Venture Capital
Last reviewed
Sources
24 citations
Review status
Source-backed
Revision
v2 · 3,271 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Sources
24 citations
Review status
Source-backed
Revision
v2 · 3,271 words
Add missing citations, update stale details, or suggest a clearer explanation.
Venture capital (VC) is a form of private financing in which professional investors put money into early-stage and high-growth private companies, usually startup-style technology firms, in exchange for an equity stake and a share of any eventual upside. Venture funds are built around a power law: most investments fail or return little, so the model depends on a small number of outsized winners large enough to return the entire fund and more. Since 2023, venture capital has been reshaped by artificial_intelligence: a single wave of AI startups absorbed the majority of global venture dollars, individual financings reached sizes that had never previously occurred in private markets, and the relationships among chipmakers, cloud providers, and model developers raised questions about whether reported demand reflected genuine end-user adoption or capital circulating inside a closed loop.
Venture capital is the slice of private financing aimed at young companies that are too risky or too unproven for bank loans or public markets but that have the potential for rapid, large-scale growth. As Silicon Valley Bank puts it, "VCs make it possible for promising entrepreneurs, some with little or no operating history, to secure capital to launch their business" [20]. In return for cash, the venture investor receives preferred shares, typically with rights such as a liquidation preference, anti-dilution protection, information rights, and often a board seat, that protect the investor if the company is sold or fails.
Most venture capital is deployed through a fund rather than by a single investor. A venture fund is a limited partnership with two kinds of participants. The limited partners (LPs) are the outside investors who supply the capital: pension funds, university endowments, sovereign wealth funds, foundations, insurers, corporations, and wealthy individuals. The general partners (GPs) are the venture firm's investors who source deals, decide where to invest, sit on boards, and manage the portfolio. LPs commit capital up front, and the GPs "call" it over the fund's investment period as deals are found. A typical fund has a roughly ten-year life: capital is invested over the first few years, and the remaining years are spent helping companies grow and exit.
The economics of the industry are usually summarized by the phrase "2 and 20." Under this model, the GP charges an annual management fee of about 2 percent of committed capital to cover salaries and operating costs, and takes carried interest ("carry") of about 20 percent of the fund's profits once the original capital has been returned to LPs [21]. According to AngelList, "A management fee usually ranges from 2% to 2.5% of committed capital and is usually charged every year the fund is in operation," while carried interest "is the percentage of profits that go to the GP of a fund," leaving the remaining 80 percent of profits for the LPs [21]. Carry across the industry generally ranges from 15 to 30 percent, but most funds settle at 20 percent [21]. This structure is meant to align incentives: the management fee keeps the lights on, while the bulk of a GP's potential reward depends on actually returning a profit to investors.
Venture financing is organized into a sequence of rounds, each priced higher than the last if the company is succeeding, with the round letter signaling a company's maturity and the investor's risk. Earlier rounds are smaller and riskier; later rounds are larger, command higher valuations, and de-risk the investment because the company has proven traction. In the United States, priced rounds from Series A onward typically follow the model legal documents published by the National Venture Capital Association (NVCA), which is why terms such as a 1x non-participating liquidation preference are market standard [22].
| Stage | What it funds | Typical investors |
|---|---|---|
| Pre-seed / Seed | An idea, a founding team, a prototype or minimum viable product, and early market research to prove the concept can grow | Angel investors, seed funds, startup accelerators such as Y Combinator |
| Series A | Shipping a commercial product, finding product-market fit, and building out the team; usually the first priced round with a lead investor and a board seat | Early-stage venture firms |
| Series B | Scaling sales, marketing, distribution, and operations to grow revenue (often from the first roughly 1 million US dollars of revenue toward 10 million and beyond) | Early- and growth-stage venture firms |
| Series C and later (growth / late stage) | Expanding into new products or markets, acquisitions, and scaling a company with proven traction toward an exit | Growth funds, crossover funds, hedge funds, sovereign wealth funds, private equity firms |
The exit, when a venture investor finally converts a paper stake into cash, normally comes through an initial public offering (IPO) or an acquisition. When public listings are scarce, as in much of the 2022 to 2025 window, investors increasingly turn to secondary sales and tender offers to realize returns, a dynamic discussed below.
The defining feature of venture capital in this period is concentration. Where the 2010s and the 2020 to 2021 boom spread funding across consumer apps, fintech, crypto, and software-as-a-service, the 2024 to 2026 cycle funneled capital into a short list of foundation_model developers and the infrastructure that trains and serves them. According to Crunchbase, just five companies, openai, scale_ai, anthropic, Project Prometheus, and xai, together raised roughly 84 billion US dollars in 2025, or about a fifth of all venture funding that year [1]. The shift was accompanied by new vehicles for deploying capital at scale, by a structural shortage of exits that left limited partners waiting for returns, and by a sustained public debate over whether AI valuations constituted a bubble.
