![\"\"\/](\"https:\/\/www.europesays.com\/ai\/wp-content\/uploads\/2026\/07\/0f0f8199a2858cfd_1785166903_inline_2.jpg\")
<\/p>\nIn early 2024, Matt Murphy walked into a meeting at Anthropic \u2014 a pre-revenue AI lab valued north of $4 billion, run by a team that hadn’t yet launched a commercial model. The deal was structurally awkward for Menlo Ventures, a firm that manages roughly $600 million per fund. It was too big, too early-stage for the valuation, and demanded a check size that made traditional venture math uncomfortable. Two weeks after a pivotal LP presentation, Murphy signed a term sheet. That $10 million starter check and the $500 million-plus SPV that followed would become the defining bet of his career.<\/p>\n
“We literally came out of that meeting and said, ‘All right, we’ve got to do this. We’ve got to figure out a way to lead the round,'” Murphy recalled, speaking on 20VC. “Two weeks later, we signed a term sheet.”<\/p>\n
The story of how Menlo got into Anthropic \u2014 and what Murphy learned from deploying more than half a billion dollars into a single AI company \u2014 is more than a case study in deal-making. It’s a window into how venture capital is fundamentally rethinking risk, ownership, and what it takes to win in an era where missing the outliers is the only mistake that matters.<\/p>\n
The Anthropic Playbook: Flexibility Over Formula<\/p>\n
Murphy’s path to Anthropic began through Anjney Midha, now a general partner at a16z, who made the introduction to co-founders Dario Amodei and Tom Brown. The company was pre-revenue, pre-model launch \u2014 an unusual target for a firm accustomed to more traditional growth metrics. The internal resistance was predictable: “Why are we doing this out of a venture fund?”<\/p>\n
Murphy credits Menlo’s senior partnership for overriding that reflex. “The easy part was the technology and the founder,” he said. “The hard part was ‘wait, why are we doing this out of a venture fund?’ Fortunately, I have partners who said, ‘Let’s just do this.'”<\/p>\n
The sequence that followed reveals the deliberate architecture behind what looked like a leap of faith. Menlo spent months building rapport with the Anthropic team after the initial check, watching revenue begin to compound following the model launch. The firm secured relationships with capital and distribution partners \u2014 including Amazon and Google, both of whom would become critical infrastructure providers. Then came the LP meeting that crystallized everything.<\/p>\n
The SPV itself was Menlo’s first, raising over $500 million from limited partners and strategic investors in a process Murphy describes as the single hardest part of the entire deal. “Never done before, over $500 million, first time,” he said. But the round was oversubscribed \u2014 and critically, it was done “in full partnership with the company,” a distinction Murphy draws sharply against later secondary-market SPVs that would draw founder criticism for operating without company consent.<\/p>\n
The nerve-wracking moments didn’t end with the SPV close. Murphy lived through the DeepSeek crash in early 2025 \u2014 “you can’t even remember it now,” he noted \u2014 and the Dow volatility moment in 2026. Both tested the thesis. Both passed.<\/p>\n
Why Frontier Models Won’t Be Commoditized<\/p>\n
The most persistent question in AI investing today is whether open-source models will eventually match frontier performance, rendering expensive proprietary models obsolete. Murphy’s answer is nuanced \u2014 and it’s backed by portfolio data that most investors don’t have.<\/p>\n
OpenRouter, a Menlo portfolio company that operates as an intelligent inference routing layer, already sees a natural floor forming: companies using roughly 50% Anthropic and 50% open-source or fine-tuned models. But the crucial insight is where each type of model gets deployed.<\/p>\n
Use CaseBest Model TypeWhyCustomer-facing AI, high-retention experiencesFrontier (Anthropic, OpenAI)Marginal improvement in retention and revenue justifies cost premiumInternal automation, cost-sensitive batch processingOpen-source or fine-tuned80% of performance at 10% of costRapid prototyping, early-stage startupsOpen-source defaultSpeed over optimization; migrate later if needed<\/p>\n
“I don’t think it goes to 96% open-source for enterprise,” Murphy said, pushing back against predictions that frontier models will be squeezed into a narrow premium niche. “What companies are seeing is if they use Anthropic, customer retention goes up, revenue goes up, engagement goes up. For certain API calls, it’s worth the premium.”<\/p>\n
This is the core of Murphy’s thesis: the performance gap between frontier and open-source models isn’t closing where it matters most. In applications where user experience directly drives business outcomes \u2014 where a 5% improvement in model quality translates to measurable gains in retention and revenue \u2014 companies will pay the premium. For everything else, open-source wins on cost.<\/p>\n
The chip-level vertical integration now underway at OpenAI, Anthropic, Meta, and others doesn’t change this calculus. Murphy frames it as an inevitable optimization for companies approaching $100 billion in revenue. “If your compute bill is big enough, you’d be stupid not to try to design a chip for your specific workload,” he said. But he adds a note of hard-won realism: “The chip business is hard. Good luck.”<\/p>\n
<\/p>\n
What Makes an AI Application Defensible<\/p>\n
