Author: Zen, PANews
On August 11th, the veteran venture capital firm Accel announced the completion of a new $3.5 billion fundraising round. Earlier in April, the firm had just raised a $5 billion late-stage investment fund. Within a mere four months, Accel has loaded $8.5 billion into its arsenal.
Accel views AI as a technological "super cycle" still in its early stages. In their assessment, AI is significantly compressing the timeline for startups to progress from product conception to scaling.
However, in contrast to the "efficiency and cost-cutting" narrative of the technology, the AI primary market is becoming increasingly expensive. Seed funding rounds are growing larger, early-stage valuations are continually rising, and some AI companies without mature products or revenue models are already able to secure capital that was once reserved for growth-stage enterprises.
AI has reduced some of the costs of starting a company, but it has simultaneously driven up the price of acquiring equity in promising AI firms.
Startup Costs Fall, Capital Bets Rise
AI tools are enabling some software, SaaS, and fintech startups to complete product development and early validation with less capital. A team of limited size can now accomplish work that previously required many more engineers, sales, and operational staff.
According to research data from equity management service provider Carta, the median team size for seed-stage startups in the US is currently just 4 people; the average employee count for Series B rounds has decreased from 53 in 2023 to 45, and Series D has dropped 29% from its peak to 131.
In terms of startup financing structures, AI is pushing the primary market towards polarization, with the fundraising market gradually showing a clear "barbell" structure.
On one hand, for lightweight startups that significantly reduce fixed costs by relying on AI tools, the initial capital required for product development and business validation is noticeably decreasing. On the other hand, a handful of startups with top-tier teams and technical backgrounds are experiencing a comprehensive increase in fundraising scale, securing far more capital and higher valuations than ordinary startups in their early stages.
Carta statistics show that in Q1 2026, approximately 3,000 startups in the US completed Pre-Seed funding, with a final estimated fundraising amount of about $2.9 billion, remaining roughly flat compared to past quarters. AI startups' share of this funding has already reached 50%, whereas this proportion was about 30% a few years ago.
In terms of capital distribution, mid-sized rounds between $1 million and $2.5 million have decreased from 24% in Q1 2023 to 18%, smaller rounds under $1 million are more common, while the share of large rounds over $2.5 million has remained stable.
Furthermore, valuations for top projects are widening the gap further. Among SAFE transactions with fundraising amounts exceeding $2.5 million tracked by Carta, the top 10% of startups by valuation now have valuation caps exceeding $100 million. In Q2 2026, the top 5% of seed round valuations tracked by Carta reached approximately $200 million, a 177% increase from $72.2 million in the same period of 2025.
AI is redefining the traditional concept of "early-stage funding," a change that has been particularly pronounced in recent months.
A group of core researchers and executives from top AI companies like Google and OpenAI have recently left to start their own ventures. While these projects are still in their very early stages, capital valuations have rapidly entered the range of hundreds of millions or even billions of dollars.
In early August, Jeff Dean, who worked at Google for nearly 27 years and served as Chief Scientist, along with several other key researchers from Google and DeepMind including Sanjay Ghemawat, Quoc Le, and Oriol Vinyals, left to found the AI scientific research company Discovery Loop.
At the time of its official establishment, the company hadn't even completed team assembly and office setup, but it had already secured support from institutions like Radical Ventures, Khosla Ventures, Lightspeed, and Kleiner Perkins, with Alphabet also participating as a founding investor. Subsequent reports indicated that Discovery Loop was in discussions for a funding round of approximately $10 billion, with a corresponding valuation of around $100 billion.
Former OpenAI Chief Product Officer Kevin Weil, after leaving this year, is preparing an AI science company whose name and product have not yet been disclosed. According to the latest report in August, this project is seeking about $150 million in funding, with a corresponding valuation of at least $750 million.
