The Firm That Funds the Things That Probably Won't Work
Vinod Khosla left a top-tier partnership to write checks nobody else would. Two decades and one $50 million OpenAI bet later, the willingness to be wrong looks like a strategy.
In the folklore of Silicon Valley, the venture capitalist is a pattern-matcher - a professional who studies what worked last time and looks for the next version of it. Vinod Khosla built a firm on the opposite instinct. When he founded Khosla Ventures in 2004, after co-founding Sun Microsystems and spending eighteen years at Kleiner Perkins, he wanted the freedom to fund what he has called "imprudent science experiments": ideas so early, so technical, or so improbable that a conventional partnership would never touch them. Two decades later, that appetite for being wrong has produced one of the most consistent track records in the business.
The Menlo Park firm now manages roughly $16 billion and has invested in more than 800 companies. Its portfolio reads less like a spreadsheet and more like a science fair with a checkbook: an early stake in OpenAI, a plant-based-meat company, a nuclear-fusion reactor, rockets, robots, payments infrastructure, and cancer-detection tools. What ties them together is not a sector. It is a tolerance for the kind of risk most investors are structurally built to avoid.
The ThesisA firm organized around failure
Most venture funds are quietly terrified of losing money. Khosla Ventures inverts that fear. The firm openly expects many of its boldest bets to return nothing, and treats those write-offs as the entry fee for the rare investment that returns a hundred times its capital. To make the math work without contaminating its main fund, the firm runs a dedicated seed vehicle - a place where the riskiest science experiments can live without having to justify themselves like normal investments.
The screening test is unusual. Rather than asking whether a market already exists, the firm's partners tend to ask whether the underlying physics is sound - even when the scientific establishment is skeptical. Founding partner Samir Kaul, who wrote the firm's first and largest check into OpenAI, has built his career on exactly that: outsized first-round bets on companies where the science is defensible but the consensus has not caught up.
That framing changes what a "good deal" looks like. A traditional growth investor prizes predictability: recurring revenue, a proven playbook, a market with visible edges. Khosla Ventures is willing to trade all of that away in exchange for the chance that a technical breakthrough resets an entire industry. The failure rate on such bets is high by design. The point is not to be right often. It is to be occasionally, spectacularly right about something that others were structurally unable to fund.
The bets that landed
The payoff for that discipline is a portfolio of companies that became household names. Khosla Ventures was an early backer of DoorDash, Instacart, Affirm, Block (formerly Square), Stripe, Okta, and GitLab - a run of fintech and consumer-internet outcomes that alone would justify a firm's existence. On the deep-tech side, it funded Guardant Health, Nutanix, QuantumScape, Rocket Lab, and Commonwealth Fusion Systems, the MIT spinout chasing commercial fusion power.
The single bet that best captures the firm's method is OpenAI. Khosla Ventures wrote what was reported to be one of the first large venture checks into the company - roughly $50 million in 2019 - back when a research lab pursuing artificial general intelligence was not an obvious business. When the company's valuation climbed into the hundreds of billions, that early conviction became one of the most valuable venture positions of the decade. In October 2024, the firm raised a separate $405 million special-purpose vehicle to put still more capital into OpenAI, pooling outside investors rather than drawing solely on its own funds.
The breadth is the point. Khosla Ventures has spread its bets across consumer and enterprise software, financial services, semiconductors, health and biotech, agriculture and food, sustainable energy, and robotics. Some of those categories - fusion, synthetic meat, next-generation batteries - carry timelines measured in decades and capital needs measured in the hundreds of millions. A firm that cannot stomach that kind of patience does not fund them at all. Khosla built one that could, and that willingness to wait has become part of the moat.
From a personal checkbook to $16 billion
The firm's scale did not arrive all at once. Khosla capitalized the earliest funds personally, keeping them closed to institutional investors so he could make unconventional bets without a committee to answer to. Only in 2009 did the firm open to limited partners, raising a $1 billion early-and-growth fund alongside a $300 million seed fund. The climb since has been steep: about $1.4 billion across vehicles in 2021, then $3.1 billion across three funds in 2023, including a $1.6 billion eighth flagship fund.
The momentum has not slowed. In early 2025 the firm began raising $3.5 billion across three funds - about 17% more than its 2023 haul - with roughly $650 million earmarked for seed-stage bets. By mid-2026 it was reported to be in talks to raise as much as $5.5 billion, which would be the largest fundraise in its two-decade history, split across seed, early-stage, and a large opportunity fund that concentrates capital into its best performers as they scale.
Engineers with a checkbook
Khosla Ventures is engineer-led in a way that shows up in its decisions. Vinod Khosla grew up in an Indian Army household with no business or technology background, studied electrical engineering at IIT Delhi, and came to the United States for graduate degrees before co-founding Sun Microsystems in 1982. That technical grounding runs through the partnership. Samir Kaul, a founding partner and managing director, focuses on AI, health, sustainability, and food, and has led the firm's investments in companies from Guardant Health to QuantumScape.
The firm's pull on talent is its own quiet signal. Investor Keith Rabois joined Khosla Ventures in 2013, left for Founders Fund in 2019, and then returned as a managing director in January 2024 - a round trip that says something about a culture people want to come back to. Around 240 people support the operation, including a platform team the firm calls Venture Assistance, which helps founders with recruiting, strategy, product, and follow-on fundraising rather than leaving them alone with a wire transfer.
The MarketWhere it sits, and who it competes with
In a venture landscape crowded with large multi-stage firms, Khosla Ventures occupies a distinct corner: early, technical, and comfortable with capital-intensive science. It competes for the same frontier deals as Andreessen Horowitz, Sequoia Capital, Founders Fund, Lux Capital, and Breakthrough Energy Ventures, but its identity is less about scale of assets and more about willingness to underwrite technical risk that others discount to zero. For a founder building something that sounds absurd today, that reputation is the product. It is why the firm often gets the first call on the next thing that probably won't work - and occasionally does.
The business itself is straightforward venture economics. The firm raises money from endowments, foundations, pensions, and family offices; invests it primarily in early-stage private companies; and earns returns through management fees and carried interest when those companies exit through acquisition or public listing. What is unusual is not the model but the risk tolerance layered on top of it - a firm that has decided, deliberately, that the surest way to a rare enormous return is to fund a lot of things that might return nothing at all.
What Founders GetBeyond the wire transfer
For a founder, the practical value of a Khosla Ventures term sheet is twofold. The first is the money: seed capital for the earliest, unproven idea, larger early-stage rounds as the company finds its footing, and the opportunity fund standing ready to concentrate more capital into the winners. The second is harder to price. Being backed by a firm known for funding the improbable is a form of credibility - a signal to future investors, recruits, and customers that serious technical people looked at a difficult idea and decided it was worth underwriting.
The Venture Assistance team turns that reputation into hands-on help. Instead of a partner who appears at quarterly board meetings, founders get access to operators who work on recruiting engineers, refining go-to-market strategy, opening doors to business partners, and preparing the next fundraise. For companies attempting something genuinely new - where there is no established playbook to copy - that operational support can matter as much as the check.