USV’s New $900M Fund: Fred Wilson, Rebecca Kaden & Mike Mignano
Summary
USV’s $900 million fund cycle is a bottoms-up response to larger rounds, costlier scaling, and wider outcomes—not a billion-dollar vanity target. The firm expects roughly 30-40 investments, with Series A checks generally between $10 million and $30 million and substantially more seed activity. Kaden said several LPs surprised her by asking, “Are you sure you shouldn’t raise more?”
USV believes AI is entering the application-layer opening that followed the internet infrastructure buildout around 2003. Its four declared hunting grounds are AI applications, AI infrastructure, physical AI, and energy; the last already represents roughly 20-30% of the portfolio. The core bet is that “the outcomes are bigger, the swings are bigger,” while the decisive applications remain largely unwritten.
Making software has become cheaper, but turning it into a defensible company may require more capital than before. Startups must buy compute, subsidize inference against free offerings from OpenAI, Anthropic, and Gemini, and spend aggressively to escape a flood of easy-to-build competitors. Wilson’s historical yardstick: Twitter’s financings went from $5 million to $25 million to $100 million; in today’s market, “multiply by 10.”
Revenue velocity is not USV’s preferred proof of product-market fit; usage retention and continued product innovation matter more. Annual subscriptions can produce impressive ARR before users have demonstrated that they actually like a product. Kaden and Wilson’s diagnosis is stark: “Trial behavior is at an absolute high” while “engagement behavior is at an absolute low,” forcing investors to infer durable behavior from incomplete cohorts, product judgment, and the team’s iteration.
USV’s emerging moat thesis combines personal agents, multiplayer behavior, open infrastructure, and trusted consumer brands. Wilson expects service-specific agents from Amazon, Shopify, and Etsy, but also a user-controlled agent that coordinates them; Kaden thinks multiplayer behavior may supply the network effects missing from today’s single-player tools. Kaden rejects the idea that this collapses everything into chat: “The user interface is not going away.”
“Obliterate, don’t automate” means replacing an incumbent’s service rather than merely selling it productivity software. USV says 20-30% of its own legal work no longer touches a human lawyer, while Doctronic, SuperTakes, and similar products aim to put medical, investment, or professional expertise directly in consumers’ hands. The opportunity is strongest where an incumbent cannot respond without cannibalizing the people and revenue on which its current business depends.
USV does not need every winner to reach $1 trillion because early ownership in single-digit-billion-dollar exits can still drive exceptional fund returns. The firm nevertheless wants expanding markets and increasingly favors the top one or two companies as value aggregates faster, rather than underwriting the third- or fourth-place player. Wilson’s calibration: many of USV’s biggest winners reached roughly $10 billion, with rarer outliers at $50 billion or $100 billion, and that has already produced “enormous returns.”
Deep dive
1. The $900 million fund was modeled from the portfolio upward
USV raised $900 million in its new fund cycle, “by far the most USV has ever raised” in a new fund cycle. The partnership says the number follows intended ownership, check sizes, stages, and company count—not a desire to cross the psychologically appealing $1 billion line.
The planned portfolio remains concentrated at roughly 30-40 companies. USV still expects to lead or co-lead meaningful early-stage positions, but Series A checks that began near $3 million when the firm started now “start with 10”; the anticipated range is approximately $10 million to $30 million.
Kaden described the fund as large for USV yet small relative to many peers. That preserves a bottom-up process: the firm can select companies matching strong hypotheses rather than needing exposure to a predetermined percentage of the available market simply to deploy the capital.
Long-running LP communication made the shift less contentious than it might have been several years earlier. USV discusses strategy, portfolio successes, and failures quarterly, so investors had watched the market move; some responded to the proposal with, “Are you sure you shouldn’t raise more?”
2. AI now resembles the web just before its defining applications appeared
Wilson’s historical analogy starts in 2003: the commercial internet and much of its infrastructure existed, alongside Google, Amazon, Yahoo, and eBay, but Facebook, Twitter, YouTube, and the iPhone had not arrived. Today, AI infrastructure is similarly advanced while its defining application companies are still emerging.
Kaden sees products such as Granola and Suno as early evidence, followed more recently by Instinct, Muse, and Grockbot. Her call is that agents will power “a whole new type of application layer” across both consumer and enterprise markets.
USV has organized that opportunity into four public theses—AI applications, AI infrastructure, physical AI, and energy. The partnership argues that being conspicuously opinionated narrows inbound deal flow productively: founders working within those areas seek USV out, while founders elsewhere know it is not their natural destination.
