Deep tech now accounts for 36% of all global venture capital funding — up nearly threefold from 2016, when the category stood at barely 12%.
The shift is structural: AI infrastructure, defense tech, and bioconvergence are drawing capital that once went to pure software.
For private-market investors, the question is not whether to allocate to deep tech, but how to build the diligence capacity to deploy into it.
The headline number comes from Celesta Capital's analysis presented at Deep Tech Week NYC in March 2026. The underlying data — drawn from Dealroom, BCG, PitchBook, and the Bureau of Economic Analysis — tells a story that extends well beyond the AI boom that accelerated it. U.S. private investment in computing equipment alone has roughly doubled since 2023, reaching $280 billion annually in 2025. As StartUs Insights confirmed in its Deep Tech Market Report published July 2026, the shift is not a spike — it is a reconfiguration of how venture capital allocates across the technology stack.
As we wrote in June, global startup funding hit a record in Q1 2026, driven by AI mega-rounds that reshaped the venture landscape. The current data suggests that trend is accelerating, not peaking.
Deep tech share of venture capital
From 12% in 2016 to 36% in 2025 — a structural reallocation of venture dollars toward hardware, infrastructure, and science-based startups · Celesta Capital / BCG, Apr 2026
What is rising
Three categories are driving the increase. AI infrastructure — chips, data centers, energy — drew the largest share. U.S. compute equipment investment hit $280 billion annually by 2025, a figure that has no precedent in the hardware investment cycle since the dot-com era. AI-related deals made up 61% of all VC investments in 2025, with total AI venture capital surging past $200 billion annually. The concentration is visible at the sector level: Dealroom tracks AI agents as the hottest category with 320 startups and $14 billion in funding, up 182%, followed by humanoid robotics at 84 startups and $6.2 billion, up 140%.
Defense tech raised a record year in 2025, led by rounds like Anduril's $2.5 billion Series F at a $30.5 billion valuation, and the pace accelerated into 2026 with fusion, space, and biotech defense applications pulling in additional capital. Defense tech capital grew from $1.7 billion across 91 deals in 2015 to over $12 billion in 2025, with 755 dual-purpose companies and 284 pure defense tech startups receiving funding. Bioconvergence — the intersection of biology with hardware and AI — grew from a niche thesis a decade ago into a distinct category with dedicated funds and multiple unicorns. Synthetic biology companies using AI-designed proteins raised over $2 billion in 2025 alone, per Dealroom.
The concentration is visible at the fund level. SOSV, the Princeton-affiliated deep-tech investor with $1.5 billion in AUM, has 8 unicorns and 2,500 portfolio companies. Celesta Capital, founded in 2013, has deployed over $1.1 billion across 100+ investments spanning semiconductors, intelligent systems, and bio-convergence. The number of active hard-science venture capital firms globally now exceeds 120, up from roughly 40 a decade ago, per Waveup's 2026 guide. Lux Capital leads the cohort with 20+ years and 66 IPOs across AI, biotech, neuroscience, and frontier sciences.
What is declining
Deal count, not total dollars. Global venture investment hit $469 billion in 2025 — a 47% increase year-over-year — but the number of deals fell 17% in the same period. Capital is concentrating in larger, more selective bets. The average deep tech deal size nearly tripled from $50 million in 2022 to $135 million in 2025, reflecting bigger and more concentrated bets on fewer companies. Software-only SaaS startups, which dominated the 2020–2021 cycle, are seeing their share of VC shrink as capital rotates toward companies with physical products, regulated markets, and longer commercialization timelines.
The compositional shift is straightforward: the technology itself is becoming more capital-intensive at the infrastructure layer, and venture capital is following the same gravity. Software venture is not in decline — its share is shrinking because the denominator is growing faster.