Artificial intelligence went from one sector among many to the single dominant destination for venture capital in the span of two years. In 2024, total global venture funding reached close to 314 billion US dollars, up about 3 percent from 304 billion in 2023, and AI-related companies took in more than 100 billion US dollars of that, up more than 80 percent from about 55.6 billion in 2023 and amounting to close to a third of all global venture funding [23]. In the United States the tilt was already pronounced: the PitchBook-NVCA Venture Monitor reported that nearly 30 percent of completed US venture investments went to AI companies in 2024 [22]. The single largest financing that year was openai's 6.6 billion US dollar round in October 2024, led by thrive_capital at a 157 billion US dollar valuation, which was at the time the largest venture round ever closed [24].
By 2025 AI had crossed from leading sector to outright majority of venture dollars, a level without historical precedent. Estimates of AI's share vary by data provider because of differing definitions of what counts as an AI company and which capital flows are included, but every major tracker placed the figure above half of global dollars in 2025.
PitchBook reported that AI startups received about 192.7 billion US dollars of roughly 366.8 billion US dollars in total global venture investment in 2025, a share of approximately 52.5 percent and the first time AI claimed a majority of venture dollars worldwide [2]. Crunchbase, using a broader tally of global venture funding it put at about 512.6 billion US dollars, calculated that AI accounted for roughly 52.7 percent, with AI funding of about 211 billion US dollars, up roughly 85 percent from about 114 billion US dollars in 2024 [1]. The Organisation for Economic Co-operation and Development cited an even higher figure, reporting that AI firms captured about 61 percent of global venture capital in 2025 [3].
The concentration was sharper in the United States. PitchBook data indicated that US AI companies captured close to two-thirds of all domestic venture capital in 2025, with AI investment rising from about 73 billion US dollars in 2022 to more than 200 billion US dollars in 2025 [2]. By multiple accounts, mega-rounds drove the bulk of this total: Crunchbase reported that more than a third of global funding in 2025 went to the 68 companies that each raised 500 million US dollars or more, up from about a quarter of funding in 2024 [1].
The period produced the largest private financings in history, repeatedly resetting records. The table below summarizes major AI venture rounds reported between 2025 and early 2026. Figures are post-money valuations as reported by the cited sources, several of which involved committed rather than fully wired capital.
| Company | Round / date | Amount raised | Valuation | Lead investors |
|---|---|---|---|---|
| openai | Mar 2025 | ~40 billion USD | ~300 billion USD | softbank [4] |
| openai | Mar 2026 | ~122 billion USD committed | ~852 billion USD | Amazon, nvidia, softbank, andreessen_horowitz [5] |
| anthropic | Sep 2025 | ~13 billion USD | ~183 billion USD | ICONIQ and others [6] |
| anthropic | Feb 2026 (Series G) | ~30 billion USD | ~380 billion USD | GIC, Coatue [6] |
| xai | Jan 2026 (Series E) | ~20 billion USD | ~230 billion USD | Valor, nvidia, Cisco and others [7] |
| safe_superintelligence | Apr 2025 | ~2 billion USD | ~32 billion USD | Greenoaks [8] |
| thinking_machines_lab | Jul 2025 (seed) | ~2 billion USD | ~12 billion USD | andreessen_horowitz [9] |
| mistral_ai | Sep 2025 (Series C) | ~1.7 billion EUR | ~14 billion USD | ASML [10] |
OpenAI's trajectory illustrates the velocity of repricing. Its March 2025 financing of about 40 billion US dollars, led by softbank with a syndicate including Microsoft, Coatue, Altimeter, and thrive_capital, valued the company at about 300 billion US dollars and was, per PitchBook, nearly three times the largest amount previously raised by any private technology company [4]. SoftBank disclosed that its share could be reduced if OpenAI did not restructure into a for-profit entity by the end of 2025; the company completed that restructuring and SoftBank closed its portion, ultimately about 41 billion US dollars, in late December 2025 [11]. In March 2026 OpenAI announced an even larger financing, reported by Bloomberg and CNBC at about 122 billion US dollars of committed capital and a post-money valuation of about 852 billion US dollars, with Amazon committing about 50 billion US dollars, nvidia and SoftBank about 30 billion US dollars each, and a notable portion of Amazon's commitment contingent on OpenAI reaching a public listing or an agi milestone [5].
Two startups stand out for raising on reputation rather than revenue. safe_superintelligence, founded by former OpenAI chief scientist ilya_sutskever, reached a reported 32 billion US dollar valuation in April 2025 with no product and roughly 20 employees [8]. thinking_machines_lab, founded by former OpenAI chief technology officer mira_murati, closed a 2 billion US dollar seed round at a 12 billion US dollar valuation in July 2025; a subsequent attempt to raise as much as 50 billion US dollars reportedly collapsed without a deal by January 2026, and the company remained at its seed valuation [9].