The question that haunts every AI application founder \u2014 “won’t the foundation model companies just build this and crush you?” \u2014 gets a practical, rather than theoretical, answer from Murphy’s portfolio.<\/p>\n
Lovable, an AI-first no-code app builder, rocketed from zero to $300 million in annual recurring revenue in roughly a year. Its defensibility comes not from model exclusivity \u2014 the underlying AI is a commodity \u2014 but from the product experience. The company targets “99% of people who were never coders,” Murphy explained, transforming them into creators. The model doesn’t matter. The experience does.<\/p>\n
Legora, an AI platform for legal workflows, is even harder to dislodge. The sale requires onboarding lawyers and full-time equivalent staff across organizational boundaries \u2014 client, law firm, opposing counsel. “It’s not an n-squared problem, but it’s complicated,” Murphy said. Founder Max is building a platform that spans professional services broadly: tax, accounting, compliance. Every additional workflow layer and every additional human stakeholder makes the platform more entrenched.<\/p>\n
Then there’s Cursor, the canonical example of an AI application that competed directly with Anthropic’s own enterprise offering and still achieved “a pretty darn good outcome.” The lesson: owning the user experience and workflow can trump owning the model.<\/p>\n
This aligns with broader signals from the enterprise market. A wave of large U.S. employers \u2014 from Alphabet to industrial stalwarts like CSX \u2014 are rediscovering that AI doesn’t eliminate the need for human workers. After a year and a half of betting that AI agents could replace white-collar roles, companies are quietly restarting recruitment. Lattice CEO Sarah Franklin observed that many firms “thought AI agents could fully fill the human gap, but in practice, humans still have to work alongside AI.” The human-in-the-loop deployment models that companies like Legora build around aren’t transitional \u2014 they’re structural.<\/p>\n
The Barbell Strategy and the Death of Series A<\/p>\n
Menlo’s current deployment strategy is a barbell, and Murphy is blunt about why. On one end: seed-stage “tracker checks” of $100,000 to $1 million into more than 50 companies per fund. These are relationship-building instruments, not return drivers. On the other end: concentrated, later-stage bets above $10 million in ARR, where a company has already emerged as a category leader.<\/p>\n
The middle \u2014 the traditional Series A \u2014 is, in Murphy’s words, “the worst place to be today.”<\/p>\n
StageARR at EntryValuation MultipleMenlo’s ApproachPre-seed \/ Seed$0\u2013500K$10\u201350M$100K\u2013$1M tracker check, no board seatSeries A$1\u20133M$200\u2013400M (200x+ ARR)Largely avoided; too expensive, too undifferentiatedGrowth \/ Breakout$10M+High but revenue-backedAggressive lead checks via fund or SPV<\/p>\n
The tracker check strategy has a measurable multiplier effect. Murphy pegs it at roughly 10x: getting that initial wedge position makes Menlo 10 times more likely to lead or participate significantly in the next round. “Get a wedge, then pounce” is how Murphy pitches it to his own limited partners, showing them how 1% seed positions in companies like OpenRouter, Whisper, and Axiom later graduated into concentrated rounds.<\/p>\n
The ownership math that once defined venture portfolio construction has been inverted. “You’re better off being in them at a very small percent than owning a large percent of a company that exits for $300 to $500 million,” Murphy said. “Those just aren’t going to move the needle.”<\/p>\n
Harry Stebbings, the host, admitted he learned this lesson the hard way. “The single biggest mistake for me is focusing on ownership,” he said. “We passed on companies where we were offered 1% \u2014 and all of them would have returned huge amounts of money.” He named Deal, Eleven Labs, and StarCloud as deals where a 1% stake, declined over ownership concerns, would have generated outsized returns.<\/p>\n
Murphy echoed the same regret. “If I had to look back at the biggest mistake, it’s not looking at a company and saying ‘we can’t do 1% ownership.’ I’ve now seen several \u2014 ElevenLabs, StarCloud \u2014 where 1% would have returned huge money.”<\/p>\n
Relationships: The Real Moat in Venture<\/p>\n
In a market where top-tier firms now compete at every stage \u2014 Benchmark recently added a growth vehicle, joining Sequoia, a16z, Founders Fund, Thrive, and Lightspeed in the all-stage arena \u2014 Murphy argues that one competitive advantage remains durable: being first to build a genuine relationship.<\/p>\n
“The most common reason you lose a deal is being late,” he said. “You come in a month before the round, and somebody else has a year-long relationship. That’s usually a death knell.”<\/p>\n
This isn’t platitude. Murphy structures Menlo’s sourcing to be “very intentional about getting out ahead of things,” building founder relationships that often precede a fundraise by a year or more. The firm’s relatively compact size \u2014 12 partners, which Murphy calls “small and mighty” \u2014 is a structural advantage, avoiding the fragmentation that plagues larger multi-team funds.<\/p>\n
“Relationships mean so much because high trust matters,” he said. “It’s hard to establish that in a shotgun-wedding sprint.”<\/p>\n
