies around the AI stack. Competitive significance: The raise reinforces the distinction that will shape the next cycle: AI funding is becoming an infrastructure story as much as an application story. Firms that can underwrite across infrastructure, data, platforms, and applications will have structural advantages over firms limited to a single layer.
Training Data Becomes a Billion-Dollar Infrastructure Layer as Snorkel AI Triples Valuation
Company: Snorkel AI | Stage: Series E | Amount: $350 million | Valuation: $3.5 billion (up from $1.3 billion) | Lead Investors: Insight Partners, S32 | Participating: Addition, Lightspeed, Greylock, GV, Wells Fargo, Third Point, March, Blumberg, Allegis, Standard VC, Frontline | What it does: Builds training datasets and reinforcement-learning environments for AI labs and enterprises. Traction: Annualized revenue run rate reached $375 million, according to the company, as demand for specialized training data accelerates. Why investors are backing it: Training data is increasingly being treated as infrastructure, not services. The AI value chain is moving from Models → Applications toward Data → Training → Models → Agents → Applications. Snorkel sits much closer to the data and training layer. Competitive significance: The round is the clearest signal yet that AI data infrastructure is now a first-class investment category — one with the potential to support a public company.
Basecamp Research Raises $140M for AI-Designed Therapeutics With Nvidia and Anthropic's Anthology Fund
Company: Basecamp Research | Stage: Series A | Amount: $140 million | Valuation: ~$800 million | Lead Investor: S32 | Participating: Nvidia, Anthropic's Anthology Fund, Catalio Capital Management, and others | What it does: AI-based drug developer using computational biology and proprietary scientific datasets to design new therapeutics. Why investors are backing it: AI is increasingly being financed as a general-purpose technology layer that can be applied to scientific and industrial problems — not just to software. The participation of Nvidia and Anthropic's Anthology Fund signals that the strategic layer of the AI ecosystem sees biotech as a durable destination for AI. Competitive significance: The round shows that AI capital is moving beyond software into biotechnology, drug discovery, scientific datasets, computational biology, and AI-designed therapeutics. AI biotech is no longer a niche vertical — it is a capital formation category.
Ema Raises $77M as Enterprise AI Moves Toward Digital Employees
Company: Ema | Stage: Series A | Amount: $77 million | What it does: Enterprise AI platform designed to perform complete workflows rather than provide chatbot-style assistance. Why investors are backing it: Ema is a case study in the continuing shift toward AI systems designed to execute enterprise work — not just help humans do it. Its model integrates with existing enterprise software and orchestrates workflows across systems, positioning it closer to a digital employee than a traditional SaaS tool. The key question: Are investors funding another application category, or a new enterprise software layer? The answer determines whether Ema and companies like it become features of incumbent platforms or independent categories. Competitive significance: The economics of replacing or augmenting repetitive knowledge work are substantial, but so are the integration and reliability challenges. Early-stage capital is betting that the platforms that solve those challenges will define the next generation of enterprise software.
Vantora Raises $100M to Build Physical AI Startups Inside Industrial Corporations
Company: Vantora (formerly UP.Labs) | Stage: Series A | Amount: $100 million | What it does: Builds startups around physical AI for industrial corporations, combining startup creation with corporate partnerships. Why investors are backing it: Vantora's model combines elements of venture capital, corporate venture building, startup incubation, industrial technology, and M&A. Participating corporations may become customers — or may ultimately acquire the startups. That creates a demand-validated pipeline most AI startups lack. Competitive significance: Physical AI is beginning to create its own capital ecosystem rather than remaining a robotics subcategory. Expect more hybrid startup-builder and corporate-partnership models as industrial corporations look for AI-native capability without building it internally.
The AI Data Economy Is Expanding
Snorkel is the entry point into a much larger category. The current funding environment includes companies focused on:
- Human-generated training data
- Synthetic data
- Reinforcement-learning environments
- Domain-specific datasets
- Data labeling
- AI evaluation
- Data infrastructure
TechCrunch reports that companies such as Mercor, Handshake, and Micro1 have also experienced rapid revenue growth in the AI data market. The market is being repriced as the value of high-quality training data becomes clearer to investors.
The distinction that matters: Data labor companies sell hours of human expertise. Data infrastructure companies sell repeatable products that become part of the AI development workflow. The latter may command a different strategic position — and different multiples — because their revenue is more durable and their products harder to replace.
Capital Flow Watch: Where the Money Is Moving
This edition's framework organizes the current funding cycle around five capital layers. September 24 activity spans all five.
| Layer | What It Covers | Today's Example |
|---|---|---|
| 1. AI Infrastructure | Compute, networking, data centers, inference, specialized hardware | Bessemer growth fund ($4B) |
| 2. AI Data | Training data, synthetic data, evaluation, RL environments | Snorkel AI ($350M) |
| 3. AI Platforms | Foundation models, agent infrastructure, developer tools | Bessemer seed fund ($1.75B) |
| 4. Enterprise AI | AI employees, automation, workflow platforms | Ema ($77M) |
| 5. Applied AI | Healthcare, biotech, robotics, finance, cybersecurity, industrial | Basecamp Research ($140M), Vantora ($100M) |
The Valuation Signal
Snorkel AI's move from a $1.3 billion valuation to $3.5 billion in roughly 17 months is the clearest valuation example of the week. That is not merely a company marking itself up — it is a repricing of an entire category.
The right question is not whether Snorkel is worth $3.5 billion in an absolute sense. It is: what has changed in investor expectations about the value of AI training infrastructure?
Three things appear to have shifted:
- Training data is no longer commoditized. Frontier labs are paying premiums for proprietary, expert-curated data — because the supply of high-quality public text is finite.
