In 2025, venture capital (VC) investments in AI firms globally made up over half (61%, USD 258.7 billion) of all VC investment (USD 427.1 billion), doubling its 2022 share (30%).
For generative AI firms specifically, VC funding surged from about 2% (USD 2.8 billion) to 12% (USD 15.3 billion) of total AI VC investments between 2022 and 2023. Since then, VC investments in generative AI firms grew further, reaching USD 35.3 billion in 2025, or about 14% of all AI VC investments.
Firms in the United States attract the largest share of VC by a wide margin, comprising approximately 75% (USD 194 billion) of global AI VC deal value, followed by the EU27 (6%, USD 15.8 billion), the People’s Republic of China (hereafter ‘China’) (5%, USD 13.9 billion), and the United Kingdom (5%, USD 13.8 billion). United States VC investors also are the most active, representing about 56% (USD 124 billion) of the worldwide value of outgoing VC investments in AI in 2025, followed by investors in the United Kingdom at 9% (USD 20.7 billion), China at 8% (USD 17.2 billion) and EU27 investors at 7% (USD 14.5 billion).
Since 2023, AI firms have attracted a declining share of early-stage VC relative to all funding rounds, possibly because capital is concentrating in “mega deals” of over USD 100 million. Mega deals have continued to rise, comprising about 73% of total AI investment value in 2025.
Since 2023, AI firms working on IT infrastructure and hosting attracted the most VC investment, overtaking other industries to reach a total of USD 47.4 billion in 2024 and USD 109.3 billion in 2025, more than two-thirds as much as all other industries combined (USD 149.4 billion). Between 2012 and 2025, this comes to a cumulative USD 256.1 billion of investment, reflecting a rush to build AI compute infrastructure critical to scaling advanced AI systems.
While long-term prospects for AI remain strong, investment markets are cyclical. These findings regarding past trends in AI VC should therefore be interpreted with caution in seeking to anticipate future trends.
Venture capital investments in artificial intelligence through 2025
Venture capital (VC) has been central to accelerating recent advances in artificial intelligence (AI). By supplying the substantial upfront financing required to develop, train and deploy advanced AI systems, VC has enabled both rapid scale-up and commercialisation. While VC did not create AI’s scientific foundations, developed over decades of government-funded research, its investments have drastically accelerated AI’s diffusion. Many of today’s most transformative AI-based products and services, especially in generative AI since late 2022, grew through critical stages of venture backing.
VC functions within a broader investment ecosystem of domestic, international, public and private actors that finance AI start-ups and firms worldwide (see Figure 1). This brief leverages data hosted on the OECD.AI Policy Observatory from Preqin, a capital-markets analysis firm in the United Kingdom (OECD, 2026[1]). Preqin defines VC as “provid[ing] capital to new or growing businesses with perceived long-term growth potential”. In short, VC focuses on non-public companies, particularly those with innovation-based business models and high growth potential, such as AI start-ups and firms.
National investment ecosystems differ. Investment ecosystems in different countries will have varied mixes of investment instruments (e.g. equity, debt) and investor types (e.g. angel investors, VC, private equity, pension funds, sovereign wealth funds, corporate investors including internal investment and corporate VC) depending on the national context and financial market structure. Investment partnerships often blend these sources. Foreign Direct Investment (FDI), including cross-border mergers and acquisitions (M&A), also shapes the landscape. While FDI brings important sources of capital and expertise, it can raise policy concerns about maintaining beneficial ownership and control of home-grown AI firms.
This policy brief focuses on AI VC investment (including corporate VC) trends through 2025. Analysis of other investment areas and cross-referencing datasets would further contribute to understanding the AI investment ecosystem, and the comparative role of VC, and could be the topic of future analysis. For example, recent research, using System of National Accounts data, shows that AI-related capital formation within EU27 countries could total as much as EUR 294 billion in 2023 (OECD, 2025[2]). This sum, while based on an experimental methodology, is about 2.4 times higher than the 2023 total AI-related VC presented in this report, highlighting the important role of internal corporate and public sector investments in the AI ecosystem.
This policy brief focuses on venture capital investment (including corporate venture capital) trends through 2025
Note: This figure is illustrative and non-exhaustive. VC is a form of investment that can be undertaken by many different actors. As such, some investors fall into multiple categories e.g. VC is often considered a subset of private equity; sovereign wealth funds are also a form of public investment; accelerators and incubators can also be part of corporate VC arms, etc.
