Lucas Swisher on AI Growth Investing, Fund Size and the Price of Scale

Harry Stebbings asks Lucas Swisher, who co-leads Coatue’s growth fund, why public software shares have fallen and how private markets compare. Swisher says AI has made investors question the durability of future software revenue and profits. His explanation is an investor’s interpretation of market sentiment, not a verified account of why each stock moved.

He argues that uncertainty can make investors less willing to overlook stock-based compensation or differences between GAAP and adjusted earnings. He also says the same software company can have credible optimistic and pessimistic AI scenarios. As a result, he believes some investors have moved away from the sector while they wait for clearer evidence.

To evaluate incumbents, Swisher would watch sequential revenue growth, new annual recurring revenue and retention. He cautions that reported earnings look backward, so those figures may not immediately capture a fast-moving change. The suggested indicators are a framework, not proof that any particular business is safe or threatened.

Stebbings asks whether apparently cheap public companies offer better risk-adjusted opportunities than expensive private rounds. Swisher says either side can be attractive: a low price may reflect real problems, while a high price can still work if a company grows sufficiently. Both outcomes depend on future performance, entry price and liquidity.

Swisher values the liquidity of public markets but believes some fast-growing private companies offer exposure that public investors cannot directly buy. He names OpenAI, Anthropic and other businesses as examples. Coatue has invested in several of the companies discussed, including OpenAI, Anthropic, Revolut and OpenEvidence; those relationships matter to his assessment.

He describes a group of large private “platform companies” that have remained private longer and developed multiple products. His estimate of how many would have been public a decade earlier is an illustrative judgment, not a measured counterfactual. Stebbings notes that longer private lives have also expanded the opportunity for growth investors.

On the durability of AI revenue, Swisher says model leadership can change quickly and argues that earlier architecture shifts also challenged incumbents. He looks for teams that can repeatedly reinvent their products and follow new demand rather than relying solely on an existing revenue stream. That is his investing framework, not a guarantee of future retention.

He cites Databricks, where he says Coatue has invested since 2019, as a company he believes has expanded across several product waves. He credits its founder with moving beyond the original data-transformation use case. His praise is an investor’s account of a portfolio company, not an independent assessment of its future growth.

Stebbings describes investing in Lovable when reported revenue grew substantially between agreeing terms and completing documentation. Swisher says exponential growth can change the apparent multiple quickly, but the challenge is identifying whether the curve is durable. An interim financing mark or a rapidly changing revenue run rate does not itself establish a realized return.

Swisher explains that Coatue considers valuation after examining the scale of the idea, market demand and company quality. His illustrative sequence of $20 million in ARR rising to $200 million, $600 million and then $3 billion is a hypothetical path, not a forecast for a named company. Price still affects returns if that growth fails to materialize.

He says a high entry price requires conviction that a market and company can become exceptionally large. Coatue’s former illustrative test of a $10 billion public company has, in his telling, shifted toward the possibility of a much larger enduring company. The $50 billion and $100 billion figures he mentions are investment scenarios, not expected valuations for every portfolio company.

Stebbings asks whether such an investment is the best use of capital compared with alternatives. Swisher values a flexible mandate that allows the fund to invest at different stages and compare opportunities. He also wants an initial investment to leave room for a larger follow-on if the company executes; access to a follow-on is a possibility, not a guaranteed right or attractive price.

Swisher cites an internal chart suggesting a higher share of companies in some larger valuation bands went on to increase tenfold. He does not provide the sample, period, methodology or treatment of survivors, so the interview does not establish a general probability of future returns. His point is that scaling further can remain possible after a company is already large.

Asked to choose between market size and founder quality, Swisher says both matter but begins with the market. He thinks a founder’s ability to develop additional products can expand the opportunity over time. A strong founder in a narrow market may build a good company, he says, but the outcome required by his strategy can be harder to achieve.

On price, Swisher says it matters but often comes last in his assessment. There is still a point at which a price erodes expected returns enough to make him pass. He says investors who initiate rounds may influence terms, but his willingness to pay depends on a belief about company execution and the size of the ultimate market.

One qualitative test is whether, after investing at a given price and seeing the company execute, he would want to invest again at a higher price. He contrasts that with judging only the next mark-up. The test asks whether the business can remain attractive as it becomes more expensive; it does not eliminate dilution, timing or exit risk.