The disappearance of stage boundaries \u2014 the old swim lanes where certain firms did seed, others did growth \u2014 is permanent, Murphy believes. Menlo’s $3 billion in total fund commitments is calibrated to stay competitive: large enough to write lead checks, small enough to maintain cultural cohesion.<\/p>\n
![\"\"\/](\"https:\/\/www.europesays.com\/ai\/wp-content\/uploads\/2026\/07\/0f0f8199a2858cfd_1785166987_inline_5.jpg\")
<\/p>\n
Where the Market Is Getting It Wrong<\/p>\n
Murphy flagged two sectors where market sentiment and reality are misaligned.<\/p>\n
The most overheated: “neo-labs” \u2014 new foundation model companies. More than 60 have launched, most pitching some variation of “we’ll build something researchy.” Only a handful, including Chai and Axiom, have clear application focus. Murphy expects a brutal consolidation in which most cannot survive as independent companies.<\/p>\n
The most underinvested: the developer tooling and infrastructure stack that was too early in the 2021\u20132023 cycle but is now essential. “Three years ago we invested in this area and nothing came out,” Murphy said. “Now these companies are really taking off because everyone needs to manage multiple models, optimize spend, and not get locked into a single chip provider.”<\/p>\n
Two Menlo investments illustrate the opportunity. OpenRouter, the inference routing marketplace, is already “insanely profitable” and on track to be a major independent company. Gimlet, an abstraction layer over chips and CUDA that obfuscates hardware specialization, addresses a need that barely existed three years ago and is now mission-critical for any company running multi-model infrastructure.<\/p>\n
Murphy expects the infrastructure layer \u2014 routing, observability, chip abstraction \u2014 to mature into a market worth more than $50 billion, producing multiple multi-billion-dollar outcomes.<\/p>\n
The Afterglow Problem and Europe’s Grit Advantage<\/p>\n
Every firm that catches a generational winner faces the same question: how do you stay hungry? Anthropic’s carry alone could approach $10 billion. Murphy credits Menlo’s “challenger mentality” \u2014 an identity forged over 11 years since he and partner Venky rebuilt the firm \u2014 with keeping the partnership focused on offense rather than wealth preservation.<\/p>\n
He also makes a counterintuitive argument: financial success makes investors better.<\/p>\n
“Richer investors are better investors because they’re not worrying about LP re-ups,” Murphy said. “They focus on ‘what happens if this works,’ not ‘what happens if this fails.'” Freed from back-to-back fund anxiety and downside mitigation, successful investors can take swings that others can’t afford to.<\/p>\n
On geography, Murphy sees San Francisco’s AI renaissance as real \u2014 talent concentration in the city is 10 to 100 times better for context and serendipity than anywhere else. But European founders, including the teams behind portfolio companies Lovable and Legora, benefit from what he calls “hard mode”: lower density forces more grit. Menlo won’t open a London office, but Murphy plans to spend significantly more time sourcing in Europe.<\/p>\n
His personal investing interest aligns with Menlo’s portfolio exposure to medical AI \u2014 eight model companies focused on drug discovery, including Chai, Zaira, and Vilia, plus Sword Health for healthcare delivery. Murphy, whose mother has multiple sclerosis, expects therapeutic AI breakthroughs to transform chronic disease management within the decade.<\/p>\n
The closing watchpoint is consolidation. The neo-lab proliferation of 60-plus companies will correct sharply, with only a handful surviving as independents. The infrastructure layer will mature into a massive market. And the firms that mastered the barbell \u2014 tiny seeds for relationship capital, monster checks for the outliers \u2014 will dominate the next cycle.<\/p>\n
Menlo’s Anthropic journey began with a $10 million check that didn’t fit the traditional venture mold. It ended with a $500 million SPV, a redefined firm strategy, and a thesis that the venture industry’s old math \u2014 maximize ownership, minimize round sizes, stay in your lane \u2014 has been permanently retired. In its place is something simpler and harder to execute: get into the companies that matter, by any structure necessary, at whatever ownership you can get.<\/p>\n","protected":false},"excerpt":{"rendered":"In early 2024, Matt Murphy walked into a meeting at Anthropic \u2014 a pre-revenue AI lab valued north…\n","protected":false},"author":2,"featured_media":120445,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8],"tags":[53,61011,53712,6541,3261,968,61009,22961,8573,33180,15128,157,5013,61010,61012],"class_list":["post-120444","post","type-post","status-publish","format-standard","has-post-thumbnail","category-anthropic","tag-anthropic","tag-axiom","tag-chai","tag-cursor","tag-dario-amodei","tag-elevenlabs","tag-gimlet","tag-legora","tag-lovable","tag-matt-murphy","tag-menlo-ventures","tag-openai","tag-openrouter","tag-starcloud","tag-sword-health"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/120444","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=120444"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/120444\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/120445"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=120444"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=120444"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=120444"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}