Lucas Swisher: The 20 Companies That Own the Future

Lucas Swisher co-leads Coatue’s growth fund and rarely does podcasts. In this conversation with Harry Stebbings on 20VC, he lays out a framework for growth-stage investing that treats nearly every traditional SaaS metric as obsolete. His core claim: we are in the middle of the largest value creation event in technology history, and the old rules of valuation, margin analysis, and portfolio construction will get you killed.

The Episode

Swisher’s portfolio includes OpenAI, Anthropic, Harvey, Canva, Deel, and Open Evidence. He previously worked under Mary Meeker at Kleiner Perkins and started his career cold-calling 50 CEOs a week as a 19-year-old analyst at Insight Partners. The conversation moves through the SaaS valuation crisis, why private markets now hold the future, how to value companies growing 10-50x per year, the death of spray-and-pray, margin dynamics in AI, whether kingmaking is real, OpenAI vs Anthropic, and lessons from two of the best investors he’s worked with.

The Terminal Value of SaaS Is Being Questioned for the First Time

For decades, SaaS companies were treated like insurance annuities: perpetual revenue streams with predictable profit pools. AI coding models from Anthropic, OpenAI, and others have shattered that assumption. Investors can now construct both a bull case and a bear case for virtually every public SaaS company.

Take a design tool. You can argue it will integrate AI to generate more value than ever. Or you can point out that people just create designs in ChatGPT now. When that level of ambiguity exists for the entire sector, most investors simply walk away.

“For the first time ever with this AI wave, people are questioning the terminal value of SaaS.”

Swisher identifies two compounding dynamics. First, stock-based compensation accounting advantages disappear when terminal value is uncertain. Second, investors can’t differentiate which companies will be affected. The result is a sector-wide exodus. The leading indicators he watches: sequential revenue growth, net new ARR trajectory, and retention dynamics. But with earnings data being retroactive, visibility for the next 3-9 months is extremely limited.

Harry raises the public-private arbitrage question. Monday.com trades at 1.5x revenue. Wix at 2.5x with a $4.5B market cap on $2B revenue. Surely that’s better risk-adjusted than the $10B private rounds? Swisher’s answer is surprising: things that look cheap often look cheap for a reason.

Where the Future Actually Lives

Harry shares a telling observation: when asked for his top three stocks, he named Anthropic, Revolut, and Open Evidence. None are available in public markets. A decade ago, all of them probably would have been public at this stage.

Swisher confirms the trend with data. Roughly 20 private “platform companies” exist globally, and 18 of them would have been public a decade ago. They’re massive, growing far faster than anything in public markets, and multi-product. But they choose to stay private.

For public market investors, this means being structurally excluded from the future. If you want to be levered long on the “token factory” thesis, on the idea that human inputs will become machine inputs, you have to own privates. Public markets give you some exposure, but durable 30%+ growth is almost exclusively private.

Companies stay private for good reasons, but Swisher sees limits. Three forces will eventually pull them public: real capital at scale (true liquidity, not layered SPVs), the public market as a powerful feedback mechanism (he cites Netflix’s disc-to-streaming transition as an example where public analysts provided valuable early pressure), and protection through scale. When you’re a big public company, it’s harder for competitors or adversaries to mess with you.

Valuation Comes Last

Coatue’s framework inverts the traditional order. When a company is growing 10x or 50x year-over-year, valuation is the last question they try to answer.

The math: a company at $20M ARR with a $3B post-money valuation looks insane. But if ARR goes to $200M in year one, $600M in year two, and $3B in year three, the entry point was actually cheap. The Lovable case made this concrete: at Series A, revenue was $3M. By the time legal documents closed, it was $20M. The multiple went from 70x to 10x without the price changing.

“Price does matter, but I think it matters least.”

The prerequisite for this elastic approach: the company must be in a gigantic TAM. Coatue’s internal test used to be “can this be a $10 billion public company?” That bar has moved. Now the question is whether a company can be an enduring $50-100B+ entity.

Swisher’s litmus test is deceptively simple: if the company executes well and raises at double the valuation in six months, would you want to put more money in? If the answer is yes, the current price is right. If you wouldn’t want to invest at 2x, you shouldn’t invest at 1x either.

The Power Law Is More Extreme Than You Think

Swisher cites a statistic that should shake every portfolio manager:

“20 companies have generated 80% of the enterprise value. Four companies have generated 65%.”

This is the entire private market. Twenty companies, 80% of all enterprise value. Four companies, 65%. The implication is stark: everything in venture and growth investing is about being in those companies. Nothing else comes close to mattering.