Swisher says Coatue prefers a small number of large investments to many small early positions followed by selective additions. He cites internal estimates that a few private companies account for a very large share of enterprise value. The figures and the universe behind them are not independently documented in the interview, so they should be treated as his illustrative concentration estimates.

Stebbings observes that some large firms hold competing companies. Swisher says this can be difficult for founders and may make winning a deal harder, particularly in directly competing early-stage businesses. At larger scale, companies that began in different markets may expand into one another’s territory, creating conflicts that investors and founders must manage.

He cites Coatue’s historical investments in both Snowflake and Databricks, saying the companies began with different emphases and later moved closer together. He notes that Snowflake went public earlier. His example shows how competition can develop after an investment, rather than proving there are no conflicts when a firm owns competing assets.

The conversation turns to mega-funds. Swisher thinks a $3 billion early-stage fund can face difficult ownership and return mathematics because it must capture enough of the relatively few very large outcomes. A smaller fund, he says, can succeed without owning every leading company. This is his assessment of fund construction, not a calculation of actual returns.

He argues that a $5 billion growth fund can work if large private rounds and eventual outcomes support concentrated positions. His simple illustration is that $1 billion invested at a tenfold gross return produces $10 billion, or twice a $5 billion fund’s size before fees, expenses, losses and other investments. The example is arithmetic, not a promised fund return.

Swisher expects AI businesses that address labor-related spending to create larger potential outcomes than traditional software alone. He says a very large fund should focus on companies capable of returning substantial capital, while acknowledging that vertical software can still produce valuable businesses and exits. His market-expansion thesis is a forecast.

Asked whether a threefold return on a late-stage deal is enough, Swisher says he looks for more upside because other investments may return less or lose money. In an equal-sized two-investment illustration, a 1x and a 5x average 3x gross, while a zero and a 6x also average 3x gross. Net fund results would additionally depend on sizing, timing, fees and expenses.

He wants to see not only a plausible path to a threefold gain but also a reason another investor might buy the company at that higher valuation. He describes consulting colleagues on Coatue’s public side about eventual demand for a stock. That is his liquidity discipline, not evidence that a buyer or exit will be available.

Swisher says secondary sales can provide liquidity before an IPO and recognizes why early-stage funds use them. Such sales may help some investors realize gains while a company remains private, although access, price and timing vary. He continues to favor businesses he believes can eventually attract substantial liquidity.

Asked about mistakes in follow-on investments, he says the firm has sometimes overestimated market size or a company’s ability to launch multiple products. He says its style has generally produced a low loss ratio, but supplies no audited figures. He treats the lesson as a reason to raise the bar for the expansion thesis.

Stebbings asks about software companies growing well but more slowly than the fastest AI businesses. Swisher says he does not know their ultimate exit paths and rejects the idea that all such companies are poor businesses. They may generate real profits, he says; they simply may not fit Coatue’s concentrated strategy and the outcomes it seeks.

On AI margins, Swisher says the same underlying business principles still apply but early gross margin can be misleading during an infrastructure shift. He expects inference costs, model choices and workload routing to create opportunities to improve margins over time. That cost curve and each company’s ability to capture the benefit are uncertain.

He agrees that AI applications may have lower gross margins than earlier software because they pay for cloud and models. Stebbings proposes that more efficient engineering, sales or legal work could reduce operating expenses and potentially support higher operating margins. This is a scenario, not a measured outcome or a guarantee that lower gross margin will be offset.

Swisher says pre-revenue companies at very high valuations are generally outside his fund’s strategy. He prefers businesses with real products, demand and evidence of growth, while recognizing that others invest earlier. His risk-reward assessment reflects Coatue’s mandate rather than a claim that every early-stage company is overvalued.

Stebbings describes losing a seed deal to an investor offering a much larger check at a much higher valuation. Swisher says larger early rounds and prices can make it harder for seed funds to buy their desired ownership. He also suggests some AI businesses need more capital than earlier software startups; the effects vary by business.

Can an investor excel at both Series A and Series D? Swisher says it is possible but hard because the skills differ: early investors may rely on repeated founder meetings, while later investors must also analyze financial statements. He says Coatue uses different specialists for public and private markets because each arena has distinct information and competitors.

