GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 -2.3%
GPT-6A 820 —
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 585 -0.7%
INKL 531 —
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 -2.3%
GPT-6A 820 —
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 585 -0.7%
INKL 531 —
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
← Back to feed

AI Absorbed $222B and 66% of US Venture Capital in 2025 — Up From 10% a Decade Ago

Two in three venture dollars deployed in the United States last year went to AI. That is not an estimate — it is the figure PitchBook calculated across its complete US VC dataset for 2025: $222 billion of $339 billion in total deal value, representing 65.6% of the market.

The trajectory is unambiguous. In 2015, AI was a niche allocation at roughly 10% of US VC. By 2024 it had reached 47.2%. The jump to 65.6% in 2025 means AI absorbed nearly all of the growth in the venture market last year while non-AI deal value stagnated.

Concentration at the Top

The structural story inside that $222B is one of extreme concentration. OpenAI’s $40 billion round in Q1 2025 is among the largest single private funding rounds in PitchBook’s dataset. Andreessen Horowitz’s $15 billion fund close in January 2026 — partly deployed in Q4 2025 commitments — anchors the LP-side picture. Foundation model companies and the infrastructure layer beneath them absorbed a disproportionate share of that capital.

US AI unicorns now represent $4.3 trillion in aggregate valuation. That figure includes OpenAI, xAI, Anthropic, SpaceX (valued partly on Starlink AI infrastructure contracts), and a growing cluster of infrastructure companies — CoreWeave, Baseten, Groq — that have crossed the $10B threshold by providing compute and inference capacity rather than models themselves.

What Happened to Non-AI Venture

The $339B total for 2025 is a record year for US VC in absolute terms. But the non-AI portion — roughly $117B — is not growing at the pace the headline suggests. Traditional SaaS, fintech, and biotech saw deal counts contract even as AI valuations pulled up the aggregate.

Investors have responded to the concentration by either going deeper into AI or repositioning portfolio companies around it. Companies describing themselves as AI-adjacent received materially higher valuations in 2025 than comparable companies that did not. The market is pricing a winner-takes-most assumption into every AI infrastructure bet.

The Debt Layer

The PitchBook data covers equity venture capital. The credit and private debt layer running beneath it is larger and accelerating faster. Apollo and Blackstone have each disclosed multi-billion-dollar AI infrastructure credit facilities in 2026. CoreWeave closed $8.5 billion in investment-grade GPU loans. CME Group launched GPU compute futures this year. AI infrastructure is now an asset class with its own debt markets and derivative products — a development that has no precedent in prior technology cycles.

Goldman Sachs estimated the AI spending cycle at $5 trillion over five years, with nearly half financed by debt. The equity concentration PitchBook documents is the visible part. The leverage underneath it is not captured in VC statistics.

The Competitive Geometry

The $222B figure also reveals the US-China dynamic in precise capital terms. Chinese AI investment, while substantial domestically, does not appear in the US VC dataset. The ATOM report published this year showed China surpassed the US in open-weight model downloads for the first time — 1.15B vs 723M — achieved with a fraction of the equity capital. DeepSeek raised its first external capital at $20B and operates with cost structures that make US foundation model economics look fragile.

Ten years into the AI investment cycle, venture capital has not distributed across the technology economy. It has concentrated into the frontier of intelligence itself — the models, the compute to run them, and the infrastructure to move the resulting tokens at scale. The 2025 data is the clearest quantification of that compression yet.