- Revenue durability is being rewarded. Snorkel's $375 million run-rate, if it holds, positions the company closer to a SaaS or infrastructure business than a services firm.
- Strategic position matters more than growth rate alone. Companies that sit at a bottleneck in the AI development pipeline attract capital at premium multiples — even if their absolute revenue is smaller than application-layer peers.
The valuation signal this week is not that AI valuations are rising across the board. It is that the market is willing to pay premiums for companies positioned at genuine bottlenecks — and less willing to pay premiums for companies that can be commoditized by the next model release.
The Venture Capital Strategy Shift
The September funding environment points toward five shifts worth tracking:
1. From applications to infrastructure. Investors are funding the layers underneath AI applications — data, compute, evaluation, and identity. The application layer is still fundable, but it is no longer where the largest checks go.
2. From SaaS to AI-native economics. AI companies can have substantially different infrastructure and compute costs than traditional software. That changes the metrics investors use to evaluate margin profiles and capital efficiency — and rewards companies that can control their compute and data costs.
3. From software to physical systems. Robotics, industrial AI, and scientific applications are attracting larger rounds. Vantora's physical AI raise and Basecamp Research's biotech round are the clearest signals — AI is no longer a purely software story.
4. From products to ecosystems. AI startups increasingly depend on relationships with hyperscalers, chip companies, data providers, and enterprise customers. Deals that combine capital with strategic partnerships — like Basecamp's round with Nvidia — are becoming more common.
5. From company funding to capital formation. Large venture firms are themselves raising increasingly large pools of capital dedicated to AI. Bessemer's $5.75 billion raise is the clearest example of this cycle — capital formation is now as important a signal as individual company rounds.
What to Watch Next
- AI infrastructure funding. Watch whether capital continues moving into compute, networking, storage, and data centers. Bessemer's $4 billion growth fund is a signal the market expects this trend to continue.
- Training-data companies. Monitor valuations and revenue growth in the AI data economy. Expect more Series E and growth rounds at premium multiples from companies positioned at the data layer.
- AI agent infrastructure. Look for funding behind agent reliability, orchestration, identity, and observability. This is the layer where safety and governance frameworks will be built.
- Physical AI. Track robotics and industrial AI rounds. Vantora's model may be replicated — expect more hybrid startup-builder and corporate partnership vehicles.
- AI biotech. Watch whether AI-native drug discovery companies continue attracting large Series A and B rounds. Basecamp Research's $800 million valuation sets a benchmark for the category.
- Late-stage AI valuations. Monitor whether private-market valuations continue expanding faster than underlying revenue — a signal that could indicate a repricing risk building in the market.
- VC fund formation. Track whether additional major venture firms raise dedicated AI funds. Bessemer's $5.75 billion is a benchmark — expect comparable raises from competitors in the coming quarters.
The CODEW Capital Flow Framework
Five questions for every funding cycle:
- Where is the capital coming from? VC firms, corporate investors, sovereign capital, strategic investors, and growth funds.
- Where is it going? Infrastructure, data, platforms, applications, and emerging physical systems.
- What layer is being funded? The application, platform, infrastructure, or underlying technology.
- What does the funding buy? Compute, talent, distribution, acquisitions, product development, or market expansion.
- What does it signal? The technology categories investors believe can support the next phase of enterprise and AI growth.
Venture capital is moving deeper into the technology stack — and the largest opportunities are no longer limited to companies building applications on top of AI models. Investors are funding the data, infrastructure, platforms, enterprise systems, and scientific technologies required to make AI economically useful.
For founders: If you are building an AI company, the funding question is no longer "Is this AI?" It is: Which layer of the AI stack does this company own, and why is that layer defensible? Companies at genuine bottlenecks — training data, agent reliability, enterprise workflow execution, physical AI deployment — will attract capital at premium multiples. Companies competing on the application layer alone will face harder diligence.
For investors: Fund formation is now as important a signal as individual company rounds. Bessemer's $5.75 billion across two AI-focused funds — with an explicit multi-layer strategy — shows that leading firms are positioning for a decade-long AI capital cycle, not a single wave. Investors who can underwrite across infrastructure, data, platforms, and applications will have structural advantages over those confined to a single layer.
For enterprise buyers: The funding landscape is a preview of the vendors that will be offering enterprise AI capability over the next 24 months. Companies attracting large rounds today — Ema in enterprise AI, Snorkel in data infrastructure, Basecamp in biotech — are the incumbents of the next enterprise software cycle. Procurement teams should be evaluating these categories now, not when they become mandatory.
Sources
Data sourced from Techmeme, TechCrunch, The Economic Times, Atom Startup Funding Tracker, and other primary reporting covering funding rounds, fund formation, valuations, and investor activity from September 22–24, 2026. Rounds reported but not officially closed are labeled accordingly. Company funding and VC fund formation are presented separately.
All factual claims regarding funding amounts, valuations, and deal terms are drawn from contemporaneous reporting and company disclosures. Editorial analysis is clearly distinguished from reported facts throughout.
The CODEW Stat
$5.75 billion — Bessemer Venture Partners' new AI-dedicated fund pool, split between $1.75 billion for seed and early-stage and $4 billion for growth-stage. It is the largest single AI fund formation of the current cycle and signals that the institutional capital cycle around AI is still expanding, not contracting. Add Snorkel AI's $350 million, Basecamp Research's $140 million, and Enveda's $311 million, and the week's AI-related capital formation exceeds $6.5 billion.
Reviewed by Erwin Castro
on
Thursday, September 24, 2026
Rating:
