Accel said on Aug. 11 that it had closed a new $3.5 billion raise. The announcement came only four months after the firm raised a $5 billion late-stage investment fund, leaving the long-established venture investor with $8.5 billion in fresh capital in a span of four months, all pointed at AI.
Accel describes AI as a technology “supercycle” that is still in its early stage. Its view is that AI is sharply shortening the time between a startup’s initial product idea and large-scale expansion.
Yet the private market is not getting cheaper. The gap between the productivity story around AI and the pricing reality in venture is widening: seed rounds are getting larger, early-stage valuations are moving higher, and some AI startups without mature products or revenue models are already attracting capital once more typical of growth-stage companies.
AI may lower part of the cost of building a company, but it is also pushing up the cost of acquiring equity in the best AI startups.
Lower startup costs, bigger venture bets
AI tools are allowing some software, SaaS, and fintech startups to build products and complete early validation with less capital. A small team can now handle work that previously required a larger group of engineers, sales staff, and operations hires.
Research from cap table management provider Carta shows that the median team size for seed-stage startups in the US is now just four people. Average headcount at the Series B stage fell to 45 from 53 in 2023, while Series D headcount dropped 29% from its peak to 131.
Funding structures, though, are not becoming more even. AI is pushing the private market toward a more polarized shape, with fundraising increasingly taking on a barbell structure.
On one side are lightweight startups using AI tools to cut fixed costs. These companies need less initial capital for product development and commercial validation. On the other side are a small number of AI startups with elite teams and technical pedigrees, and they are seeing round sizes and valuations rise across the board from the earliest days of company formation.
Carta said about 3,000 US startups completed pre-seed financings in the first quarter of 2026, with total funding expected to come to roughly $2.9 billion, little changed from recent quarters. AI startups accounted for 50% of that capital, up from about 30% several years ago.
The distribution of checks has shifted as well. Mid-sized rounds between $1 million and $2.5 million fell to 18% of the total from 24% in the first quarter of 2023. Financings below $1 million became more common, while the share of rounds above $2.5 million held steady.
The valuation gap is widening. In SAFE deals above $2.5 million tracked by Carta, the top 10% of startups by valuation now carry valuation caps above $100 million. By the second quarter of 2026, the top 5% of seed financings in Carta’s data reached about $200 million in valuation, up 177% from $72.2 million a year earlier.
That means AI is redrawing what “early-stage financing” looks like, and the change has become especially visible in recent months.
A number of key researchers and executives from Google and OpenAI have left to launch new companies. Many of those projects remain at a very early stage, yet private market pricing has already moved into the hundreds of millions or even tens of billions of dollars.
In early August, Jeff Dean, Google’s chief scientist after nearly 27 years at the company, left with Sanjay Ghemawat, Quoc Le, Oriol Vinyals, and other key Google and DeepMind researchers to start Discovery Loop, an AI scientific research company.
When Discovery Loop was formally established, the company had not even finished building out its team or office setup. Still, it had already secured backing from Radical Ventures, Khosla Ventures, Lightspeed, and Kleiner Perkins, with Alphabet also participating as a founding investor. A later report said the company was discussing a financing of about $1 billion at a valuation of roughly $10 billion.
Kevin Weil, OpenAI’s former chief product officer, left this year and is preparing an AI science company whose name and product have not yet been made public. According to the latest report in August, the project is seeking about $150 million at a valuation of at least $750 million.
In April, Ineffable Intelligence, founded by former DeepMind reinforcement learning researcher David Silver, completed a $1.1 billion seed round at a $5.1 billion valuation, one of the largest seed financings in Europe.
In this AI startup cycle, capital is placing a price on the future potential of elite teams almost all at once, long before operating milestones would traditionally justify it.
Higher valuations raise the cost of ownership for VCs
In the past, product strength, user growth, and revenue were the usual drivers of higher valuations over time. Now, for teams coming out of top labs such as OpenAI and Google DeepMind, research credentials, talent density, and the possibility of building a platform company can be capitalized almost from day one.
That matters because venture capital has a plain business model: returns depend heavily on entry price and ownership percentage. For early-stage funds, building a large enough initial position when valuations are still low and then maintaining ownership through later rounds has long been a key part of generating outsized returns.
In the current AI rush, the window for getting quality companies at lower prices is shrinking quickly.
Carta’s July data add another layer. Over the past six months, dilution in US software financings continued to fall. Median dilution was about 18% for both seed and Series A rounds, 12% for Series B, and below 10% for Series C. Over the same period, median seed valuation reached $24.3 million, while Series A and Series B valuations climbed to $80 million and $191 million, respectively.
In other words, round sizes and company valuations are rising, but founders are not giving up a larger share of the company in return. For VCs trying to gain or preserve meaningful ownership, the amount of capital needed to secure the same stake is rising sharply.
Take a company with a $90 million pre-money valuation. A VC investing $10 million could end up with roughly 10% ownership. If that same company is priced at $490 million at a similar stage, getting close to a 10% stake would require around $50 million.
That changes the game for venture firms. They need to get in, and they need to keep up. Large firms now need two capabilities at once: securing enough initial ownership early, then reserving enough capital to follow on after valuations rise.