The OECD, including through its OECD.AI Policy Observatory, has provided analysis of global AI venture capital activity since 2019 using data from Preqin (OECD, 2019[3]; Tricot, 2021[4]). This policy brief updates prior analyses and examines VC investments between 2012 and 2025 in more than 33 000 AI firms worldwide, covering nearly 85 000 transactions. It provides updated analysis of the countries, sectors and technology verticals of VC investments in AI firms.
Estimates leverage the most recently available data, in this case 2025, on the OECD.AI Policy Observatory, to show timely trends about where, how and at what rate AI is being developed and used across sectors. Firms are identified as AI firms based on Preqin’s cross-industry and vertical categorisation, as well as on the OECD’s automated analysis of the keywords contained in the description of the activities of the firms captured in the Preqin dataset (OECD.AI Policy Observatory, n.d.[5]).
The scope of this policy brief is limited to VC investments in AI firms; caution should be taken when drawing broader conclusions from this data as it offers only one vantage point relative to other forms of investment such as internal investments in AI made by public companies or by governments. Historical revisions and methodological improvements to the Preqin dataset can also impact results as deals, particularly smaller deals, are often added retroactively. This brief captures the available data at the time of publication and thus over time may differ from the live data on the OECD.AI Policy Observatory or from other OECD analysis. Data is nominal and not adjusted for inflation. Dollar values are rounded to one decimal place and percentages to the nearest whole number; as a result, percentage totals may not sum to 100.
It is important to note that investment markets are cyclical and past trends are not always accurately indicative of future performance. AI has already seen sizeable fluctuations in attention and funding through waves of breakthroughs (OECD, 2019[3]). While long-term signals point to major advances in AI, short- to medium-term investment outcomes remain uncertain. Accordingly, the findings in this brief should be interpreted with caution with respect to future outcomes and do not represent financial advice.
Global VC investment in AI firms is surging. This reflects the growing maturity of AI-enabled products and services and investor confidence in its transformative potential. Recent data on VC investments in AI firms show continued momentum, as reflected by the significant value and number of investments in AI firms, particularly in the United States. The global annual value of VC investments in AI firms rose dramatically, from about USD 8.3 billion in 2012 to USD 258.7 billion in 2025 (Figure 2). By 2025, AI accounted for about 61% of the total value of all VC investment globally, double its 2022 share of 30%, underscoring AI’s growing weight in global VC activity.
Overall global VC investment peaked in 2021 at over USD 800 billion, reflecting investments linked to the COVID-19 pandemic in areas including remote work and communications and telemedicine for example, and have since declined in 2025 to stabilise at roughly pre-pandemic levels. VC investment into AI firms has increasingly become the driver of total VC flows: In 2025, total VC investment (including AI VC investment) totalled roughly USD 427.1 billion annually, representing an approximately 47% decrease since 2021. At the same time, while VC investment in AI firms experienced a decrease from its peak in 2021 (USD 257.3 billion) to 2023 (USD 123.6 billion), annual VC investment in AI firms rose again between 2023 and 2025 by about 109%, recovering back to 2021 levels of 258.7 billion. Between 2022 and 2023, VC funding for generative AI surged from USD 2.8 billion to USD 15.3 billion, an increase from about 2% of total AI VC investments to 12%. This corresponds to the release of ChatGPT in late 2022 and the subsequent wave of generative AI tools. By 2025, VC investments in generative AI firms totalled USD 35.3 billion, now representing over 14% of AI VC investment.
Sum of venture capital investment overall and for AI-related venture capital investment (2012-25, USD billions)
Note: Please see the methodological note for more information.
Source: OECD.AI (2026), data from Preqin, last updated 2026-01-01, accessed on 2026-01-09, https://oecd.ai/
AI firms generally follow the VC lifecycle, from seed and early-stage rounds to late-stage funding and eventual exit, such as through an initial public offering (IPO), M&A or bankruptcy (Table 1). Each stage reflects different levels of technology readiness, investor risk and company maturity. Seed rounds typically fund proof-of-concept and AI model development; early-stage rounds usually support productization, scaling and customer acquisition; while late-stage rounds support expansion into new markets, company monetisation and preparation for eventual exit.
Analysing funding stages and exits can help inform policymakers on the strengths and weaknesses of domestic AI ecosystems, pointing to gaps where targeted government support may be warranted to help AI firms grow past certain stages. Examining exit trends can be particularly helpful to policymakers as it indicates whether firms remain private, list publicly through an IPO, merge or are acquired (often by large technology firms) or face bankruptcy. It should be noted that M&A activity can also reflect strategic positioning rather than technology quality or commercial success. For example, larger technology firms can acquire smaller players, their key employees, or their intellectual property (IP) as part of a defensive posture against competition.
