Logan Bartlett: public software companies trade at 4x revenue while private comps fetch 400x ARR — the optimism disconnect explained

Summary

Logan Bartlett — investor at Redpoint Ventures and a backer of Ramp — argues that the selloff in public software stocks is not really about deteriorating financials. Revenue and retention at most large SaaS companies remain intact. The market is discounting terminal value, not near-term cash flows.

The concern is structural. AI spending is now a larger pool of net-new dollars than all of software combined — by roughly 50%, according to Bartlett's research. Companies that fail to capture that incremental spend will see growth rates drift toward zero. A business with flat growth still has value, but not the 30x kind. Bartlett draws the analogy to newspapers, whose earnings stayed relatively stable for about five years after the internet arrived, even as their long-term value collapsed. Public investors, unable to tell which of Salesforce, ServiceNow, Snowflake, or CrowdStrike will win the AI transition, are rotating into Nvidia and Google and waiting for the picture to clarify.

The valuation framework question

The conversation turns on whether software should be valued on revenue multiples at all. For a company with $100M of EBITDA and a $3B market cap, the question isn't whether the business survives — it's whether a 30x EBITDA multiple makes sense when growth is flat and discount rates are at 6%. At zero interest rates, cash flows twenty years out were nearly equivalent to cash flows today. That math no longer holds. Bartlett suggests the SaaS bond analogy — high retention, predictable annuity streams, almost debt-like certainty — was what justified the premium. That certainty is now in question, not because churn is spiking, but because the value being created on top of existing platforms may accrue somewhere else.

Buy vs. build

Bartlett's deck includes a worked example on the economics of replacing off-the-shelf software with in-house builds. Using Slack as the illustration, he prices a 1,000-person deployment at roughly $250,000 a year. Building an equivalent in-house runs around $2M annually, before accounting for integrations, SSO, compliance controls, admin tooling, and the opportunity cost of engineering time diverted from core product. For most companies — where IT spend is under 1% of revenue — the arithmetic of saving $40K on a Slack equivalent is irrelevant. What moves the needle for a mid-size industrial or healthcare company is reducing workforce turnover from 70% to 60%, not trimming a software line item.

Private market bubble

Bartlett is more cautious about early-stage venture than the ambient optimism in the market implies. Portfolio mortality rates will be higher than historical norms, he tells his LPs. Every AI-native category — AI CMO, AI CFO, vertical SaaS with an AI layer — is crowding fast, and the category that creates value may not be the company that captures it. Model providers are the likeliest vacuum for equity value, and any individual early-stage bet in a crowded category could be rational and still return nothing. He draws on the airline analogy: aviation was transformative and still proved to be a poor investment, while business travel and airport lounges were the second-derivative opportunities that actually paid.

His best historical comparison for the current moment is the industrial revolution rather than the internet bubble or the railroad boom — not because of the technology, but because of what it implies for the labor force. No prior technological shift had the same potential to fundamentally reorganize where humans deploy themselves economically.

Diffusion lag

The capability overhang is real. Bartlett estimates a meaningful share of Americans — he puts it at roughly 25%, by his own admission a rough guess — couldn't name an AI company, and around 70% would say "that ChatGPT thing." Brett Taylor at Sierra, he notes, describes constantly building infrastructure that gets productized by model providers six months later. Closing the gap between what the models can do and what an industrial or healthcare worker is actually using takes far longer than the technology timeline would suggest.

Hiring for agency

Bartlett argues that investment banking, historically Redpoint's primary recruiting pipeline, is now a structurally worse source of talent. The model was defensible when analysts brought financial modeling skills that were genuinely hard to replicate. Claude now matches or exceeds most junior analysts on those tasks. What banking doesn't produce is agency — the ability to walk into an unstructured environment and figure out what matters. As remedial tasks get automated, agency may be the only input that retains value. Redpoint is rethinking where it hires from: entrepreneurs, project managers, people with non-linear paths. The old doors, he says, are not just closed — they're cemented shut.