Carnegie Mellon's Miles Dieffenbach: Can Mega Funds Still Produce Venture Returns? | 20 Oct 2026 | Origins: Inside Venture Capital, Hosted by a GP and LP · Nuggets
Podcast//21 min read/youtu.be
Carnegie Mellon's Miles Dieffenbach: Can Mega Funds Still Produce Venture Returns? | 20 Oct 2026 | Origins: Inside Venture Capital, Hosted by a GP and LP
"Would it hurt for you if this fund didn't do well, would you experience the pain that we do, or are you just getting a free call option on our money and if it hits you win, if it doesn't you've collected $30 million in management fees?" - Miles Dieffenbach [00:00:10]
"90% of LPs shouldn't even be in venture without real access to top decile managers... today's mega funds may not be able to produce venture style returns at all." - Nick Charles (summarizing Miles Dieffenbach's thesis) [00:01:52]
Disclaimer: Orignal content owned by or sourced from third parties. It does not represent the views of 'Nuggets' platform or it's team. AI is used extensively across this platform including for summaries. Accuracy is not guaranteed, there can be mistakes. Any info or content on this platform is not a financial, legal, or investment advice. Do your own research. Refer for complete disclosures:- Terms of Use · Full Disclaimer
"The beta of venture is going to be much worse into the future than it was in the past. The fund sizes and the asset base has increased 10-15x over the past 15 years." - Miles Dieffenbach [00:07:52]
"If you took the Qs over the last 10 years, I think it's like a 6x and it's like a 19 or 20 IRR. If you put a 2 and 20 on that, you'd have the same exact return as the S&P 500. So you strip away all the alpha..." - Miles Dieffenbach [00:16:46]
"10 to 15 years ago there was 15 or 20 firms that you could go to if you were a founder and you wanted to raise a couple million bucks from, and they basically all had $200 to $400 million funds and that was it. That was probably 95% of the capital." - Miles Dieffenbach [00:09:10]
"The innovation engine of the United States has to be the most coveted thing that any country in the world would want... The fact that you can be that okay with losing money or failing as a founder or failing as a company—that is why we have Nvidia, that's why OpenAI was created..." - Miles Dieffenbach [00:39:12]
"We're looking for great pickers... investors that can back and find incredible companies that I'd say in general they are finding before they're legible to the broader beta of the industry." - Miles Dieffenbach [00:20:22]
Speakers & Credentials
Miles Dieffenbach: Managing Director of Investments at Carnegie Mellon University (CMU). Manages CMU's $4 billion endowment, allocating roughly 25% ($1 billion) specifically to Venture Capital. Former Penn State Big Ten football team captain and NFL offensive lineman for the Pittsburgh Steelers. Transitioned to real estate brokerage at CBRE, real estate private equity, and has spent nearly a decade allocating institutional capital across global asset classes at CMU 00:01:01, 00:02:36.
Nick Charles: Podcast Host & General Partner at Asylum Ventures, an early-stage venture capital firm 00:00:53.
Beaser Clarkson (Don Beaser Clarkson): LP Co-Host & Managing Director at LGT Capital Partners, a premier global institutional LP firm 00:00:53.
1. Executive Summary
a. Summary
Miles Dieffenbach oversees Carnegie Mellon University's $4 billion endowment, allocating nearly 25% ($1 billion) to venture capital—a significantly higher exposure than peer institutional endowments due to CMU's strategic academic advantage in AI and robotics 00:01:05, 00:06:48.
He asserts that ~90% of institutional LPs should not allocate to venture capital at all, as general venture beta will underperform public equities following a 10–15x expansion in venture fund sizes and total asset base over the last 15 years 00:01:52, 00:08:07.
High-fee multi-stage mega-funds ($6B+) face severe structural mathematical constraints to delivering true venture returns (underwritten to a 4x net / 6–7x gross target, ~25% IRR) because massive late-stage rounds (e.g., Anthropic's $18B valuation round requiring a 20x return just to move the needle) cannot return multi-billion dollar vehicles 00:11:19, 00:13:08.
Benchmarking private VC against the Nasdaq-100 (QQQ)—which has delivered a ~17–18% net CAGR (~35–37x net DPI on a rolling 10-year basis) since 1998—reveals that charging standard 2% management fees and 20% carry on QQQ public index returns yields the exact same net performance as the S&P 500, stripping away all net private market alpha 00:12:41, 00:16:46.
Dieffenbach highlights a tight financial circularity between private AI unicorns and public megacaps: private AI revenue/COGS flows directly into public hyperscalers (Microsoft, Amazon, Alphabet), while AI capex flows directly into semiconductor giants (Nvidia, Broadcom), making private growth untethered from public markets a financial impossibility 00:15:25.