Coatue’s strategy follows directly: few investments, big checks. The spray-and-pray approach fails because you can’t afford to invest time and resources in the wrong company when the distribution of outcomes is this extreme.

Jeff Horing (Insight Partners) told Swisher that the best round is the double-down round. Get access first, then deploy more capital when the company has proven itself at a higher price. Brian Singerman (Founders Fund) articulated a complementary insight: the next double is far easier than the first. Harvey going from $6B to $12B is dramatically easier than a company going from zero to $6B.

Coatue has a counterintuitive internal dataset: as you move up market cap bands, the percentage of companies that 10x actually increases. A company at $10-100B has a better shot at delivering a 10x than one in earlier bands. This inverts the conventional wisdom that early-stage means higher multiples.

Market Size Beats Founder Quality (Almost)

Harry pushes on the tension: do you need gigantic markets more than incredible founders? Swisher’s answer is nuanced but ultimately clear: market size comes first.

A great founder in a small market with a wedge that can’t easily expand will build an incredible business. But without a massive core market and a macro trend behind it, getting to $100B is nearly impossible. The best founders aren’t the ones with grand vision at the start (Swisher and Stebbings both call “founder vision” overrated), they’re the ones who can reinvent their companies across multiple S-curves.

Databricks is the quintessential example. Ali Ghodsi took the company through at least four reinventions: ELT data transformation, model training, the center of all enterprise data, and now AI-era workloads. Most founders would have stopped at wave one.

“You unlock the next chapter through progression and continuing.”

Canva follows the same pattern. Melanie Perkins and Cliff Obrecht started with a yearbook business, transitioned online, moved to SaaS, and now operate a suite of a dozen products. Swisher notes that Cliff called Coatue about integrating AI before ChatGPT launched, showing the kind of anticipatory thinking that separates platform companies from single-product companies.

Margin Matters, But Not the Way You Think

Traditional SaaS demanded 80% gross margins. Swisher argues that applying this standard to AI companies is a category error.

The hyperscalers had terrible margins early. Snowflake and Databricks had terrible margins early. Many investors passed on those rounds because “in SaaS you have to have 80% gross margin.” Those passes were expensive.

In AI, inference costs are declining so rapidly that today’s margin profile is misleading. A company that was margin-negative two quarters ago might be at 10% today, heading toward profitability as it optimizes across frontier models for complex workloads, proprietary models for routine tasks, and cheap small models for everything else.

“Margin matters, but early it can be a misleading indicator.”

The critical reframing: AI companies will have structurally lower gross margins than SaaS (paying both cloud and LLM costs), but their operating margins may actually be higher. AI tools make engineering, sales, and legal teams dramatically more efficient, compressing OpEx in ways SaaS companies never could. The end result could be better terminal economics despite worse unit economics at the input layer.

But there’s a hard rule for low-margin AI companies. If your gross margin is low, your retention must be extraordinary. There’s no margin for error (literally). One wrong move, one churn spike, and a low-margin business has no cushion.

The Seed Investing Squeeze

Swisher is blunt: it has never been harder to be a seed investor. Two forces are compressing the economics.

First, mega-funds can destroy seed economics with overwhelming terms. Harry shares a case: his fund offered $3M on $15M pre-money. A mega-fund came in with $10M on $100M, no liquidation preference. Harry told the founders to take the deal.

Second, AI startups need more capital than SaaS startups did. They’re more capital-intensive to build, which means larger seed rounds at higher valuations, which means buying less ownership. The structural advantage is that these capital-intensive businesses may be more defensible at scale (harder for the next entrant), but the seed investor’s economics are compressed either way.

For a $3B seed fund, the math is nearly impossible: too hard to capture enough ownership in the few companies that actually generate liquidity. For $5B+ growth funds, the math works because companies stay private longer (enabling billion-dollar single-round investments) and AI-era outcomes are larger than anything SaaS produced.

Kingmaking Isn’t Real (But It Helps)

“I don’t think the kingmaking concept is a real thing.”

Swisher distinguishes between advantage and determinism. More capital is an advantage, particularly when a company has real product-market fit and needs to hire aggressively. But the idea that Coatue, Sequoia, and Kleiner all investing in a company means game over for competitors? That goes too far.

Some companies with too much capital and insufficient PMF were actually disadvantaged. Capital without product-market fit breeds complacency.

Philippe Laffont (Coatue’s founder) has a recurring framework:

“Who’s going to want to help you and who’s going to want to hurt you? Because that ultimately matters.”

Having many people invested in your success is a powerful position. But it’s not destiny. Swisher cites Snowflake and Databricks: Coatue was the only private investor in both simultaneously, and they started in completely different areas before growing into competitors. Big markets naturally produce overlapping bets.