Note: Examples are an illustrative selection. Funding round definitions vary across investor communities and can overlap (e.g. angel investment and seed funding). Some funding rounds for AI firms are not publicly disclosed. This analysis clusters funding rounds as defined by Preqin.
Source: OECD analysis using sources indicated in-text and verified with Preqin.
Preqin defines early-stage VC as capital directed toward start-ups and seed funding. Over the last decade, early-stage investment has consistently accounted for roughly a quarter of the total value of AI VC deals (Figure 3). This share experienced a trough in 2021, likely due to investors favouring more established companies amid market uncertainty caused by the COVID-19 pandemic. While the share of early-stage deals in total AI investment value has remained relatively stable, their share of deal count has grown over the last five years, reaching a historic high of just over 75% in 2025 (Figure 4). This trend reflects the increasing prevalence of “mega deals” in later-stage funding, which skews overall investment value. In 2025, early-stage VC deals average around USD 11.8 million, compared to USD 131 million for later stage deals. Consequently, early-stage VC represented only 14% of all AI-related mega deals in 2025 (among deals with a specified stage).
Share of value of worldwide AI venture capital deals by stage in USD
Note: Deals with an unspecified stage are dropped and assumed to have a similar proportion by stage.
Source: OECD.AI (2026), data from Preqin, last updated 2026-01-01, accessed on 2026-01-09, https://oecd.ai/
Share of number of worldwide AI venture capital deals by stage
Note: Deals with an unspecified stage are dropped and assumed to have a similar proportion by stage.
Source: OECD.AI (2026), data from Preqin, last updated 2026-01-01, accessed on 2026-01-09, https://oecd.ai/
The dollar values of AI VC deals have continued to grow over time, with “mega deals” (deals over USD 100 million in value) becoming more frequent (Figure 5) and larger (Figure 6). From 2014 to 2025 the mean AI VC deal size increased from about USD 11.2 million to USD 35.8 million, though the median deal size in 2025 remained relatively modest at USD 5 million, indicating that a few very large deals are skewing the average. Indeed, in 2025, the top five mega deals collectively accounted for nearly USD 63 billion in funding, representing about a quarter of the total sum of investments. Mega deals themselves are also getting bigger, comprising around 73% of total investment value of AI VC deals in 2025. Of these mega deals, deals over USD 1 billion represent almost half the value of all VC investments in 2025 (Figure 6).
Share of number of worldwide AI venture capital deals by size in USD
Note: This figure includes only deals with reported values. Some data on the OECD.AI Policy Observatory fills in missing values using a median value calculation and thus will differ from the calculation used in this figure. Please see the methodological note for more information.
Source: OECD.AI (2026), data from Preqin, last updated 2026-01-01, accessed on 2026-01-09, https://oecd.ai/
Share of value of worldwide AI venture capital deals by size in USD
Note: This figure includes only deals with reported values. Some data on the OECD.AI Policy Observatory fills in missing values using a median value calculation and thus will differ from the calculation used in this figure. Please see the methodological note for more information.
Source: OECD.AI (2026), data from Preqin, last updated 2026-01-01, accessed on 2026-01-09, https://oecd.ai/
In several countries, AI now makes up a majority of total VC activity. In 2025, more than half of all VC investment in the United States, the United Kingdom, Israel and Canada flowed to AI firms in those jurisdictions (Figure 7). The flatter trend in the shares of AI VC investment in the People’s Republic of China (hereafter ‘China’) come largely due to increases in non-AI-related VC directed toward energy, raw materials and utilities as well as mobility and autonomous vehicles. The occurrence of mega deals can also quickly reshape national VC totals. For example, three large VC deals with United Kingdom AI firms totalling USD 6.7 billion made up about half of the country’s sum of AI venture capital inflows in 2025.
Percentage of AI in total venture capital investment by top jurisdiction, 2012-25
Note: Percentages are calculated as the sum of deals invested in AI divided by the total value of all venture capital deals in that country/region. Please see the methodological note for more information. The EU27 percentage is calculated as total AI VC investment across all EU countries divided by total VC investment in the EU27. It therefore represents an aggregate EU share rather than the average of country-level shares.