In April of this year, Ineffable Intelligence, founded by former DeepMind core reinforcement learning researcher David Silver, completed an $11 billion seed round at a $5.1 billion valuation, becoming one of the largest Seed financings in Europe.
In this era of AI entrepreneurship, capital is effectively "cashing out" the future potential of top-tier teams in one go.
Rising Valuations Increase VC's Cost of Ownership
In the past, product capabilities, user growth, and revenue were often key bases for gradual valuation increases. Today, for teams hailing from top labs like OpenAI and Google DeepMind, research track records, talent composition, and the potential to build a platform company in the future can themselves be capitalized on in the early stages of a company's founding.
Venture capital has a rather straightforward business model; its returns depend significantly on entry price and ownership stake. For early-stage funds, building a sufficiently large initial position when company valuations are still low and maintaining a corresponding equity share through subsequent funding rounds is a crucial foundation for achieving outsized returns.
However, in the current AI investment frenzy, the window to acquire equity in high-quality projects at low prices is rapidly closing.
The latest data released by Carta in July this year further shows that over the past six months, equity dilution percentages in US software company fundings continued to decline. The median dilution percentage for seed and Series A rounds was about 18% each, dropping to 12% for Series B, and falling below 10% for Series C. During the same period, the median seed round valuation reached $24.3 million, with Series A and B reaching $80 million and $191 million respectively.
In other words, while fundraising amounts and company valuations continue to rise, the equity share given up by founding teams has not increased proportionally. For VCs hoping to acquire or maintain a relatively high ownership percentage, the capital required to obtain the same equity share is rising significantly.
Suppose a startup has a pre-money valuation of $90 million. A VC investing $10 million could roughly secure a 10% stake. If the valuation for a company at the same stage rises to $490 million, obtaining a similar 10% stake would require an investment of approximately $50 million.
This also means that VCs need both the ability to get in and the capacity to keep up. This requires large VCs to possess two capabilities simultaneously: securing sufficient initial ownership in the early stages, while also reserving ample capital for follow-on investments as company valuations skyrocket.
This trend has become a consensus among all top VCs. In Accel's recently raised $3.5 billion, $1.35 billion was allocated to a global expansion fund, a key purpose of which is to support larger initial early-stage investments and rapid follow-ons. The $5 billion late-stage fund raised in April provides Accel with the capital to maintain investment capacity as portfolio companies enter the growth stage.
In January of this year, a16z completed a fundraising of over $15 billion in one go, with $6.75 billion earmarked for growth-stage investments and $1.7 billion specifically allocated to AI infrastructure.
B Capital subsequently closed a $500 million early-stage fund, doubling the size of its previous fund of the same type. Its management pointed out that with massive capital entering the early-stage market, valuations and deal sizes for some early-stage financings are increasingly resembling those of past growth stages.
Matthew Effect: Capital Further Concentrates on a Few Top AI Projects
Primary market capital is increasingly concentrating in a handful of top projects.
The latest data from Carta shows that in the first half of 2026, the companies it covers completed a total of $58.7 billion in venture funding, higher than the $56.5 billion in the same period last year. However, the divergence between different funding stages has significantly widened. Seed round funding dropped from $6.5 billion to $3.8 billion, Series B decreased from $13.5 billion to $10.4 billion, Series A remained roughly flat at $12.7 billion, while Series C and later stages grew from $23.9 billion to $31.8 billion.
Crunchbase statistics show that in Q2 2026, over 70% of global startup funding flowed to AI companies, a significant increase from less than 50% in the same period last year. During that quarter, 16 companies completed funding rounds exceeding $10 billion, totaling $108.6 billion and accounting for 53% of the quarter's total venture capital.
The capital attraction effect of leading companies is even more pronounced. In the first half of this year, OpenAI and Anthropic alone raised a combined $217 billion, constituting 43% of the total fundraising amount for global startups. This indicates that the current AI investment boom is not a diffusion of capital evenly across many startups, but rather an increasing concentration on a few foundational model companies and market-recognized top projects.