OpenAI vs Anthropic: Two Distinct Theses

Harry puts Swisher on the spot: one dollar, OpenAI or Anthropic? Swisher doesn’t pick but offers deeply considered analysis of both.

OpenAI’s bull case rests on three pillars. An incredible consumer franchise with outstanding retention and growth. Emerging enterprise strength through Codex and coding. And a wild card: the Jony Ive acquisition creates a SpaceX-like “unknown unknown” upside vector. How do you value that hardware play in 5-10 years?

Anthropic’s bull case is more structural. Coding was a brilliant strategic focus because it’s the first AI use case to genuinely take off, and since everything in the digital world is code, dominance in coding gives Anthropic a beachhead for all enterprise analytical tasks.

But the most underappreciated advantage, in Swisher’s telling, is Anthropic’s infrastructure strategy. From day one, they architected to run on every cloud and every chip platform: Trainium, TPUs, and GPUs. In a compute-constrained world where demand outstrips supply, being able to absorb capacity that others can’t use makes them more cost-effective and gives them deployment flexibility nobody else has.

“Who’s going to want to help you and who’s going to want to hurt you? Because that ultimately matters.”

Swisher applies Laffont’s framework implicitly: Anthropic has more entities rooting for its success because it’s a customer and partner of every major cloud and chip vendor simultaneously.

The Data Prerequisite

Swisher worked for Mary Meeker at Kleiner Perkins. Her lesson: being able to express a complex company in a few lines in Excel, telling stories with data, spotting an error by glancing at cell 95. She has an extraordinary ability to lean against the grain when data tells a story others miss.

But data has limits. Thomas Laffont once told Swisher during a Databricks investment committee:

“You’re missing the forest for the trees. Just because net new ARR didn’t accelerate in a quarter doesn’t mean the trend isn’t happening.”

Swisher’s synthesis:

“Data is a prerequisite. It is not the answer.”

Mamoon Hamid showed him the other side. Swisher considers Mamoon the best Series A investor of the SaaS era: Figma, Glean, Rippling, Slack. His superpower is identifying inflection points with almost no data.

The Figma story illustrates this. Swisher prepared the data when Figma had $500K ARR and an incumbent (InVision) was the consensus winner. Mamoon looked at it for 30 seconds, saw the usage curves inside Google, Square, and Amazon, and said “We’re doing it.” Thirty seconds. Five hundred thousand dollars of ARR. He saw the kink.

Getting Off the Linear Path

Swisher’s hardest career decision was leaving Insight Partners for Kleiner Perkins. His life had been a series of linear choices: SATs, Harvard, Insight. Going solo to the West Coast as the only associate at a firm he barely knew was the first genuinely non-linear move.

“The safe path is so much less safe than you think. The risky path is actually less risky than you think.”

His biggest investment miss: passing on Anduril’s billion-dollar round. As a SaaS-focused, P&L-driven investor at the time, he looked at an ugly income statement and walked away. He missed the founding team and the defense-tech mega-trend. It’s the clearest example of “missing the forest for the trees” in his own career.

What excites him most about the next decade: the products themselves. Using Claude Code this year changed his conviction that tokens will genuinely replace human labor inputs. Twelve months ago, he wasn’t fully convinced. Now he is. The shift from assistant world to agent world happened through personal experience, not theoretical argument.

“The outcomes this generation in technology are going to be so much bigger than the outcomes from the last generation.”

A Few Observations

This conversation is most valuable not for any single insight but for the coherence of its framework. Swisher isn’t offering tips; he’s describing an integrated system for growth investing in a period when all the old systems are breaking.

• The 20-company, 80% statistic is the single most important data point. If true (and Coatue’s data is proprietary but specific), it means the vast majority of private market investing is noise. The only thing that matters is being in the twenty.
• Valuation-last is counterintuitive but internally consistent. If the power law is this extreme and exponential growth is this fast, obsessing over entry multiples is optimizing the wrong variable. The variable that matters is whether you’re in the right company at all.
• The margin reframing deserves more attention than it gets. Lower gross margin but higher operating margin due to AI-driven OpEx efficiency is a genuinely novel structural argument. If correct, it means AI companies will look worse on traditional SaaS metrics while actually being better businesses.
• Anthropic’s multi-chip strategy as competitive advantage is the most underreported insight. Most AI competition analysis focuses on model quality. Swisher focuses on infrastructure optionality: who can use capacity others can’t? In a supply-constrained world, that’s a first-order question.
• The seed squeeze is structural, not cyclical. Higher starting valuations, more capital-intensive businesses, and mega-funds willing to offer absurd terms combine to fundamentally compress seed economics. This isn’t a phase; it’s the new landscape.