3. Cheap product creation does not mean cheap company creation
Mignano’s distinction is between building and breaking out. AI can make a prototype dramatically faster and cheaper, but that produces many more credible competitors; capital then migrates into positioning, marketing, distribution, and the effort required to rise above the noise, particularly in consumer markets.
Compute creates a second cost block whose eventual “asymptotes” remain unknown. Mignano said the fundamental compute costs companies face are in “a totally different hemisphere,” while Kaden said physical-AI, sensor, robotics, and energy businesses carry still more varied capital requirements than conventional software.
Inference may also become a deliberate customer-acquisition subsidy. A startup competing with free core offerings from ChatGPT, Claude, or Gemini cannot necessarily charge before proving value, so it may have to finance substantial usage while seeking the scale at which adoption becomes self-sustaining.
Wilson’s comparison is Twitter: its original financing was $5 million, followed roughly a year later by $25 million and then $100 million. Those looked large in 2007; his rough translation for the current market was, “Multiply by 10,” before even accounting for two decades of inflation.
4. Agents will proliferate, but interfaces and network effects still matter
Wilson expects Amazon, Shopify, and Etsy each to offer an agent, yet believes users will also want a personal agent that represents them and negotiates with those service agents. Today’s agents often impersonate users by logging in directly; eventually, he expects standards for agent-to-agent interaction.
The near-term conflict is already visible: Wilson cited Amazon telling Manus that Manus would not be allowed to access Amazon. His conclusion is not that service agents win outright, but that the market is in “the early innings” of defining permissions, representation, and interoperability.
Kaden’s durability thesis borrows from the social era. Scale and distribution may separate similar consumer agents initially, but lasting retention might require turning today’s single-player behaviors into genuinely multiplayer ones: “It’s not that we don’t need network effects anymore. It’s just that we need to get there.”
Kaden’s pushback against interface extinction is equally important. Agents may shop or file taxes, but they will not watch Netflix or YouTube for users; interfaces remain where people build trust, understand concepts visually, and perform activities they actually enjoy. “The user interface is not going away.”
5. USV is pushing further into seeds, founder bets, and controlled formation
Easier prototyping is encouraging USV to invest before conventional proof exists. Mignano expects more “pure founder bets,” despite their being less traditional for a thesis-driven firm, and selective internal incubations where the partnership has enough conviction to build and test an initial product itself.
Wilson described a continuum from true incubation to “proprietary seeds.” USV might originate an idea, recruit a founder, help shape the company, and provide its first two checks—retaining what he called “full ball control” without literally serving as the founder.
Venture partners Jared Hec and Scott Belsky, together with Mignano’s product background, expand what the firm can build around an idea. Wilson estimated that incubations, founder bets, and proprietary seeds could ultimately represent one-third to one-half of USV’s seed investments.
6. Usage retention matters more than headline ARR
Molly O’Shea revisited a line from her previous Sourcery appearance—“ARR is a fake metric”—with a more precise argument from the discussion. Annual subscriptions and paywalls can generate rapid revenue before proving affection or habit, so USV clicks beneath revenue into daily, weekly, and monthly usage retention.
Kaden framed the present consumer market as a trial-retention divergence: “Trial behavior is at an absolute high” and “engagement behavior is at an absolute low.” People want to sample every prominent new product because AI dominates the discourse, but very few products will become integrated into their lives.
Consumer diligence therefore combines direct product judgment with cohorts. The best signal is when the experience and the data corroborate one another—felt retentiveness appears in returning users, or an apparent reason to churn surfaces in the numbers as an actual retention problem.
Perfect evidence generally arrives too late for USV’s desired entry price. The firm instead looks for proxies: founders who identify the small hooks that refresh a product, teams capable of iterating, and early data suggesting durable behavior before a larger check makes the answer obvious.
7. The best companies and founders rarely travel in straight lines
Abridge began with the idea that patients would record medical appointments; that consumer path did not work, and the team navigated toward the right product, buyer, and market. Kaden’s lesson is that USV had not truly backed the initial customer thesis—it backed Shiv’s ability to carry a directionally right kernel toward the right product and go-to-market.
After making all its money from one product for five or six years, Abridge launched two new lines within 12 months. Wilson called the result a “three-legged stool”: the offerings reinforce one another, and customers become progressively more embedded as they adopt the second and third products.
Wilson sees two or three, perhaps four, stars driving each venture fund. Across USV’s history he cited Twitter, Etsy, Indeed, MongoDB, Twilio, Stripe, Coinbase, Duolingo, and Abridge; Coinbase itself advanced through long plateaus and intermittent breakthroughs rather than a smooth climb.
