Hi, I'm Charles Costa, researcher, strategist, and founder of Lexora Labs®. Each week, I dig into research on AI, knowledge management, and the future of work so you don't have to.
EDITOR'S NOTE
Going public used to mean one main path: file an S-1, roadshow, list. AI and related tech companies now have a menu. They can list with founders still holding the vote, go through a SPAC, or stay private for years while employees cash out through tender offers. The window for a large IPO looks thin, and investors are pushing back on valuations and on deals that mostly cash out insiders.
The question running through this issue is who gains on each route, and what public shareholders are actually buying when they show up late to a story that has already been priced, financed, and partially harvested in private markets.
THIS WEEK'S READING:
Why It Matters: Ashley Capoot reports that Anthropic's planned listing would leave a founder LLC with 50.1% of the votes, while a Long-Term Benefit Trust elects a majority of the board through Class T shares. The company would be a Nasdaq "controlled company," in the same family as Elon Musk's roughly 85% voting control at SpaceX or Mark Zuckerberg's roughly 60% at Meta.
Dual-class structures are not shocking after Google and Meta. What is significant here is how locked the control is. Public buyers are taking a leap of faith: an economic bet on management over the long run, not a seat at the table.
If the vote stays private, the next question is what the filing itself reveals about the bet.
Burry says Anthropic S-1 shows AI is 'catastrophically vulnerable' | Business Insider
Why It Matters: Theron Mohamed covers Michael Burry's reading of Anthropic's S-1. Burry flags enormous compute purchase commitments relative to run-rate revenue, hyperscalers sitting as lenders, vendors, and shareholders at once, structured financing involving Broadcom, Apollo, and Blackstone, and quarter-over-quarter growth that has slowed sharply from earlier rates. He draws a parallel to vendor financing in the dot-com era. This is investor commentary on a filing, not peer-reviewed research, and the headline numbers are Burry's interpretation of the document.
Circular financing matters. So do the other headwinds: companies adjusting depreciation, and datacenters that cannot come online because of utility constraints. Buyers should ask how growth actually gets delivered with those brakes on, not just trust the S-1 narrative.
One way to delay that public reckoning is to never list until a lot of the upside is already spoken for.
Why It Matters: Marina Temkin describes how tender offers at companies like Clay, Linear, and ElevenLabs now function less as founder windfalls and more as retention tools. Employees can cash out while the company stays private. Ken Sawyer of Saints Capital argues that this enables firms to stay private longer and reduces liquidity for venture investors. Clay ran a tender at a $5 billion valuation; ElevenLabs sold about $100 million of stock at a $6.6 billion valuation.
Looked at holistically, that pattern means something simple. Tenders keep talent and delay the listing. By the time the IPO lands, a lot of the upside is already harvested privately, leaving the public holding more of the bag.
A faster on-ramp does not automatically fix that problem.
AI, Quantum Computing Feed SPAC Revival | Morningstar
Why It Matters: Michael Bodley covers a revival in SPACs as a faster route that bypasses parts of the traditional IPO process, especially for AI and quantum names. De-SPAC results still look weak: Kodiak AI went public via SPAC at about $2.5 billion and later traded near $1.4 billion. In the first quarter of 2026, 62 SPAC IPOs raised $13.2 billion, while PitchBook counted only 48 venture-backed IPOs in all of 2025.
I am always skeptical of SPACs, especially in AI. The bag problem does not go away because the listing was faster. Ultimately investors still need to do diligence and make sure they are focusing on the right things, not the structure that got the company public.
Even the label on the prospectus has its own pricing history.
Artificial intelligence disclosures and IPO underpricing | International Review of Economics & Finance
Why It Matters: Maher Kooli and colleagues study 1,138 U.S. IPOs from 2013 to 2024 and find that mentioning AI in an S-1 filing was associated with higher first-day returns. The effect concentrated among non-tech firms when investor sentiment was hot, and the premium fell sharply after ChatGPT made the label less novel. The sample ends in 2024, so it does not include the current mega-deal wave, and the authors say they lack definitive evidence that investors "learned" rather than simply grew familiar with the buzzword.
The paper shows that AI was a hot factor in how new issues got picked. Whether that stays is an open question. Looking at the Renaissance IPO ETF, which more than doubled in the COVID window and has since been far less dramatic, some companies can keep delivering value after the initial hype. The job is still picking on fundamentals, not on ticker buzz.
Taken together, the five pieces describe one pattern. Founders can keep the vote, filings can paper over headwinds, secondaries can harvest upside before the listing, SPACs can speed the clock without fixing diligence, and an AI label can buy a first-day pop that fades once the word is ordinary. The public market is not always the first chance to get rich off the company. Often it is the first chance for outside buyers to inherit the hard parts of the story.
Weigh the fundamentals against the vision. Price what the company cannot control, not just what it can build.
Know someone still treating the IPO as the finish line instead of a transfer of risk? Forward this to them.
Have a take on any of this? Reply directly. I read every response.
Charles