AI infrastructure funding round: deal sizes and market data
How large are AI infrastructure funding rounds? 2025 OECD data on the 109.3bn USD segment, mega-deal concentration and the mean-versus-median gap.
In this article6

Global venture capital deployment for infrastructure
Global venture capital deployment into artificial intelligence infrastructure and hosting reached USD 109.3 billion in 2025, according to analysis published by the Organisation for Economic Co-operation and Development (OECD) in February 2026[1][1]. Total venture capital invested across all AI sectors reached USD 258.7 billion in 2025, representing 61% of total global venture capital investment (USD 427.1 billion)[1]. That share is double the 30% recorded in 2022[1]. Cumulative venture capital allocated to AI infrastructure and hosting companies between 2012 and 2025 totalled USD 256.1 billion, reflecting the heavy capital intensity of hardware acquisition, data center colocation, and model hosting networks[1]. The underlying series is Preqin data held by the OECD.AI Policy Observatory, last updated 1 January 2026 and accessed 9 January 2026[1].
Geographic distribution demonstrates significant concentration in capital origin and destination. Venture capital investors based in the United States provided 56% (USD 124 billion) of outgoing global AI venture capital in 2025, while US-based companies attracted 75% (USD 194 billion) of total deal value[1]. Investors in the European Union (EU27) supplied 7% (USD 14.5 billion) of outgoing capital, with EU27 companies attracting 6% (USD 15.8 billion) of global deal value[1]. United Kingdom investors generated 9% (USD 20.7 billion) of outgoing venture investment, while UK companies received 5% (USD 13.8 billion)[1]. Chinese investors accounted for 8% (USD 17.2 billion) of outgoing capital, with domestic companies securing 5% (USD 13.9 billion)[1].
| Region | Outgoing VC Share (%) | Outgoing VC Value (USD) | Attracted VC Share (%) | Attracted VC Value (USD) |
|---|---|---|---|---|
| United States | 56% | $124.0B | 75% | $194.0B |
| United Kingdom | 9% | $20.7B | 5% | $13.8B |
| European Union (EU27) | 7% | $14.5B | 6% | $15.8B |
| China | 8% | $17.2B | 5% | $13.9B |
Capital concentration toward mega-rounds has intensified alongside overall sector expansion. Transactions exceeding USD 100 million represented about 73% of total AI venture capital investment value in 2025[1]. AI firms have attracted a declining share of early-stage venture capital relative to all funding rounds since 2023, as capital concentrates in mega deals. That is a value effect rather than a volume one: early-stage deals still made up just over 75% of AI deal count in 2025, a historic high[1].
Deal sizes: the gap between the mean and the median
Headline totals hide how uneven individual rounds are. The mean AI venture capital deal grew from about USD 11.2 million in 2014 to USD 35.8 million in 2025, while the median deal in 2025 stayed at just USD 5 million, a spread that shows a small number of very large transactions pulling the average up.[1] The stage split is starker: early-stage deals averaged around USD 11.8 million in 2025 against USD 131 million for later-stage rounds, and the five largest deals of the year together accounted for nearly USD 63 billion, about a quarter of the total sum of AI venture capital investments.[1] Deals above USD 1 billion alone represented almost half of total 2025 value.[1] Dealroom's separate tracker is not directly comparable to the OECD's Preqin series: it puts 2025 AI funding at USD 215.9 billion, against the OECD's USD 258.7 billion. On that basis it recorded USD 399.0 billion raised by AI startups in the first six months of 2026, of which USD 142.3 billion fell in a closed second quarter.[2]
Inference platform and hardware capital events
Private market deployment into dedicated inference platforms has accelerated as enterprise workloads shift from model training to production inference execution. Baseten raised USD 1.5 billion in Series F funding in June 2026, its fourth fundraise in 18 months.[3] The round was led by Altimeter Capital, Conviction and Spark Capital and co-led by Sands Capital and Wellington Management.[3] It followed the company's USD 300 million Series E, announced in February 2026.[4] Round sizes at that cadence show how quickly inference-serving specialists are having to refinance to keep pace with compute commitments.
- Capital allocation directed toward reserving dedicated GPU clusters and high-density colocation space.
- Enterprise transition from general-purpose cloud instances to specialized, low-latency model serving frameworks.
- Increasing demand for customized inference orchestration layers across the European provider ecosystem.
- Financial expansion driven by real-time API serving requirements and enterprise SLA commitments.
Inference infrastructure capital requirements differ from training clusters due to ongoing operational costs, continuous power drawn by inference nodes, and strict geographic latency constraints. Specialized hosting platforms require substantial balance sheets to secure long-term capacity agreements with data center operators and hardware vendors. What those capacity agreements cost is set by the accelerator price curve underneath them, which moves on its own cycle.
