Bessemer Gathers $5.75 Billion With $4 Billion for Growth
Bessemer Venture Partners closed its biggest fund ever at $5.75 billion, steering $4 billion toward growth rounds as AI startups stay private and capital needs balloon.
Bessemer Venture Partners has raised $5.75 billion in new capital, its largest fundraise ever, with $4 billion earmarked for growth-stage investments as AI companies raise bigger rounds and stay private longer, partner Sameer Dholakia said on Bloomberg Tech.
Bessemer Venture Partners has closed on $5.75 billion of new capital, the largest fundraise in the firm's history, and has set aside $4 billion of it for growth-stage investments. The split is the story: roughly seven of every ten dollars is aimed not at first cheques into unproven teams but at companies that already exist, already have revenue, and increasingly have no intention of going public any time soon.
Partner Sameer Dholakia laid out the reasoning in an interview with Ed Ludlow on Bloomberg Technology, describing a "special moment" for investing across AI applications and what he called physical AI \u2014 robotics and machines that act in the world rather than only on a screen. He also returned to a framing he has used before: that intelligence itself could become an economic input on the order of oil, something bought, priced and consumed at industrial scale.
Why the money is weighted toward later rounds
Venture firms have traditionally led with seed and Series A capital, where a small cheque buys a large percentage. That model strains when the companies that matter most are absorbing capital at a rate no early-stage fund can match. Dholakia's explanation for the $4 billion growth allocation is straightforward: AI companies are raising larger rounds and remaining private longer.
Both halves of that sentence create the same problem for a fund manager. Larger rounds mean the price of staying at the table rises with every financing \u2014 an investor who does not have reserves gets diluted out of exactly the position it worked hardest to establish. Staying private longer means the public market, which used to be the point at which venture returns were realised and recycled, arrives years later than it once did. Capital that would historically have come back to limited partners instead has to be replaced with new capital.
A dedicated growth pool is the structural answer. It lets a firm defend ownership in its winners through successive rounds without cannibalising the early-stage budget, and it gives limited partners exposure to companies at a stage that used to belong to crossover funds and mutual funds.
The Anthropic position, and the safety argument
Bessemer was an early backer of Anthropic, and Dholakia used the interview to argue that the AI developer has been consistent on safety \u2014 a point that matters commercially as much as ethically. Enterprise buyers signing multi-year commitments to a model provider are underwriting a vendor's judgement as much as its benchmarks, and consistency of posture is one of the few things a buyer can actually verify over time.
It is also the clearest illustration of the fundraising logic. A position established early in a company that has since become one of the largest private capital absorbers in the industry is only worth what the investor can hold onto. Without growth capital, the early conviction gets progressively priced away by each new round of larger, later investors.
Intelligence priced like a commodity
Dholakia's oil analogy is the sort of line that sounds rhetorical until you look at how AI is actually sold. Model access is metered, priced per unit of consumption, and increasingly procured the way a manufacturer procures energy \u2014 in volume, on contract, with an eye on unit cost trending down over time. If intelligence behaves like an input rather than a product, the investable layer shifts to whoever controls supply, whoever refines it into usable form, and whoever builds the machinery that consumes it.
That framing also explains the interest in physical AI. Software applications monetise intelligence at the desk. Robotics and autonomous systems monetise it in warehouses, factories and vehicles \u2014 a far larger addressable base of labour and capital spending, and one where the buyer is already used to signing eight-figure equipment orders.
A cautious tape for a bullish thesis
The announcement lands on a soft session for public risk assets. As of the last trade at 19:27:54 GMT on Wednesday, the S&P 500 tracker (NYSEARCA: SPY) was at $767.50, down 0.76% from the prior close of $773.38, having traded between $766.50 and $773.59. The Nasdaq 100 tracker (NASDAQ: QQQ) was weaker still at $740.68, off 0.91% against a $747.46 close and near the low of its $738.19\u2013$747.30 range. The Dow tracker (NYSEARCA: DIA) sat at $514.52, down 0.67%.
The tech-heavy benchmark leading the decline is the relevant detail. Private capital commitments of this size are made on a ten-year clock and are largely indifferent to a single red session \u2014 but the exit environment is not indifferent. A fund weighted toward growth-stage positions needs public markets to be receptive when those companies eventually list, and every stretch of AI-led weakness in the listed market is a data point on what that reception might look like.
What to watch from here
Three things will determine whether the $4 billion growth allocation reads as prescient or as a top-tick.
- Round sizes. If private AI financings keep scaling, a growth pool of this size is the minimum ticket to participate. If they plateau, the same capital chases fewer opportunities at bid-up prices.
- Time to listing. The longer companies stay private, the more of the value accrues to growth investors rather than public shareholders \u2014 and the longer limited partners wait for distributions.
- Physical AI economics. Robotics has a long history of impressive demonstrations and disappointing gross margins. Whether intelligence-as-an-input really translates into hardware businesses that earn their capital is still unproven.
What is not in dispute is the direction of travel. The biggest fund in Bessemer's history, with the majority of it reserved for companies that already exist, is a bet that the AI cycle's returns will be captured after the seed round \u2014 and that the capital needed to stay in the game has permanently reset higher.
Key facts
- Total raised: $5.75 billion — Bessemer's largest fundraise ever
- Growth-stage allocation: $4 billion of the total
- Nasdaq 100 (QQQ): $740.68, -0.91%, as of 19:27:54 GMT Sept 23, 2026
- S&P 500 (SPY): $767.50, -0.76%, prev close $773.38
Frequently asked questions
How much did Bessemer Venture Partners raise?
Bessemer Venture Partners raised $5.75 billion in fresh capital, the largest fundraise in the firm's history. Of that total, $4 billion is dedicated specifically to growth-stage investments, with the remainder available for earlier-stage commitments. The firm disclosed the raise in an interview with Bloomberg Technology on September 23, 2026.
Why is so much of the fund allocated to growth stage?
Partner Sameer Dholakia pointed to two linked trends: AI companies are raising larger financing rounds and are staying private for longer. Both make it more expensive for an early investor to maintain its ownership stake, and both delay the public listings that traditionally return capital to limited partners. A dedicated growth pool addresses that directly.
What is Bessemer's connection to Anthropic?
Bessemer was an early investor in Anthropic. In the Bloomberg Technology interview, Dholakia said the AI developer has been consistent on AI safety — a point he raised as part of a broader discussion of the firm's AI investing thesis. Bessemer has not disclosed the size or terms of its position through this lead.
What does 'physical AI' mean in this context?
Physical AI refers to systems that apply machine intelligence to the physical world — robotics, autonomous machines and equipment that act rather than only generate text or images. Dholakia named it alongside AI applications as one of two areas where Bessemer sees a "special moment" for investment across the current cycle.
What did Dholakia mean about intelligence and oil?
He argued that intelligence could ultimately become an economic input comparable to oil — something purchased, metered and consumed at industrial scale rather than sold as a discrete product. The analogy implies value accrues to whoever controls supply and to the systems that consume intelligence in volume across the economy.
How were markets trading when the news broke?
Equity benchmarks were lower. As of the last trade at 19:27:54 GMT on September 23, 2026, the S&P 500 tracker SPY was at $767.50, down 0.76%; the Nasdaq 100 tracker QQQ was at $740.68, down 0.91%; and the Dow tracker DIA was at $514.52, down 0.67%. Tech-weighted exposure led the decline.
Sources
- Bessemer: Anthropic Has Been Consistent on AI Safety — Bloomberg Technology
