Private AI Platforms Outsmart Public SaaS — Scale, Stickiness, Margins
Issue
Private AI Platforms Outsmart Public SaaS — Scale, Stickiness, Margins
Distilling Deep AI Insights from Leading Podcasts
![Cover art for Invest Like the Best with Patrick O'Shaughnessy: Dan Sundheim - The Art of Public and Private Market Investing - [Invest Like the Best, EP.460]](/_next/image?url=https%3A%2F%2Fmegaphone.imgix.net%2Fpodcasts%2F6935741a-1133-11f1-8658-83eef776ecd3%2Fimage%2F6ba553323d1caefe2f648a5990d7bb69.jpg%3Fixlib%3Drails-4.3.1%26max-w%3D3000%26max-h%3D3000%26fit%3Dcrop%26auto%3Dformat%2Ccompress&w=256&q=75)
Dan Sundheim - The Art of Public and Private Market Investing - [Invest Like the Best, EP.460]
Invest Like the Best with Patrick O'Shaughnessy by Colossus · 75 min
Listen to episode →Dan Sundheim walks through how he thinks about investing across public and private markets, the differentiated economics of large LLM businesses, and firm-level lessons from adversity. He contrasts current private-market opportunity sets with highly competitive public markets and explains why late-stage private deals can be more attractive today. On AI, he frames major LLM platforms as capital-intensive businesses with durable moats that blend characteristics of Netflix (front-loaded infrastructure) and Spotify (personalization and stickiness). He discusses implications for hyperscalers, portfolio construction changes after the GameStop episode, and why rigorous public research can both move markets and launch careers.
Key Takeaways
- 01
Private markets today can offer less competitive, high-quality late-stage opportunities compared with public markets.
Sundheim argues fewer people thoroughly vet each private deal and capital isn't as concentrated on single opportunities, creating windows where disciplined investors can find attractive risk/return. This matters for allocation choice: sophisticated investors might tilt to late-stage private exposure when public markets are overcrowded, but success requires deal access and strong diligence capabilities.
- 02
Large LLM/API businesses are capital-intensive, differentiated, and unlikely to be fully commoditized.
Training and operating state-of-the-art models require enormous upfront spend and specialized talent, so scale and differentiation matter; Sundheim expects only a handful (4–5) long-term leaders. The implication for investors is to evaluate scaling laws, monetization paths, and whether customer lock-in (data, personalization) justifies the initial capital outlay—technologies referenced include LLMs, model training infrastructure, and inference APIs.
- 03
Think of major LLM firms as a hybrid of Netflix (fixed, front-loaded investment) and Spotify (personalization-driven stickiness).
This analogy emphasizes two value drivers: heavy, early-capex on models/infrastructure and the long-term value from personalized experiences that increase retention and monetization. For founders and investors, it highlights the importance of early focus (consumer vs enterprise), data capture for personalization, and designing pricing/usage models that turn scale into durable economics.
- 04
Hyperscalers will benefit near-term from AI demand but face longer-term margin and concentration risks.
Sundheim expects AWS/Azure/GCP to see rapid growth as AI workloads drive cloud demand, yet large AI players could internalize compute or negotiate leverage, pressuring margins and creating customer-concentration risk. Investors should monitor infrastructure capex cycles, customer self-provisioning trends, and the evolving split of value between cloud providers and AI-native firms.
- 05
Adversity (e.g., GameStop impact) shifts firm behavior toward resilience, clearer communication, and lower-risk portfolio construction.
The GameStop shock forced Sundheim's firm to emphasize investor communication, operational resilience, and de-risked positioning—lessons that apply broadly to investment firms managing liquidity and reputational risk. For fund managers, this underscores governance, stress-testing portfolios, and the trade-off between returns and business durability.
- 06
Publishing rigorous, public research can materially move markets and accelerate careers.
Sundheim’s anonymous Value Investors Club short on Orthodontic Centers of America both affected the stock price and opened professional doors, illustrating how well-documented public analysis creates real-world impact. For analysts and investors, disciplined, transparent research is a high-leverage activity that can establish credibility and lead to opportunities.