Kaden’s example of endurance was Stedi founder Zach Kanter: ten years of work, seven before the business truly accelerated, then three years of strong performance when market, team, and product aligned. The conversation offers a similar maturation example in Suno’s Mikey Schulman—founders eventually begin to “see the Matrix” and operate at an entirely different level.
8. The “Rebel Alliance” is an open-stack wager against dependency
Wilson’s label covers developers rebelling against closed AI stacks from OpenAI, Anthropic, Grok, or Gemini. Their fear is historically grounded: companies once built atop Facebook only to be cut off when Facebook entered their markets, and builders now worry Claude or another provider could do the same.
The investable stack is broader than open-weight models. Mignano included model alignment, memory, routing, harnesses, and the tooling enterprises need to manipulate systems for their own requirements; that depth makes the Rebel Alliance a complete USV thesis rather than a single open-source bet.
At USV’s CEO summit, Mignano learned from Coinbase’s CTO that the company had moved away from Claude and built a coding-agent harness atop the open-source product Pi. Mignano said he had not known about the adoption and noted that USV is an investor in a portfolio company identified in the transcript as Arendelle, which put Pi into the market.
9. “Obliterate, don’t automate” attacks the incumbent’s business model
USV contrasts the prior software era—automating established workflows—with AI products that restructure markets by giving end users intelligence once controlled by expensive institutions. Molly noted that Wilson used the phrase “Obliterate, Don’t Automate” in 2014; AI supplies a radically stronger toolkit for that longstanding thesis.
Mignano’s healthcare example is Doctronic: not software sold merely to make doctors more efficient, but “a doctor in your pocket.” It can write prescriptions in Utah, though the speakers cautioned that controlled substances were probably excluded, illustrating both the ambition and the still-evolving boundary.
Wilson made the thesis literal: “Don’t hire a lawyer, use an AI lawyer. Don’t hire an accountant, use an AI accountant.” He said 20-30% of USV’s legal work no longer touches a human lawyer; routine term sheets can be generated from the firm’s prior documents without outside counsel.
Incumbents struggle when matching the new product destroys their present economics. A law firm cannot readily give away work that supports its employees; Google is a partial counterexample because YouTube, Waymo, Gmail, and Android give it room to cannibalize search with Gemini. The harder question, Kaden added, is not only why incumbents cannot respond, but “why can’t anyone do it?”
10. Trust and security could become consumer-facing AI categories
Kaden’s simplest security warning is password reset: an agent with access to a user’s email may automatically possess the recovery path to every other account. That shifts cybersecurity from a primarily enterprise concern toward ordinary consumers surrendering sensitive data and authority to agents.
Kaden pointed to portfolio company Common Fabric, a sandbox intended to reconcile “the security that they want” with “the power that they need.” She added that privacy products previously lacked sufficient consumer incentive, but on-device and private models, biometrics, and proof of humanity may now receive a larger market opening.
Mignano said he and Molly had just seen a company running local models on phones: “We’re there.”
Crypto supplies the cautionary analogy. Wilson has held crypto for 15 years, been hacked four times, and says he still retains about 99%; Molly spent roughly 15 minutes speaking with a scammer before her wife stopped her. Coinbase’s advantage was making a dangerous market feel normal through cold storage, two-factor authentication, delayed withdrawals, and trust.
11. USV needs expanding markets, not trillion-dollar companies
Wilson distinguishes company outcomes from market potential. USV will underwrite robotics, sensor proliferation, or nuclear energy based on what those markets might become in 2030, 2040, or 2050; evaluating Coinbase solely against crypto spending at investment would have missed the market expansion that followed.
Wilson believes value now aggregates to the leading one or two players faster, leaving less interest in the long tail; a third- or fourth-place company that previously could have accumulated meaningful value may no longer fit USV’s return case.
Wilson rejected trillion-dollar underwriting as unnecessary for USV’s ownership and fund size. Single-digit-billion-dollar winners can return the fund handsomely, while many of its historic successes reached around $10 billion; $50 billion and $100 billion outcomes exist, but they are outliers rather than the base requirement.
Valuations will still overshoot. The hosts pointed to ZIRP-era companies once marked at $11 billion or $15 billion and later acquired near $1 billion or $2 billion; Wilson called that perfect territory for Bending Spoons, which can acquire companies after “the venture journey’s over” and grow them with a different capital structure.