Despite macro fee headwinds, CMU remains heavily committed to early-stage VC, believing that true early-stage investing retains a "protected moat" by backing non-consensus, unconventional ideas before they become legible to consensus capital 00:08:23, 00:37:42.
LP alignment requires rigorous scrutiny of GP skin-in-the-game (personal cash commitment) to verify that VCs experience actual downside financial pain rather than holding a risk-free call option financed by tens of millions in fixed management fees 00:00:10, 00:22:28.
b. Tactical Allocations & Explicit Positioning
Long Positions / Overweight:
True Early-Stage / Seed Venture Capital: Concentrated investments in non-consensus early-stage managers who secure high dollar-weighted entry ownership in unconventional founders prior to industry legibility 00:08:23, 00:20:22.
Value-Tilted Private Equity / Lower Middle Market Industrials: Cleanest equity portfolio hedge away from AI capex spending concentrations 00:29:22.
Fixed Income (TIPS / Treasury Bonds): Uncorrelated asset class offering primary portfolio diversification away from broad tech and AI equity concentration 00:29:40.
Short Positions / Underweight:
Multi-Stage Mega Venture Funds ($6B+ Vehicles): CMU holds zero active primary allocations to multi-stage mega-funds due to mathematical inability to generate 4x net returns under 2/20 fee structures 00:13:37, 00:26:39.
Passive Venture Beta / Generalist Fund-of-Funds: Avoided for institutional LPs lacking access to top-decile managers 00:01:52, 00:07:52.
2. Chronological Table of Contents
00:00:00 - Teaser: GP Alignment, Skin in the Game, & Management Fee Friction
00:00:40 - Show Introduction & Guest Background: From NFL to CMU Endowment
00:02:25 - Miles Dieffenbach's Career Arc: Penn State, Steelers, & CBRE
Theme 1: From NFL Lineman to Endowment Allocation: Career Arc & CMU Portfolio Dynamics
Miles Dieffenbach's path into institutional asset management originated from a severe professional setback: getting cut as an offensive lineman by the Pittsburgh Steelers after playing four NFL preseason games 00:02:49.
Weighing 310 lbs at his peak and burning through personal savings to fund a 7,000-calorie daily diet, Dieffenbach urgently entered commercial real estate at CBRE before transitioning into real estate private equity 00:03:20, 00:04:02.
Joining the Carnegie Mellon University (CMU) endowment office provided a generalist platform overseeing a global multi-asset mandate 00:04:35.
CMU's endowment stands at $4 billion, maintaining a venture capital allocation of nearly 25% ($1 billion)—notably larger than peer university endowments 00:01:05, 00:06:33.
CMU justifies this outsized venture exposure through its institutional competitive moat: global preeminence in artificial intelligence, computer science, and robotics, providing advantaged deal flow and domain expertise 00:06:48.
In professional sports, performance is absolute and transparent; in venture capital, long feedback loops and surging asset growth allow underperforming managers to hide behind legacy track records while continuing to raise massive funds 00:05:03, 00:05:47.
Theme 2: The Death of Venture Beta & Why 90% of LPs Should Exit the Asset Class
Dieffenbach contends that 90% of institutional LPs currently allocating to venture capital should completely exit the asset class unless they possess guaranteed access to top-decile managers 00:01:52.
The foundational driver of this thesis is the structural inflation of venture capital: the aggregate asset base and fund sizes have expanded 10–15x over the past 15 years, degrading broad market venture beta 00:07:52.
A decade ago, roughly 15 to 20 dominant VC firms controlled 95% of institutional venture capital, deploying fund sizes strictly constrained between $200 million and $400 million 00:09:10.
Today, capital availability has exploded into thousands of multi-stage mega-funds, growth equity platforms, and micro-VCs, creating hyper-competition and compressing forward risk-adjusted returns 00:09:39.
Average venture capital returns are historically abysmal; investors without access to the top 1–5% of funds are paying high illiquidity fees to receive returns below public market benchmarks 00:18:59.
Theme 3: The Broken Math of Mega-Funds & Public Market Benchmark Comparisons (QQQ vs. 2/20 Fees)
Institutional LPs must underwrite funds against clear bogeys: CMU utilizes the Nasdaq-100 index (QQQ) as its primary Public Market Equivalent (PME) benchmark 00:12:25.
Since 1998, QQQ has compounded at a ~17–18% net annual CAGR, delivering a 35x to 37x net DPI on a rolling 10-year basis 00:12:41.
To compensate for illiquidity, operational risk, and 10-to-15-year lockups, CMU requires venture funds to deliver a 4x net DPI bogey (requiring a 6x to 7x gross fund return, or mid-20s net IRR) 00:13:08.
Multi-stage mega-funds ($6 billion+ funds) face impossible mathematical friction to hit a 4x net return:
Anthropic's $18 billion valuation round was entered almost exclusively by Spark Capital at earlier stages, while subsequent mega-fund participation occurred at valuations where even a 20x return on price-per-share ($360B EV) yields only ~1x net return to a $6B mega-vehicle 00:11:19.