This concentration further intensifies the competition among large funds for these top projects.
If ultimately only a handful of AI companies can become global platforms, then a large fund's failure to enter the shareholder list of these companies could directly impact the return performance of the entire fund cycle. Compared to expanding the number of investments, an increasing number of institutions are opting to reduce investment targets and deploy larger amounts into a few high-conviction projects.
This has created a self-reinforcing mechanism: high-quality AI companies grow faster, prompting VCs to compete for entry earlier. Mutual competition drives up early-stage valuations, increasing the capital required to maintain ownership stakes. Subsequently, capital further concentrates on top projects, strengthening the advantage of large funds in follow-on rounds.
Ultimately, AI is not only widening the funding gap between startups but also reshaping the competitive landscape within the VC industry itself. Smaller-scale funds are particularly affected.
A $100 million fund could previously spread investments across dozens of seed projects. But if a single round for a top AI project reaches tens of millions of dollars, and investing institutions also need to reserve capital for subsequent rounds, the ability of small funds to participate in hot AI projects will significantly decline.
Large institutions, however, can cover a company's entire capital cycle through funds at different stages: early-stage funds establish positions, growth funds continue follow-ons, and late-stage funds further maintain ownership stakes. Accel raising $8.5 billion within four months is a typical microcosm of this trend.
High Valuations Are Pricing in Future Growth Expectations Early
However, the expansion of VC fundraising does not necessarily mean achieving returns will be easier. In fact, higher entry valuations impose greater demands on a company's future growth and exit scale.
If a startup receives investment at a $100 million valuation, growing to $1 billion brings a tenfold increase in valuation. But if a company's early-stage valuation is already $10 billion, achieving the same tenfold growth requires a final valuation of $100 billion.
Thinking Machines' first funding round already reached a $12 billion valuation; SSI's valuation rose to $32 billion in less than a year since its founding. These valuations reflect the scarcity of top AI teams, but also mean that a significant portion of future growth expectations has already been priced in.
The risk in the current AI primary market lies precisely here. If several platform companies emerge in the future with revenue and profit scales substantial enough to support multi-hundred-billion-dollar market caps, today's high valuations might still be digested. However, if most AI companies ultimately fail to build sufficiently strong technological and commercial moats, excessively high entry prices will directly compress the potential return space for investment firms.
For now, this investment model still appears healthy on paper. Carta data shows that some VC funds established in 2023 and 2024 currently exhibit better paper internal rate of return (IRR) performance than some older funds from 2017-2020. But a considerable portion of these returns comes from valuation mark-ups driven by subsequent funding rounds, rather than actual cash returns from IPOs or M&As. Today's ever-rising primary market valuations will ultimately need to be validated by future revenue growth and exit prices.
But for large VCs like Accel, the more immediate concern is not waiting for valuations to rationalize, but how to avoid missing the potential few winners that may emerge in this technological cycle.
This is also one of the reasons why large funds continue to expand. The growth and fundraising cycles of AI companies are visibly shortening. Once a project gains market recognition, its valuation can rapidly increase over consecutive funding rounds. For a VC hoping to acquire sufficient equity early on and maintain its stake through subsequent rounds, it needs to prepare capital reserves far exceeding those of the past.
Accel defines AI as a technological "super cycle" still in its early stages. Under the premise that this judgment holds, the $3.5 billion early-stage and expansion fund, along with the previously raised $5 billion in late-stage capital, essentially correspond to the same strategy: enter potential winners as early as possible, and reserve space for continuous follow-on investment as their valuations rise rapidly.
This constitutes the most apparent paradox in the current AI investment frenzy: AI reduces some costs of starting a business, but does not reduce the cost of investing in promising AI companies.
For VCs, what has truly become expensive is not just the capital needed to support a startup's growth, but the price required to acquire and maintain a sufficient ownership stake in a quality project within an increasingly shrinking window of opportunity.