What is emerging
Two structural developments deserve attention beyond the headline concentration. A third is the rise of corporate venture capital models adapted to deep tech. The Venture Capital as a Service (VCaaS) model, exemplified by Pegasus Tech Ventures' partnership with automotive supplier AISIN, allows industrial corporations to access deep tech deal flow without building internal venture teams. Harvard Business School research confirms that VC funding in climate tech startups causes established corporations to measurably increase their own climate investments across capex and R&D — exposure to startup portfolios transforms corporate behavior at the balance-sheet level.
University spinouts are claiming a record share of European VC investment as investors look to academic labs for the next generation of deep tech companies, per PitchBook's May 2026 analysis. The AUTM 2024 Licensing Survey shows equity exits from spinouts up 25%, while traditional royalty income declined 40% — the market is betting on early-stage tech transfer through equity rather than licensing. The data tracks 272 venture-backed university spinouts collectively representing $15.36 billion in funding, with the University of Washington, MIT, and ETH Zurich as the most active breeding grounds. Applied Intuition, a Stanford-connected autonomous vehicle software company, raised $600 million in a Series F round at a $15 billion valuation. Kailera Therapeutics, a Harvard-connected biotech, raised $600 million in Series B for obesity treatments. These are not academic exercises — they are institutional-grade venture outcomes emerging from the university pipeline.
Family offices now represent approximately 31% of all startup funding, up from an estimated 20% five years ago. With family office assets projected to reach $5.4 trillion by 2030, according to Deloitte, this capital base is structurally longer-dated than institutional VC and better suited to the 7–10 year timelines that deep tech requires. The UBS Global Family Office Report confirms that alternatives now represent more than half of U.S. family office portfolios, with private equity and venture capital as the primary allocation channels. The shift has implications for fund structure: as more family offices move from LP positions into direct co-investment and solo GP vehicles, the capital formation model for deep tech is becoming less dependent on institutional gatekeepers.
| Sector | Share of deep tech VC | 5-year trend |
|---|---|---|
| AI / digital platforms | ✔ ~33% | Growing — infrastructure-led |
| Energy / climate | ✔ ~24% | Stable — policy-supported |
| Defense / dual-use | ✔ ~12% | Fast-growing — VC record in 2025 |
| Biotech / health | ◐ ~10% | Cyclical — AI-driven rebound |
| Space | ◐ ~5% | Growing — commercial station era |
| Other | ✗ ~16% | Fragmented |
University spinout exit values — the AUTM data shows a 25% increase in equity cash-outs; if this accelerates, it validates the spinout-as-vehicle model.
Family office direct allocation to deep tech — a shift from fund-of-funds to direct co-investment changes the capital formation dynamic.
Defense tech dual-use ratio — the share of defense tech companies with commercial revenue determines whether the category sustains or corrects.
AI infrastructure ROI — the $5.2 trillion in projected data center investment by 2030 requires real revenue from inference workloads, not just training.
What it means for private-market investors
The 36% figure is not a ceiling. If deep tech continues to grow at its current trajectory relative to overall VC — and the structural drivers (AI infrastructure, defense modernization, bioconvergence) show no sign of reversing — the category could approach 50% of global venture funding before 2030. The McKinsey estimate of $5.2 trillion in data center investment alone suggests the capital intensity floor is rising.
The deeper risk is a shortage of diligence capacity. Deep tech companies have longer development cycles, higher technical risk, and less transparent comparables than software startups. The funds that build the technical evaluation infrastructure to assess semiconductor tape-outs, FDA trial designs, and satellite manufacturing timelines will have a structural advantage over those applying software-era heuristics to hardware-era opportunities.
For Nexithon's readership, the practical takeaway is to track the composition of deep tech flows by subsector and stage, not just the aggregate. The 36% headline masks wide variance: AI infrastructure captures the bulk, but the highest risk-adjusted entry points may lie in the smaller, less crowded, and under-allocated categories — industrial biotech, orbital manufacturing, and autonomy-enabling hardware — where capital is still scarce relative to the addressable market. The returns, however, will follow those who build the diligence infrastructure to evaluate these opportunities before they become consensus allocations.
