Episode Summary
Executive Summary: Matt Murphy of Menlo Capital argues the AI market has fundamentally changed venture investing: ownership matters less than being in the best outlier companies early, pricing discipline is secondary to access, and AI winners will be defined by model quality, distribution, and capital intensity. He explains Menlo’s Anthropic, Lovable, Legora, and OpenRouter bets, and why the firm is leaning into a barbell strategy across seed and growth.
Main Topics: Anthropic: why Menlo moved fast despite valuation concerns: Murphy details how Menlo was introduced to Anthropic, why the technical edge and Dario’s leadership made the decision obvious, and why the partnership agreed to enter despite the round being too large for a typical venture allocation. Ownership, dilution, and the new venture math: The conversation repeatedly returns to the idea that huge AI outcomes make small ownership stakes more attractive than large positions in smaller exits. Murphy says the market now rewards being in the biggest outliers, even at lower ownership. SPVs and concentrated capital deployment: Murphy explains how Menlo used SPVs to expand beyond normal fund limits, especially in Anthropic, and frames SPVs as a practical tool for winning larger rounds and bringing LP capital to bear. Application-layer investing: Lovable and Legora: Murphy discusses why Menlo backed fast-growing application companies like Lovable and Legora, emphasizing founder quality, explosive revenue growth, workflow depth, and the potential to expand beyond initial verticals. Open source, model routing, and multi-model optimization: He argues open source will be useful but not fully displace leading frontier models. Instead, companies will mix proprietary and open-source models, with routing and orchestration layers like OpenRouter becoming increasingly important. Menlo’s barbell strategy and the changing stage landscape: Murphy says Series A is the hardest stage today because signal compression makes valuation jumps less meaningful. Menlo is responding by going earlier at seed and later at growth, while remaining flexible on ownership and syndication. Geography, talent, and the return of the Bay Area: The discussion closes on San Francisco regaining momentum as AI’s center of gravity, while Europe—especially Sweden and the broader DeepMind diaspora—becomes more investable due to founder grit and technical density.
Key Arguments: Frontier models like Anthropic remain hard to displace because of their performance, intelligence, and ecosystem advantages; open source is complementary, not a full substitute. In AI, the most important venture criterion is no longer ownership percentage alone; access to the biggest outlier outcomes matters more than maximizing stake in smaller companies. Firms need flexibility: rigid ownership targets or stage silos can prevent participation in generational companies. SPVs are a necessary tool when fund limits would otherwise block meaningful participation in oversized AI rounds. Application companies can be extremely defensible when they own deep workflows, cross organizational boundaries, and expand into adjacent domains. Revenue growth and distribution partnerships can be as important as model quality, especially when companies like Anthropic add cloud partners and technical channels. OpenRouter-style routing infrastructure becomes more valuable as companies optimize across price, latency, reasoning, and model quality. Series A is structurally difficult because the time from seed to A has compressed and valuation increases often outpace meaningful product or revenue signals. Menlo’s advantage comes from a small, highly aligned partnership with the ability to move across seed, growth, and thematic technical areas. The market is overhyped in some areas like Neolabs/robotics/defense, and underinvested in tooling, observability, and infrastructure around the AI stack.
Data Points: Menlo first check in Anthropic: a little over $10 million - Initial investment from Menlo’s venture fund before later SPV expansion. Anthropic follow-on SPV: $500 million+ - Menlo used an SPV to size up materially in a later Anthropic round. Menlo fund size: $600 million venture fund - Murphy explains why Anthropic’s valuation was awkward for a standard venture allocation. Anthropic valuation at entry: $4 billion+ - The investment was made when Anthropic was pre-revenue and around a $4B+ valuation. Anthropic capital efficiency: ~1/50th of the capital - Murphy says Anthropic reached comparable benchmark performance to ChatGPT while spending far less. Lovable revenue growth: from zero to ~$300 million in a year - Used to illustrate the company’s outlier trajectory. Lovable entry valuation: around $150 million - Murphy says Menlo got in when the company was around this level, after trying earlier near $30 million. Lovable earlier target valuation: around $30 million - Menlo initially tried to invest at a lower valuation before the round moved up. Legora round timing: about 6 months ago - Murphy references Menlo’s recent investment in Legora. Legora check size: sub-$50 million range - Menlo sized into the round with room to add later. AI company margin target: 60-70% gross margins - Murphy argues many AI companies can still reach strong margins despite current inference costs. Current margins in many AI companies: 20-30% gross margins - He notes many great AI businesses start with atypically low margins. Menlo seed check capacity: up to $8 million on the spot - Murphy says three partners can now write meaningful seed checks quickly. Anthology fund size: 50+ companies - Menlo’s anthology fund is used for smaller starter checks and future follow-ons. Anthology check range: $100K to $1 million - Murphy describes the smaller exploratory check sizes used in the anthology strategy. Neolabs portfolio exposure: about 7 companies - Menlo has invested in several neo-labs-focused companies.
Pivotal Quotes: "I think the foundation models, let's say specifically Anthropic, have such special models, performant, intelligent models. It's going to be hard for somebody to just kind of say, I've used open source with my data." — Matt Murphy: Explaining why leading frontier models are likely to remain hard to displace. "I think ownership today is less relevant than it ever used to be." — Matt Murphy: On how AI-era venture returns depend more on outlier access than classic ownership concentration. "Series A is the worst place to be today." — Matt Murphy: Describing stage compression and weak signal quality between seed and A.
Implications: AI venture is shifting toward flexible, high-conviction, multi-stage investing. The winners will be firms that can access outliers early, add capital aggressively, and navigate a market where models, apps, and infrastructure increasingly blend together.