Expecting multiple trillion-dollar outcomes per fund to return 4x net on $6 billion in pooled capital is mathematically implausible 00:12:01.
Highlighting fee drag, Dieffenbach notes that if QQQ's historical 6x 10-year return (~19–20% IRR) were subjected to standard 2/20 venture fees, its net returns would decline to match the S&P 500, proving that mega-fund fee structures absorb all generated alpha 00:16:46.
Sovereign wealth funds (e.g., Saudi Arabia's PIF) subsidize mega-fund fee models because writing $400 million checks grants them direct co-investment rights into mega-unicorns, a dynamic irrelevant to mid-sized endowments 00:19:15.
Theme 4: Evaluating GP Alignment: Skin in the Game, Missionaries vs. Mercenaries, & Turnover Signals
CMU evaluates VC managers through "great picker" criteria and rigorous alignment checks 00:20:22.
A core diagnostic is GP skin in the game (personal cash commitment to the fund):
Venture VCs frequently exhibit aversion to committing substantial personal net worth compared to private equity GPs 00:22:28.
While adjusting for age and baseline wealth, CMU demands that GPs experience genuine financial pain if a fund underperforms, preventing VCs from holding a "free call option" financed by $30 million in guaranteed management fees 00:00:10, 00:23:01.
CMU differentiates between "missionaries" (investors dedicated to early-stage venture building regardless of market cycles) and "mercenaries" (opportunists chasing hype trends like 2021 crypto or current AI seed rounds) 00:21:35.
CMU conducts extensive 360-degree reference calls with past co-workers and portfolio founders to test whether a GP will endure for 15–20 years 00:24:27.
Primary catalysts for LPs rotating out of established VC firms include key person departures (when top-performing dealmakers leave) and strategy drift (when a seed firm inflates fund size and morphs into late-stage growth equity) 00:26:05.
Theme 5: AI Capex Circularity, Public Market Valuations, & Macro Portfolio Hedging
The public market and private venture AI landscapes are tightly interconnected:
The top 10 companies in the QQQ represent 52–53% of the index's weight 00:15:25.
Major private AI unicorns (OpenAI, Anthropic, Databricks, Revolut, Ramp) direct virtually all operational COGS and capex into public hyperscalers (Microsoft, AWS, Google Cloud) and semiconductor giants (Nvidia, Broadcom) 00:15:43.
Therefore, a scenario where private AI growth thrives while public tech indices collapse is mathematically impossible 00:16:20.
Equity diversification away from AI exposure is difficult:
Emerging market equities offer no diversification, as TSMC, SK Hynix, and Samsung comprise 35–40% of the MSCI Emerging Markets Index 00:28:38.
Traditional value stocks are AI-adjacent; Nvidia trades at ~18x forward earnings (and 14x '27 earnings) while Micron trades at 5x 00:28:59, 00:31:38.
CMU's primary portfolio hedges against an AI capex bubble consist of lower-middle-market industrial/services private equity and fixed income (TIPS/Treasuries) 00:29:22, 00:29:40.
Theme 6: Secondary Market Mechanics, Opportunity Fund Conflicts, & The Protected Moat of True Seed
The secondary market serves as an essential liquidity valve in venture capital:
Multi-stage mega-funds cannot easily sell secondary stakes without damaging founder relationships or returning negligible fractions (~1/10th) of multi-billion dollar vehicles 00:33:16.
Early-stage seed funds urgently require secondary liquidity strategies to return capital prior to year 10–15 lockups 00:33:59.
Average multi-stage funds take 10 years just to hit 1.0x net DPI 00:34:50.
Opportunity funds introduced by seed firms create structural conflicts of interest: VCs selling secondary shares out of Fund I while deploying fresh capital from Opportunity Fund II into the same company creates an inherent LP conflict 00:36:11.
True early-stage venture capital retains an enduring "protected moat" because backing non-consensus, unconventional ideas (e.g., Airbnb in its infancy) is structurally counter-cultural and inaccessible to consensus mega-funds 00:08:23, 00:37:42.
Ultimately, US venture capital serves as the vital innovation engine of the global economy, backing transformative tail risks like Nvidia, Google's Transformer architecture, and OpenAI 00:39:12.
The Reference Vault
4. Core Frameworks & Mental Models
The QQQ / Public Factor Benchmark Framework: A risk-adjusted framework that evaluates private VC returns against the liquid Nasdaq-100 index (QQQ). If a private strategy cannot generate net returns above QQQ's historical 17–18% net CAGR (35–37x 10-year DPI) plus an illiquidity premium, institutional capital should remain in public equities 00:12:25.