The boom prompted established venture firms to raise the largest funds in their histories, often explicitly to fund AI growth-stage companies and to defend existing stakes through follow-on investment.
andreessen_horowitz, commonly called a16z, was reported in April 2025 to be seeking roughly 20 billion US dollars for what would have been the largest fund in its history, aimed at growth-stage American AI companies and follow-on positions in databricks, xai, Mistral, and Safe Superintelligence [12]. The firm did not close at that target; in January 2026 it announced raising about 15 billion US dollars across five funds, including roughly 6.75 billion US dollars for growth and about 1.7 billion US dollars for AI infrastructure [13]. thrive_capital, founded by Joshua Kushner and a longtime lead backer of OpenAI, closed a fund exceeding 10 billion US dollars, roughly double its prior vehicle, having led OpenAI's 6.6 billion US dollar round in October 2024 [14]. sequoia_capital raised about 7 billion US dollars for its largest late-stage fund, targeting AI companies including OpenAI and Anthropic [15]. lightspeed_venture_partners closed multiple funds totaling several billion US dollars in 2025, holding sizable stakes in Anthropic and Mistral [15]. founders_fund and Greenoaks were active leads on AI rounds, Greenoaks anchoring the Safe Superintelligence financing [8]. lux_capital, which specializes in deep-tech and science-driven startups, was among the established firms scaling up to back AI and frontier-technology companies during the cycle.
The most dramatic strategic shift came at softbank. masayoshi_son repositioned the conglomerate as an AI-first investor, slowing other Vision Fund dealmaking, requiring his personal approval for deals above about 50 million US dollars, and funding the OpenAI commitment in part by selling SoftBank's entire roughly 5.8 billion US dollar stake in nvidia and portions of its T-Mobile holding [11]. SoftBank also participated in stargate, the AI data-center venture announced with OpenAI and Oracle that targeted up to 500 billion US dollars of capacity in the United States [11].
The scale and illiquidity of AI deals pushed capital into vehicles beyond traditional primary rounds. Special purpose vehicles, or SPVs, which pool money from many investors into a single allocation, proliferated as backers shut out of primary rounds sought exposure to the most sought-after names, frequently at a premium. In August 2025 OpenAI publicly warned against unauthorized SPVs and similar structures, stating that sales attempting to circumvent its transfer restrictions would not be recognized and would carry no economic value, and Anthropic likewise insisted that Menlo Ventures invest through its own funds rather than syndicate to co-investors [16]. The collapse of Linqto, which had offered retail investors secondary exposure via SPVs, underscored the opacity of these arrangements [16].
The secondary market expanded as the primary route to liquidity, the initial public offering, remained largely shut. Jefferies recorded a record of about 103 billion US dollars of private-equity stakes traded in the first half of 2025, up roughly 51 percent year over year [17]. Secondaries, tender offers, and continuation funds, which allow a manager to move assets into a new vehicle and provide an exit to existing investors, became central tools for returning capital without an acquisition or listing.
This liquidity problem is captured by distributions to paid-in capital, or DPI, the share of invested capital actually returned to limited partners. Industry trackers reported that distribution rates ran in the single digits as a share of net asset value for roughly eight consecutive quarters, well below the long-run average, as venture-backed IPO volume stayed far below the 2021 peak of more than 300 listings [17]. A modest recovery in public listings during 2025, concentrated in fintech and AI infrastructure, offered the first signs that distributions might resume.
The most scrutinized concern is circular financing, in which a company invests in or lends to a customer that uses the proceeds to buy the investor's products, allowing the resulting spending to be reported as revenue. Reporting by Bloomberg and others documented arrangements in which nvidia invested in cloud and model companies that in turn became major buyers of its chips, including a large stake in coreweave, and in which OpenAI signed long-term infrastructure commitments with vendors that were themselves investors or partners, including oracle, Microsoft, Broadcom, Amazon, and CoreWeave [18]. Critics, including analysts at the CFA Institute, argued that such loops can inflate apparent demand and that a default, a credit tightening, or a plateau in AI demand could propagate losses through the interconnected parties [18]. Antitrust authorities in the United States and the European Union examined whether some exclusive AI partnerships amounted to disguised acquisitions structured to avoid merger review [18].
A broader bubble debate ran in parallel. Skepticism intensified after a July 2025 study from the MIT Project NANDA initiative reported that about 95 percent of enterprise generative-AI pilots produced no measurable profit-and-loss impact, against tens of billions of US dollars of corporate spending [19]. Prominent figures including OpenAI chief executive sam_altman cautioned that some AI startups were overvalued, and Meta chief executive mark_zuckerberg acknowledged that a collapse was a possibility even as his company raised capital expenditure sharply [19]. The fragility was demonstrated in January 2025, when the release of a competitive low-cost model from China's deepseek triggered a roughly one-day loss of about 1 trillion US dollars across AI-related equities, including the largest single-day market-value decline ever recorded for a single company at nvidia [19].
Geographically the boom remained heavily concentrated in the United States, which captured the large majority of AI venture dollars. China continued to fund a domestic frontier led by companies such as deepseek, zhipu_ai, and moonshot_ai, while Europe's most prominent challenger was France's mistral_ai, whose September 2025 round led by chip-equipment maker ASML valued it at about 14 billion US dollars and stood out as the largest European AI financing of the period [10]. Whether the cycle resolves through durable revenue growth or a sharp repricing of valuations remained, as of mid-2026, an open and widely contested question.