Episode Summary
Executive Summary: Carnegie Mellon’s Miles Diffenback argues venture LPs are not being compensated for risk, especially in large multi-stage funds and low-quality seed exposure. He lays out CMU’s endowment approach, strict diligence, and why liquidity, exits, fee alignment, and manager quality matter more than brand. He’s constructive on top-tier firms and cautious on AI, China, and venture’s current fundraising environment.
Main Topics: Miles’s cancer experience and mindset (Priority: 4/5): He describes surviving lymphoma at 26, how it changed his perspective on adversity, and why it made him resilient and less mentally shaken by later setbacks. Carnegie Mellon endowment construction (Priority: 5/5): He explains CMU’s $4B endowment structure: 85% equity, 15% fixed income, with 50% in privates and 50% in hedge funds/liquids, and a 'best athlete' allocation philosophy. Venture risk-reward is unattractive for most LPs (Priority: 5/5): He argues venture’s median and even top-quartile returns often fail to beat public-market equivalents, so most allocators should not invest unless they can access top-decile managers. Manager selection: access, picking, sourcing, and references (Priority: 5/5): He breaks down venture into sourcing, picking, winning, helping, and selling, emphasizing people, reference work, partnership quality, and alignment over brand or hype. Critique of large multi-stage funds and fees (Priority: 5/5): He says huge funds need unrealistic exit sizes to generate target returns, behave more like passive public-equity investors, and should charge much lower fees on growth strategies. Liquidity, IPOs, secondaries, and the return of public markets (Priority: 4/5): He sees the current market as liquidity-starved, expects more distributions from 2026 onward, and urges venture-backed companies to go public now that public comps are attractive again. AI, OpenAI, NVIDIA, and China risk (Priority: 4/5): He is bullish on AI’s long-term importance but warns about bubbles, capital intensity, and valuation risk. He also remains cautious on China due to geopolitical and fund-structure issues.
Key Arguments: Venture is a risk-heavy asset class, and LPs should only allocate if they can access top-decile managers; otherwise public-market equivalents are often better. CMU underwrites managers using deep reference work, partner attribution, and valuation checks on top portfolio companies rather than relying on brand names or self-reported stories. Multi-stage funds have become too large for the ownership stakes and exit sizes needed to hit venture-style net returns, making their fee structures increasingly misaligned with LP outcomes. Selling stock/cash discipline matters; CMU prefers cash distributions because distributed equity can create timing and pricing mismatch risk. The seed market is crowded and structurally difficult for small funds because average seed round sizes make meaningful ownership and diversification hard to achieve. GP-LP alignment is critical; red flags include overconfident fundraising claims, management-company secondary sales, and unclear partnership dynamics. Liquidity is the main bottleneck in venture today; better IPO markets and M&A would improve distributions, fundraising, and eventually LP re-ups. AI is real but likely to experience a bubble; companies like OpenAI may be transformative yet still carry substantial financing and execution risk. Top firms like Index, Sequoia, and Andreessen are powerful because of brand, access, and disciplined sizing, but CMU still assesses them on math and expected returns. China remains hard to underwrite because USD/RMB alignment has broken down and many best founders are choosing to raise elsewhere.
Data Points: Endowment size: $4 billion - Carnegie Mellon endowment managed by Miles Diffenback Endowment allocation: 85% equity / 15% fixed income - Top-level endowment structure Private asset allocation: 50% of portfolio - Target allocation to private assets Liquid/hedge allocation: 50% of portfolio - Remaining half allocated to hedge funds and liquid public assets Venture allocation: Just under 25% of total endowment - CMU’s overall commitment to venture globally Venture overweight vs peers: 5-10 percentage points - CMU is overweight venture relative to most endowments of similar size Median IRR for mature venture funds: ~8% net - 1998 onward, using 10-15 year mature vintages Top quartile IRR for mature venture funds: ~15% net - Same venture return set Top quartile TVPI: ~2.5x - Venture fund performance across mature vintages Top quartile DPI: ~1.8x - Venture cash-return performance Target net return for CMU venture funds: 4x net - Internal underwriting target for venture commitments Large multi-stage fund example: $7 billion - Illustrative fund size used in underwriting math Dollar-weighted ownership in example: ~5% - Average ownership across early, growth, and opportunity sleeves Implied enterprise value needed: ~$140 billion - To support the ownership profile of the $7B fund Required gross return for 4x net: ~6x gross - Approximate gross multiple needed after fees Exit value in best venture year: ~$800 billion - Used as comparison for how much exit value would be needed across large funds Public-market comparison: QQQ / NASDAQ 100 - PME benchmark for venture Fund check size range: $10 million to $80 million low end; up to $400 million to $1 billion high end - Range of commitments CMU may make depending on manager Reference calls per new fund: At least 20 - Due diligence process for manager underwriting GP references provided: 5 - References sourced directly from the GP are treated as weak evidence Endowment payout/draw: 5% typical annual draw - Standard spend from endowment to university campus Higher-ed inflation assumption: ~3% - Used to explain endowment return needs Target long-term endowment return: 8% to 10% - Needed to preserve purchasing power after draw and inflation China fund return: 20x net - Best-performing fund Miles mentioned Venture-backed IPO count since 2010s: 1,150 IPOs - Used to frame the rarity of giant exits Public market tech cash flow: ~$600 billion annual operating cash flow - Google, Meta, Amazon, and Microsoft combined Hyperscaler CapEx estimate: $1 trillion from 2024 to 2027 - AI infrastructure spending forecast mentioned
Pivotal Quotes: "My question to any new allocator or an investor is: do you think you're going to have access to top-decile managers?" — Miles Diffenback: Arguing that venture only works for LPs with elite access and selection skill "I would rather not have to believe in 800 billion of market cap IPOs and MA transactions to get a 4x net" — Miles Diffenback: Explaining why huge multi-stage funds look structurally unattractive "Now is the time. Please take your companies public." — Miles Diffenback: Calling on venture-backed companies to use improving public-market conditions for liquidity
Implications: LPs should be far more selective in venture, favoring smaller, aligned managers and demanding realistic return math. For GPs, liquidity, valuation discipline, and fee compression on large growth strategies may become unavoidable as markets normalize.