Source: OECD.AI (2026), data from Preqin, last updated 2026-01-13, accessed on 2026-01-13, https://oecd.ai/
Historically, VC investors from both the United States and China have been dominant actors in their respective domestic markets and internationally. In terms of “incoming” investment, in 2025 the United States and EU27 were net beneficiaries of AI VC investment, meaning that AI firms from these jurisdictions received more VC funds than those distributed by domestically based VC investors (as estimated by calculating total investment inflows minus total outflows). Of note, the United States is, by a wide margin, the largest recipient of international VC capital flows (Figure 10). The United States comprises about 75% (USD 194 billion) of all AI VC deal value (the sum of AI VC investments), followed by the EU27 (6%, USD 15.8 billion), China (5%, USD 13.9 billion) and the United Kingdom (5%, USD 13.8 billion) (Figure 8). In terms of number of deals, in 2025 the United States and China account for about 55% of the number of incoming deals worldwide (Figure 9).
The size of VC relative to the rest of the economy has generally stayed below 1% of Gross Domestic Product (GDP) for top jurisdictions. These shares have been gradually increasing since 2012, falling in 2022 following the previous years’ surge. As of 2024, Israel and United States had the highest VC investment per GDP, slightly less than 0.4%, followed by the United Kingdom and Canada (World Bank Group, 2024[18]). While GDP data are not available for 2025 at the time of writing, the sizeable increase in VC spending in 2025 will likely result in a rise in AI VC investment per GDP.
VC investments in AI firms by sum of investment, 2012-25 (USD millions)
Note: Please see the methodological note for more information. Top jurisdictions are shown along with the Rest of the world sum, where the latter is comprised of 79 separate jurisdictions.
Source: OECD.AI (2026), data from Preqin, last updated 2026-01-01, accessed on 2026-01-13, https://oecd.ai/
VC investment in AI firms by number of investments, 2012-2025 (number of deals)
Note: Please see the methodological note for more information. Top jurisdictions are shown along with the Rest of the world sum, where the latter is comprised of 79 separate jurisdictions.
Source: OECD.AI (2026), data from Preqin, last updated 2026-01-01, accessed on 2026-01-13, https://oecd.ai/
On the investor side (i.e. “outgoing” investment), VC investors in the United States were the most active investors in AI firms, representing about 56% (USD 124 billion) of the worldwide value of VC investments in AI in 2025 (with at least one identified investor country), followed by investors in the United Kingdom (9%, USD 20.7 billion), China (8%, USD 17.2 billion), and EU27 investors (7%, USD 14.5 billion). These distributions may reflect, in part, dynamics influenced by varying taxation regimes, as illustrated by the approximately 2% share attributed to investors from the Cayman Islands.
Flow of venture capital investments in AI firms from jurisdiction of investor to jurisdiction of AI firm, 2025
Note: The chart illustrates the approximate flow of investment in value from nationality of investor to industry and nationality of start-ups: how much VC investors from a country invest into start-ups of an industry per country of the start-up. Investor sums are calculated using fractional shares, such that if companies from more than one country invest, country-level investment is calculated as total investment divided by total number of source countries for each deal. Note that about USD 36 billion of AI VC investment does not have an identified investor country and is therefore excluded for these data. Caution is advised when comparing different versions of these data, as the AI-related concepts identified by the machine learning algorithm may evolve in time. Please see methodological note for more information.
Source: OECD.AI (2026), data from Preqin, last updated 2026-01-01, accessed on 2026-01-13, https://oecd.ai/
VC flows help reveal industry hotspots for AI. From 2012 to 2022, AI companies in mobility and autonomous vehicles attracted the most VC investments of all industries, drawing a total of USD 147.4 billion over the decade (Figure 10). Nearly all of these investments (94%) were made in companies in the United States and China. While VC investment in this area has softened in recent years, the earlier trend highlights investors’ belief that AI could address major mobility and transportation challenges.
Another notable wave of VC investment can be observed in AI companies focused on healthcare, drugs and biotechnology, prompted by the COVID-19 pandemic. Between 2019 and 2021, VC investments in this sector more than tripled from about USD 8 billion to USD 25 billion, before tapering off to about USD 12 billion in 2023 as economies largely halted pandemic related policies. As of 2025, VC investments into related AI firms in this category reached USD 20 billion.
Since about 2023, AI firms working on IT infrastructure and hosting have attracted the most VC investment, overtaking other industries to jump to USD 47.4 billion in 2024 (Figure 11). In 2025, investments in this industry sharply spiked to USD 109.3 billion, nearly as much as all other industries combined (USD 149.4 billion), representing over 42% of all AI VC investments. Investments in this industry represents a cumulative total of USD 256.1 billion of investment between 2012 and 2025. This category captures both compute infrastructure, from companies like Databricks, but also direct investments in model developers such as Anthropic, Mistral and X.AI. The increased investment in the industry corresponds to massively expanded interest in development and applications of generative AI models, as well as their compute infrastructure.