The Dollar-Weighted Entry Ownership & Sizing Model: A quantitative underwriting model measuring a fund's entry valuation, initial dollar-weighted ownership percentage, and vehicle size against required enterprise exits to determine if a 4x net return is mathematically achievable 00:10:40.
The Public-Private AI Circular Interdependence Model: A macro analytical framework demonstrating that private AI unicorn revenues and capex flow directly into public hyperscalers and chipmakers, proving that private AI valuations cannot decouple from public market software and hardware valuations 00:15:25.
The Free Call Option vs. Pain Alignment Framework: An LP diagnostic that evaluates GP commitment relative to net worth to confirm that VCs face actual downside financial loss rather than extracting risk-free management fee income 00:00:10, 00:22:57.
Missionaries vs. Mercenaries Evaluation Framework: A qualitative underwriting tool that distinguishes founders and VCs driven by lifelong conviction from opportunistic capital chasing speculative market bubbles 00:21:35.
Secondary Market Liquidity Asymmetry Framework: A structural concept showing that while seed funds can utilize secondaries to de-risk portfolios early, multi-stage funds cannot execute secondaries at scale without incurring severe founder brand damage 00:33:16.
Opportunity Fund Structural Conflict Framework: An analysis of LP agency risk when seed managers run growth vehicles, creating conflicting incentives when selling primary shares from early funds while buying new shares in later funds 00:36:11.
5. Anecdotes
The Caloric Cost of Getting Cut from the Steelers: Miles Dieffenbach recounted getting cut as an offensive lineman by the Pittsburgh Steelers after playing four preseason games. At 310 lbs and burning 7,000 calories a day, he realized that staying at playing weight without team-funded meals was rapidly draining his bank account, forcing him to take his first job in commercial real estate at CBRE 00:02:49.
Spark Capital's Early Solitary Bet on Anthropic: Dieffenbach highlighted that prior to Anthropic's massive $18 billion valuation mega-round, Spark Capital was essentially the sole venture firm with early conviction. Subsequent multi-stage VC entries occurred at valuations where even a $360 billion exit yields only a 20x return, failing to move the needle for multi-billion dollar vehicles 00:11:19.
The Golden Era of 15–20 VC Firms: Dieffenbach reflected on the industry landscape 10–15 years ago, when a founder seeking capital could approach only 15 to 20 firms. Those managers operated $200M to $400M funds representing 95% of total market capital, yielding high power-law returns compared to today's hyper-inflated mega-fund environment 00:09:10.
Sovereign Wealth Funds Trading Fees for Co-Investment Access: Dieffenbach detailed why sovereign wealth funds (e.g., Saudi Arabia's PIF) willingly pay 2/20 fees to $6B+ mega-funds. Writing $400M fund checks grants them access to equal $400M zero-fee co-investment allocations directly into mega-unicorns like Anthropic or OpenAI 00:19:15.
Public Market Skepticism on Tech Giants' Massive AI Spending: Dieffenbach noted that public equity markets are actively pricing AI risks: Broadcom shares dropped after announcing Anthropic as its largest custom ASIC customer, and Oracle stock faced pressure after taking on large OpenAI compute liabilities, demonstrating public market discipline absent in private VC 00:31:40.
The "Ghosting Seed Founder" Secondary Friction: Dieffenbach described how early-stage GPs often struggle to execute secondary sales 6–8 years post-investment because the founder, having scaled to later stages, stops responding to the seed investor's texts, blocking required transfer sign-offs 00:35:43.
Bizarre AGM Pitches and the Bizarre Beginnings of Airbnb: Dieffenbach illustrated true early-stage venture capital by citing Airbnb's origin story: pitching investors on the idea that strangers would pay to sleep on air mattresses in random people's homes sounded absurd, proving that top early outcomes stem from unconventional, non-consensus ideas 00:38:02.
6. References & Recommendations
Companies & Startups
Anthropic: AI safety and LLM research company; raised an $18B valuation round; major customer of Broadcom ASICs 00:11:19, 00:15:52, 00:31:52.
OpenAI: Creator of ChatGPT; major driver of global hyperscaler compute capex and infrastructure spending 00:15:52, 00:17:40, 00:31:47.
Databricks: Late-stage private enterprise data and AI platform whose cloud COGS flows directly to public hyperscalers 00:10:54, 00:15:43.
Nvidia: Global leader in GPU semiconductor compute; primary beneficiary of private AI capex 00:15:59, 00:28:59, 00:39:32.
Exceptional Leaders: Shake Shack: Hospitality at Scale | 16 Oct 2026 | Exceptional Leaders | Morgan Stanley
"I always felt incorrectly I felt that growing was going to be selling out. And I had to convince myself that you can also sell up. You don't have to sell out. And if you can actually use your growth to advance your culture and get better…
Historical VC Dominant Firm Count
15 - 20 Firms
Number of VC firms controlling 95% of venture capital 10–15 years ago