Lucas Swisher

The discussion opens with a direct diagnosis of the public SaaS selloff. For the first time, investors are questioning the terminal value of SaaS businesses—previously treated as annuity-like revenue streams with durable profit pools. The emergence of capable coding models from Anthropic, OpenAI, and others has introduced genuine uncertainty about whether existing SaaS products will retain their customers. When terminal value is in question, the accounting benefits that SaaS companies historically enjoyed—favorable treatment of stock-based compensation, gap versus non-GAAP earnings adjustments—also erode. The second pressure is uncertainty about which SaaS companies are exposed. Because a credible bull and bear case can be constructed for almost every public SaaS name, many investors are simply exiting the sector rather than attempting to distinguish winners from losers. The leading indicators worth monitoring, the discussion suggests, are sequential revenue growth, net new ARR trajectory, and retention dynamics—but even these are lagging signals in a market moving as fast as the current one.

A central argument is that the most important companies of this cycle—OpenAI, Anthropic, Revolut, and a handful of others—are simply not accessible in public markets. The discussion estimates that roughly 20 companies have generated approximately 80% of enterprise value across all private companies globally, with four companies alone accounting for roughly 65%. A decade ago, most of these businesses would have been public by now. This concentration has a direct implication for fund strategy: being in the wrong companies is not recoverable at scale. The prescription is few investments, large checks, with capital concentrated in what the discussion calls "platform companies"—businesses that have demonstrated the ability to hop multiple S-curves, launch multiple products, and expand into adjacent TAMs over time. Databricks is cited repeatedly as the archetype: moving from ELT data transformation to model training to enterprise data infrastructure across successive architectural waves.

On valuation methodology, the framework is explicit: when a company is growing 10x or 50x year-over-year, valuation is evaluated last. A Series C investment at $3 billion post on $20 million ARR can look cheap in retrospect if ARR reaches $200 million within a year and $3 billion within three. The internal test is not whether the current multiple is defensible, but whether, after a 3x return, a subsequent investor—including the fund's own public markets team—would want to buy the stock at that higher price. Every investment is evaluated against the question: will my public market counterpart want to own this over everything else in their book? The fund's baseline return target is approximately 3x net, which requires individual investments to generate 5x or 6x to offset misses. This math drives the emphasis on large TAMs: to justify a $5 billion entry on $50 million ARR, the investor must believe the company can reach $5 billion in revenue at 30% margins, which implies a $50 billion addressable market minimum.

On gross margin, the discussion pushes back against treating it as a gating criterion at early stages. Infrastructure businesses during architecture shifts—the hyperscalers, Snowflake, Databricks—all had poor gross margins early. In AI, inference costs are falling rapidly enough that a 10% gross margin today may have been negative two quarters ago. The more important terminal metric is operating margin, which may actually improve in AI-native businesses because AI tooling reduces headcount requirements in engineering, sales, and legal. The treatment of data follows a similar logic. The framing offered is that data is a prerequisite, not the answer. Metrics like net new ARR are described as guideposts that can cause investors to miss the larger directional trend. The specific example given is an early Databricks IC where a senior partner pushed back on over-indexing to a single quarter's ARR deceleration when the broader enterprise adoption curve was clearly inflecting upward. For low-margin AI businesses specifically, the non-negotiable data signal is retention: without high retention, a low-margin business has no margin for error.

The discussion explicitly rejects the kingmaking thesis—the idea that tier-one investors crowding into a single company makes competition impossible. Capital concentration is described as an advantage, particularly when combined with genuine product-market fit and an active land-grab opportunity, but not a determinative one. The counterexample offered is that excess capital without product-market fit historically creates complacency rather than competitive moats. On fund size, the argument is that $5 billion growth funds are viable today in ways they were not in the SaaS era, for two reasons: companies are staying private longer and getting larger while private, enabling billion-dollar check sizes with 10x return potential; and AI is addressing labor markets large enough to support outcomes well above the ceiling of the SaaS wave. Vertical SaaS, by contrast, is described as structurally unsuitable for large fund deployment given constrained TAM and AI disruption risk, though still capable of generating strong returns for smaller vehicles. --- Key takeaways: - **Terminal value uncertainty, not near-term metrics, is driving the public SaaS selloff**; investors cannot distinguish which companies are structurally impaired, so many are exiting the sector entirely rather than stock-picking. - **The most valuable private companies are staying private deliberately**, creating a structural advantage for growth funds that can deploy large checks into platform companies inaccessible to public market investors. - **Valuation is evaluated last**, not first; the operative question is whether a subsequent investor—including the fund's own public team—would want to own the stock after a 3x return, which requires believing in a TAM large enough to support continued compounding. - **Gross margin is a misleading early indicator during architecture shifts**; the more important signal for AI-native businesses is retention, which determines whether a low-margin business has any resilience. - **The "few investments, large checks" discipline is non-negotiable at scale**; the math of large growth funds only works if capital is concentrated in the small number of platform companies that generate disproportionate enterprise value, making spray-and-pray strategies structurally incompatible with the fund model.