Venture capital investments in AI by top industry, 2012-25 (USD millions)
Note: AI firms operating in IT infrastructure and hosting offer an indicative measure of AI infrastructure (AI compute) firms as Preqin’s methodology for classifying firms by industry takes a broad definition of IT infrastructure and hosting, for example, including AI model developers. Please see the methodological note for more information.
Source: OECD.AI (2026), data from Preqin, last updated 2026-01-01, accessed on 2026-01-13, https://oecd.ai/
In addition to VC investment fuelling AI developments, AI technology is also transforming how VC investors and funds operate and make decisions (Oumeima and Zouari, 2024[19]). This creates a feedback loop where AI can help enhance the efficiency of VC actors, while VC funding in turn can accelerate the development and deployment of AI technologies. Policymakers can harness this feedback dynamic to incentivise and support VC investors in funding AI products and firms that are both innovative and trustworthy.
For example, AI can help automate due diligence processes, allowing VC investors to quickly analyse vast amounts of financial, market and other data. Natural language processing (NLP) systems can scan legal documents, pitch decks and news articles to highlight risks and opportunities. AI systems also hold promise for mitigating human biases in investment decisions, for example, with AI systems trained on representative datasets helping VCs to evaluate founders and teams more objectively (Chen et al., 2025[20]).
VC funding is a key driver of AI innovation, enabling the rapid development, scaling and global diffusion of AI technologies. However, current trends show global divides in AI VC activity, as the highest deal quantities and dollar values are concentrated in a handful of countries, namely driven by VC investors in the United States, followed by China. This highlights the need for further analysis on global divides in access to capital that can help to fuel AI ecosystems, including how to foster vibrant investment environments in emerging economies. Such investment environments can help support AI firms across levels of commercialisation, strengthening scale-up pathways to support competitive and sustained growth of AI firms globally.
Promote AI literacy for VC investors and investment literacy for policymakers: Policymakers can offer tailored AI literacy programs and strengthen investment literacy for policymakers to understand investor ecosystems and how to support the scaling of domestic AI firms across maturity levels. This could include educating investors on AI risks and opportunities.
Provide policy incentives for scaling AI firms: Policymakers can support the growth of ecosystems of AI firms through targeted policy incentives, including tax credits, grants or matching public financing to capital put forward by VCs or other investors (e.g. hybrid public-private co-investment platforms). Such incentives can help support firms that promise longer-term benefits or are of national interest but might not otherwise receive necessary investment across funding rounds. Policymakers also can incentivise scaling by reducing or eliminating overly burdensome, such as overlapping, regulations.
Foster an enabling investment environment, including across countries with varying levels of economic development: Policymakers can strengthen investor confidence by establishing mechanisms that provide a clear, predictable and enabling investment environment. This may include creating regulatory experimentation mechanisms, such as sandboxes, policy and regulatory support, or establishing specialised agencies to offer guidance and legal certainty to investors and firms. As countries differ in their AI readiness and economic development level, understanding the gaps and opportunities in local investment ecosystems is crucial to help catalyse local investment in home-grown AI firms while also attracting investment from international markets.
Build resilience through diversification of investment across the AI stack: Aligned with national AI needs and policy goals, policymakers can encourage necessary investment across complementary components of the AI stack, such as in semiconductors, energy-efficient hardware, edge computing, and measures promoting AI security, robustness and resilience across the stack. Facilitation can come in the form of direct investment and reducing regulatory burden and other barriers to entry. Diversifying investment across the AI stack can help economies develop resilient AI ecosystems capable of sustaining innovation even during market corrections or technological shifts, supporting both security and longer-term economic competitiveness.
Explore what drives investments into trustworthy AI: The present analysis of AI VC investment trends can be further expanded to support a study of the sort of criteria investors use to make investment decisions, with a focus on trustworthy AI. Policymakers could benefit from assessing the extent to which investors, including VC investors, are using trustworthy AI considerations to drive their investment strategies. Such a study could take stock of present criteria used by investors and assess whether opportunities or gaps exist where standardised investment terminology around trustworthy AI may be useful. This could help investors evaluate the trustworthiness of AI firms consistently, help with regulatory compliance, and encourage alignment with trustworthy AI best practices such as the OECD Recommendation on AI.