That logic has spread across the top end of the industry. Of Accel’s new $3.5 billion raise, $1.35 billion is earmarked for a global expansion fund, meant in part to support larger initial checks and fast follow-on investments. The $5 billion late-stage fund raised in April gives the firm more room to keep backing portfolio companies as they mature.
In January, Andreessen Horowitz, or a16z, raised more than $15 billion in one go. Of that total, $6.75 billion was set aside for growth-stage investing, and $1.7 billion was dedicated to AI infrastructure.
B Capital later closed a $500 million early-stage fund, double the size of its prior comparable vehicle. Its management said that with so much capital flowing into the early market, some early financings now resemble former growth-stage deals in both valuation and size.
More capital is concentrating in a small group of AI leaders
Venture money is not spreading evenly across the AI startup universe. It is moving toward a smaller set of leaders.
Carta’s latest figures show that companies on its platform raised $58.7 billion in venture funding in the first half of 2026, up from $56.5 billion a year earlier. But the split by stage grew much wider. Seed funding fell to $3.8 billion from $6.5 billion, while Series B dropped to $10.4 billion from $13.5 billion. Series A was roughly flat at $12.7 billion. Series C and later, by contrast, rose to $31.8 billion from $23.9 billion.
Crunchbase said more than 70% of global startup funding in the second quarter of 2026 went to AI companies, up sharply from less than 50% a year earlier. During the same quarter, 16 companies raised more than $1 billion each, collecting a combined $108.6 billion and accounting for 53% of total venture funding for the quarter.
The pull of the largest companies is even more pronounced. In the first half of the year, OpenAI and Anthropic together raised $217 billion, equal to 43% of total global startup funding. That suggests the current AI investment cycle is not a broad diffusion of capital across a large field of startups. It is becoming more concentrated in a small number of foundation model companies and already validated leaders.
That concentration sharpens competition among big funds.
If only a handful of AI companies eventually become global platforms, missing their cap tables could have a direct effect on fund-level returns. Instead of increasing the number of portfolio companies, more firms are choosing to own fewer names and write larger checks into a smaller group of higher-conviction opportunities.
A self-reinforcing loop follows. Strong AI companies grow faster, which pulls VCs into competition earlier. Competition raises early valuations, which pushes up the capital needed to maintain ownership. Capital then concentrates further in the leaders, strengthening the position of larger funds in later rounds.
The result is a widening gap not only among startups, but inside venture capital itself, with smaller funds feeling the pressure more acutely.
A $100 million fund could once spread capital across dozens of seed deals. But if a single top AI round reaches tens of millions of dollars and investors also need reserves for future rounds, smaller funds will find it harder to participate in the most sought-after AI companies.
Larger firms can cover a company’s full capital cycle with different pools of money: early-stage funds establish the position, growth funds add to it, and late-stage funds help preserve ownership. Accel’s $8.5 billion raised over four months is a clear sign of that trend.
High valuations are pulling future growth into today’s price
Bigger venture funds do not make returns easier. The higher the entry valuation, the more growth and the larger the eventual exit required to justify the investment.
If a company is funded at a $100 million valuation, reaching $1 billion produces a 10x valuation increase. But if an early-stage company is already priced at $10 billion, the same 10x outcome would require a $100 billion valuation.
Thinking Machines was valued at $12 billion in its first round. SSI reached a $32 billion valuation in less than a year after it was founded. Those figures reflect the scarcity attached to elite AI teams, but they also show that a substantial part of future growth expectations has already been priced in.
That is where the risk in today’s AI private market sits. If several platform AI companies emerge with revenue and profit large enough to support market capitalizations in the hundreds of billions of dollars, current valuations may still be absorbed. If most AI companies fail to build strong enough technical and commercial moats, high entry prices will directly compress the return potential for investors.
On paper, the model still looks strong. Carta data show that some VC funds launched in 2023 and 2024 are currently posting higher gross internal rates of return than some older funds from 2017 to 2020. But a meaningful share of those gains comes from markups in follow-on financings rather than cash returned through IPOs or M&A. Rising private market valuations will still need to be validated by future revenue growth and exit prices.
For large firms such as Accel, the more immediate question is not whether valuations eventually cool. It is how to avoid missing the few winners that may define this technology cycle.
That helps explain why the largest funds keep expanding. AI company growth and financing cycles are shortening. Once a project gains market validation, valuations can rise rapidly across consecutive rounds. If a VC wants enough ownership early and wants to maintain it later, it needs to prepare capital reserves well above what was once standard.
Accel calls AI an early-stage technology supercycle. On that basis, its new $3.5 billion early and expansion capital, together with the $5 billion late-stage fund it raised earlier this year, reflects one strategy: get into potential winners as early as possible and keep enough room to follow on as valuations rise.
That is the central contradiction in the current AI investment boom. AI reduces part of the cost of starting a company, but it does not reduce the cost of investing in the best AI companies.
For venture firms, what has become expensive is not only the money required to support a startup’s growth. It is the price of winning and defending a meaningful ownership stake within a much shorter window.