Quantum computing and next-generation architectures
Venture capital is increasingly directed toward quantum computing architectures as a parallel compute track for specialized workloads. PsiQuantum closed a USD 1 billion Series E round in September 2025 led by funds and accounts managed by affiliates of BlackRock, along with Temasek and Baillie Gifford, with new participation from Nvidia's venture arm NVentures[5]. The company said the funding would let it break ground on utility-scale quantum computing sites in Brisbane and Chicago and deploy large-scale prototype systems to validate its systems architecture and integration[5].
- USD 125 million performance-based DARPA award under the Quantum Benchmarking Initiative, PsiQuantum's most valuable US government contract to date, funding verification of its utility-scale photonic architecture[6].
- The award builds on a USD 31.8 million Stage C tasking from September 2025 and broadens DARPA's hands-on evaluation framework; separately, PsiQuantum holds an anchor role, backed by state-level capital commitments, at the Illinois Quantum and Microelectronics Park in Chicago[6].
- A May 2026 letter of intent with the US Department of Commerce for USD 100 million in proposed direct funding under the CHIPS and Science Act[6].
- A system-integration milestone at the Milpitas facility, where three cabinets holding roughly 250 chips each have been networked together[7].
- Construction of a Chicago facility intended to host hardware for a million-qubit scale, fault-tolerant quantum computer.[7]
Institutional backing of photonic quantum hardware and next-generation accelerator silicon reflects a capital allocation strategy aimed at mitigating scaling limits associated with traditional silicon CMOS processors, providing specialized processing capacity alongside standard GPU infrastructure.
Frequently asked questions
What was the total venture capital investment in AI infrastructure in 2025?
Venture capital investments in artificial intelligence firms focused on IT infrastructure and hosting reached 109.3 billion USD in 2025. This segment attracted the largest volumes of overall AI sector funding as compute and physical capacity requirements surged globally.
How big is a typical AI funding round?
There is no typical round. The mean AI venture capital deal reached 35.8 million USD in 2025 while the median stayed at 5 million USD, because a small number of very large transactions pull the average up. Early-stage deals averaged about 11.8 million USD in 2025, against 131 million USD for later-stage rounds.
How concentrated are AI infrastructure funding rounds?
Deals above 100 million USD accounted for about 73% of total AI venture capital investment value in 2025, and deals above 1 billion USD alone made up almost half of total value. The five largest deals of the year together came to nearly 63 billion USD, about a quarter of all AI venture capital invested.
Are quantum computing firms attracting AI infrastructure capital?
Yes. Quantum computing is drawing parallel capital as a next-generation compute infrastructure layer. PsiQuantum closed a 1 billion USD Series E in September 2025 to break ground on utility-scale sites in Brisbane and Chicago, and separately holds a 125 million USD DARPA award funding verification of its architecture.
Which regions capture AI funding?
Companies in the United States attracted about 75% (194 billion USD) of global AI venture capital deal value in 2025. EU27 companies took 6% (15.8 billion USD), China 5% (13.9 billion USD) and the United Kingdom 5% (13.8 billion USD).
Does AI infrastructure funding only cover compute?
No. The IT infrastructure and hosting category behind the 109.3 billion USD figure uses a broad industry classification that also counts AI model developers such as Anthropic, Mistral and X.AI, not only compute and hosting providers. Read the segment total as an indicative measure of AI compute investment rather than an exact one.
Sources
- oecd.org
- Artificial Intelligence, Dealroom Deep Dive dealroom.co
- Announcing our Series F baseten.co
- Announcing Baseten's $300M Series E baseten.co
- PsiQuantum Raises $1 Billion to Build Million-Qubit Scale, Fault-Tolerant Quantum Computers — PsiQuantum psiquantum.com
- PsiQuantum Secures $125 Million Expanded Agreement with DARPA under QBI Program quantumcomputingreport.com
- PsiQuantum has a plan to make a massive quantum computer out of light technologyreview.com
Rankings and scores on Optinest cannot be bought — no commercial relationship changes how a provider is assessed. How we score
Related reading
All Funding →A third of EU adults now use generative AI
Analyze generative AI usage among adults with 2025 statistics from the EU and OECD. Discover market sizes, demographic trends, and everyday use cases.

AI chip startups recent funding
AI chip startups raised USD 8.4 billion in Q1, over USD 6 billion in Q2 and USD 8.5 billion in Q3 2026. The largest rounds with stage and lead.

AI chip market share: who is counted and how it is measured
AI chip market share explained: which vendors are counted, revenue versus unit shipments, and which published figures are measurements versus forecasts.