The guests debate “kingmaking,” the idea that prominent backers and abundant capital can determine a winner. Swisher says capital and reputable investors can help a company hire or attract attention, but do not guarantee success. Too much money without product-market fit can also encourage waste or complacency, in his view.

He distinguishes that risk from capital deployed into a growth company with demonstrated traction and a productive use for more funding. His assessment that some fast-growing companies can earn strong returns on additional capital is an investment judgment, not evidence that every large round has positive returns.

Stebbings asks why large private companies such as Canva or Stripe would go public if secondary markets offer liquidity. Swisher expects many eventually will, citing access to capital, broader liquidity and public-market feedback. These are possible benefits and predictions; a listing also exposes a company to public scrutiny and the episode does not establish any specific listing plan.

Stebbings challenges whether Canva is truly a platform company amid AI competition. Swisher says it has moved from yearbooks to online design and a broader product suite, and praises its early AI adoption. Coatue’s investment interests, and his discussion of related AI investments, are relevant to his positive assessment; product growth claims are his account.

Swisher recalls working with Mary Meeker at Kleiner Perkins and credits her with analytical rigor and the ability to explain a company through data. He describes struggling with modeling early in his career and learning to use financial details to tell a coherent story. The lesson is his personal experience, not a rule that a spreadsheet alone can predict outcomes.

He says data is a prerequisite rather than the whole answer. In an example from Databricks, a single quarter’s new ARR did not capture the broader trend he and colleagues were evaluating. For a low-margin AI company, he especially wants to see strong retention, because weak economics and weak customer behavior leave little room for error.

Discussing Mamoon Hamid, Swisher recalls looking at early Figma customer and retention data and says Hamid recognized an inflection point. His recollection of customer names and roughly $500,000 of ARR is approximate. Swisher’s praise reflects having worked with him, and does not establish a general formula for predicting winners.

Stebbings asks him to pick among prominent fund managers. Swisher resists naming a single winner, saying different investors and strategies may suit different limited partners. His characterizations of Meeker, Hamid, Insight and other funds are personal opinions shaped by his professional relationships, not independently verified rankings.

Asked to choose between OpenAI and Anthropic, Swisher discusses the strengths he sees in both instead of declaring a winner. He praises OpenAI’s consumer distribution and enterprise opportunities, and Anthropic’s coding position and ability to deploy across cloud and chip providers. Coatue has investments in both; his comparative advantages and future applications are views from an interested investor.

He emphasizes the strategic value of partners who benefit from a company’s success, while acknowledging that supportive investors and partners do not automatically determine the outcome. The guests distinguish having an advantage from proving that any one provider will win. Competitive conditions and infrastructure arrangements can change.

Swisher says his view has shifted toward believing AI can address labor-related work and support much larger outcomes. He distinguishes his conviction from certainty about the speed of adoption. Enterprise integration and deployment remain difficult, he says, even if some AI companies have grown faster than past software businesses.

He cites publicly discussed Anthropic ARR and growth figures to illustrate that speed, comparing them with earlier cloud businesses. These are his time-specific figures and comparisons, not independently verified financial statements; ARR is an annualized recurring measure. He expects organizational change to take time despite rapid technology adoption.

Swisher recalls his first meeting with Harvey founder Winston Weinberg and says the fit between language models and legal document work was compelling to him. Coatue later became an investor, though he says its team missed the Series A. His story is a portfolio investor’s recollection, not proof that legal AI can operate without professional review.

He describes leaving Insight for Kleiner Perkins as a difficult career decision after an early role that involved extensive founder outreach. Moving to a different firm and geography meant leaving a more predictable path. His advice to younger people is to consider such changes deliberately; it is a personal career lesson, not a universal prescription.

Swisher says San Francisco can help an AI company access concentrated talent, but is not an absolute requirement. He believes competition for scarce specialists can justify exceptionally large compensation packages. Those are his views on geography and pay, not evidence that a particular package will create value.

He recalls passing on an Anduril round when he was focused on software metrics and says he failed to appreciate the team and broader opportunity. The retrospective example supports his argument that financial data alone can miss a company’s direction, but hindsight does not establish that the original decision was irrational with information available then.

Asked what excites him, Swisher points to new AI products and a future OpenAI device. He speaks as an investor in OpenAI; his enthusiasm is a personal expectation, not a product review or proof of its eventual features, availability or success.