20VC: Inside Coatue's $7BN Growth Fund: Why Price Matters Least | Why Mega Markets are the Most Important | How Mega Funds Can Still Do 5x Returns | How to Assess Durability of Revenue and Margins in AI with Lucas Swisher
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch by The Twenty Minute VC · 66 min
Listen to episode →The episode explores how AI is reshaping public and private SaaS valuations, forcing investors to reassess terminal value for recurring-revenue businesses and to watch retention and net-new ARR as leading signals. Lucas Swisher explains Coatue's growth fund approach: price is considered last, while market size, the ability to ride multiple S-curves (platform reinvention), and founder-market fit drive investment conviction. The conversation covers how mega growth funds can still achieve venture-like returns by concentrating capital in a few durable platform winners and doubling down over time. It also examines margins in an AI era, arguing early gross margins can be misleading while long-term operating margins and efficiency at scale matter more. Finally, the guests discuss investor skill — spotting inflection points via usage and retention curves — and controversial shifts in seed economics and the so-called 'kingmaking' narrative.
Key Takeaways
- 01
Price matters, but it should be considered last when evaluating generational, high-growth companies.
The hosts argue that valuation is secondary to durable market opportunity, founder-market fit, and the ability to scale into multiple S-curves. For growth-stage investing, paying up can be justified if the business can meaningfully expand TAM or become a platform — Coatue's litmus test is whether they'd want to invest more later at a higher price.
- 02
AI is re-pricing public SaaS by changing terminal value expectations for traditional SaaS annuities.
Frontier models (e.g., OpenAI, Anthropic) and rapid shifts in inference costs create uncertainty about which SaaS products will retain pricing power and margins. Investors are de-risking by exiting exposed SaaS names, so monitoring retention, net-new ARR, and sequential revenue becomes critical to separate durable businesses from ones vulnerable to AI disruption.
- 03
Market size and the ability to ride multiple S-curves (platform reinvention) matter more than near-term margins or revenue snapshots.
Companies that can expand into adjacent products or reinvent themselves around new paradigms (like integrating AI) can compound value far beyond early revenue or margin signals. This concept elevates strategic optionality — e.g., multi-product platforms or multi-cloud/multi-chip strategies — as a key determinant of long-term outcomes.
- 04
Mega growth funds can still generate venture-like returns by concentrating capital into a few durable winners and doubling down over long private lifecycles.
Rather than broad diversification, the path to high returns at scale is concentrated bets on platform companies with potential for huge outcomes, plus the discipline to follow-on aggressively. This approach leverages the fund's size to capture larger ownership in winners and exploit asymmetric upside as companies scale.
- 05
Margins matter at scale, but early gross margins can mislead during AI-driven architecture shifts.
Token/inference costs are falling and businesses can optimize compute (own models, smaller models, better cost curves), so initial low gross margins don't doom a company if it can drive efficiency and achieve strong retention. The critical metric for low-margin, AI-native businesses is very high customer retention and the ability to improve unit economics over time.
- 06
Elite investors focus on 'force through the trees' — spotting inflection points via usage and net retention curves rather than quarterly revenue noise.
Signals like disproportionate usage from large accounts, rising net retention, and early kinks in growth trajectories indicate durable inflection points even with sparse data. Skills in reading these leading indicators (a la Mamoon Hamid) separate top investors in identifying Series A and growth inflections.
Questions answered in this issue
How do private and public markets differ today for investors?
Private markets currently have fewer people examining each opportunity and less capital concentrated on each deal, which can create attractive late-stage opportunities; public markets are more competitive with many participants and diverse approaches, making it harder to generate excess returns. The optimal view depends on cycle and investor capabilities.
What is the central investment risk for LLM companies?
The main risk is the enormous upfront capital required to train models and the uncertainty whether scaling laws will continue to deliver attractive returns on that capital; outcomes could be profitable if adoption and monetization scale, or painful if returns don't materialize quickly enough.
Why are many public SaaS companies being re-rated now?
Because frontier AI models (e.g., from OpenAI, Anthropic) are changing expectations of terminal value for SaaS annuities; investors are unsure which SaaS will be disrupted and so many choose to de-risk by exiting the sector, producing broad multiple compression. Watch retention, net-new ARR, and sequential revenue as leading indicators.
How should investors think about margins for AI-native businesses?
Margins matter ultimately, but early gross margins can be misleading because token/inference costs are falling quickly and companies will optimize (own models, use smaller models) over time; focus on terminal operating margins and the ability to drive efficiency at scale.