Everyone Picks the Flattering Comp

Every venture cycle has a similar blow-up pattern, and it's not just about valuations being too high. It's also about valuations being anchored to the wrong public comp.

In 2019, Casper went public calling itself "a pioneer of the Sleep Economy." It was a mattress company. The S-1 used the word "platform" dozens of times. Investors who comped it to SaaS got a 70% drawdown in eighteen months. Investors who comped it to Tempur Sealy would have been roughly right on day one.

Six years later, CoreWeave went public calling itself "the AI Hyperscaler." The pitch comp was AWS. The S-1 told a different story: Microsoft was 62% of 2024 revenue, two customers were 77%, the company carried $7.9B in debt against rapidly depreciating GPU inventory. That's a leveraged data center operator with concentrated counterparty risk — closer to Equinix or Digital Realty than to AWS. The market is still arguing about which comp wins. The stock has been volatile partly because that argument hasn't been resolved.

This is the move I want to give a name to: category laundering. A private company in a low-multiple business gets relabeled — by founders, by VCs, by bankers, by everyone with an incentive to agree — into a higher-multiple category. The relabeling is the trade. WeWork was real estate sold as software. Casper was CPG sold as tech. CoreWeave is GPU leasing sold as core infrastructure. Each cycle picks a different victim category and a different premium category, but the mechanism is similar.

Two ways the comp goes wrong

There are actually two distinct errors that compound, and people usually only talk about one of them.

The first is the timing error. Public markets during hype cycles trade at multiples that don't last. Snowflake at 100x revenue in 2021 was the comp set for every data infra Series C that year. Eighteen months later Snowflake was at 12x, but the private rounds were still anchored to the peak. The public comp normalized through repricing. The private comp had no mechanism to normalize, so it sticky-downed instead — through dilution, structured rounds, flat extensions, and time. A 2021 vintage Series C raised at 50x ARR is still trying to grow into that mark in 2026, four years later. The public-market investor who bought Snowflake at the peak got a quick, painful 80% drawdown. The late-stage private investor who bought the comparable Series C at the peak is still waiting for an exit and will probably get one through years of slow dilution rather than a clean repricing.

The second is the category error, and this is the one that gets missed. Even at sober prices, the comp set is usually wrong on the merits. Casper at 5x revenue was still mispriced because the right comp was Tempur Sealy at 1x, not Shopify at 15x. WeWork at any price was mispriced because the right comp was Regus, not Airbnb. The timing error normalizes eventually. The category error does less— it just gets exposed at IPO, sometimes catastrophically.

These compound. A company that benefits from both — laundered into a premium category and funded at the peak of that category's hype cycle — gets the worst of both worlds when reality arrives. That's most of what happened from 2020 to 2022. It's also what's setting up for whatever the AI cohort equivalent looks like when the lockup expires after they go public.

The waves all rhyme

Walk through the last fifteen years and the pattern is similar.

Sharing economy (2010-2015). Uber, Lyft, Postmates, Instacart, DoorDash. The pitch was "Amazon for transport," "marketplace network effects," "platform economics." The actual businesses were logistics and staffing companies with thin margins and structural labor cost issues. Public comp would have been XPO Logistics or a temp agency. Hype comp was Amazon. Uber's IPO at $82B and immediate drift toward profitability via raising prices and cutting driver pay — exactly what a logistics company has to do — settled the argument.

D2C (2014-2020). Casper, Warby, Allbirds, Blue Apron, Peloton. Funded like SaaS, valued like SaaS, priced like SaaS. Actual economics: CPG, with all the inventory, returns, and CAC payback problems CPG has always had. The right comp was Helen of Troy or Levi's. The wrong comp was Shopify. Almost every D2C IPO from this cohort is down 70%+ from peak.

SaaS itself (2018-2022). This one's subtler. SaaS as a category wasn't laundered — it was the destination category everyone else was being laundered into. But within SaaS, the same trick happened at the edges. Companies that were really services businesses (consulting wrapped in software) got SaaS multiples. Companies with consumption-based revenue and high COGS got pure-software multiples. 

Defense and AI today (2023-2026). Same playbook, new clothes — and in defense tech, doubly so. Anduril, Shield AI, and the new generation of defense primes comp to Palantir. The pitch is software margins, sticky government ARR, AI-native autonomy stacks. The S-1s, when they come, will describe companies with factories, physical inventory, fixed-price production contracts, and revenue tied to appropriations cycles. The unflattering comp is Lockheed or General Dynamics at 2x sales, not Palantir at 40x. 

Why everyone is complicit

Category laundering keeps working because it's a coordination game where everyone benefits from agreeing on the wrong reference class.

Founders get higher valuations and more dilution-protected rounds. VCs get markups they can show to LPs. Late-stage investors get a narrative they can sell to their LPs about why they paid the price they paid. Bankers get bigger fees on the eventual IPO. Even employees benefit, at least on paper, because their options vest against the laundered valuation.

The only people who lose are the public market investors who eventually have to buy the company at the laundered price, and the late-stage privates who bought the last private round before the laundering got exposed. Everyone earlier in the chain has already marked up, already taken secondaries, already moved on to the next vintage.

✍️ Written by Zachary and Alex