Anthropic's tender priced at $350B. The market said $1T.

A month after the tender closed, the Forge mark is at $1T. What that gap tells us about how private markets actually price.

‍

The Big Story

Last month, Anthropic offered employees the chance to sell shares in a $6 billion tender priced at a $350 billion valuation. Many declined. Within days of the tender closing, Anthropic disclosed that its run-rate revenue had crossed $30 billion, up from $9 billion at the end of 2025. A week after that, Bloomberg reported the company was fielding investor offers at $800 billion — primary-market interest, not secondary, from buyers operating outside the tender's curated syndicate. On Forge Global, Anthropic's implied mark has crossed $1 trillion. The employees who held are up roughly three-fold in thirty days.

The trade press has covered each of these data points as separate stories. They aren't. The structural case that late-stage tenders aren't price discovery — that they're closer to a private equity continuation fund than a clearing market — has been theoretical for a while. Anthropic just made it empirical, in thirty days, with $650 billion of margin between what the company tried to mark itself at and what the actual market is willing to pay.

Thursday Thesis: Tender offers are venture's continuation funds

The consensus on late-stage tenders is generous. Companies are staying private for an average of 13 years, up from 10, according to several recent studies. Employees and early investors need liquidity. Mega-tenders give it to them. The argument extends further: a tender priced through hundreds of insiders trading against a basket of existing and new institutional investors looks more market-grounded than a primary round priced by one new lead. It looks like a clearing event because cash actually moves and a lot of parties are involved.

This view has the form of a market without the substance of one. A tender is this: the company decides who gets to bid and who gets to sell. The lead investors writing the new checks usually already hold large positions, which means they sit on both sides of the trade economically. The price is set by negotiation between parties who all benefit, in different ways, from a higher mark. There is no anonymous bid coming in below the tender price to discipline it. There is no short. There is no public clearing market. A tender is a curated transaction at a curated price. That can be useful. Employees do get cash, the company does get a defensible mark for the next round. But the language of "liquidity" and "market price" dignifies what is structurally closer to a marking event than a clearing one.

The PE precedent

The cleaner analogy isn't anywhere in venture. It's in private equity, where the GP-led continuation vehicle market crossed $115 billion in 2025, with single-asset CVs alone setting a six-month record at $33 billion in the second half. Roughly 40% of GPs surveyed for Bain's 2026 Global Private Equity Reportsay they expect to explore a continuation vehicle in the next year or two — on top of the quarter who already have one in the books. The structural template is identical: the same firm that originally invested in the asset orchestrates a sale of that asset, often into a vehicle that the same firm continues to manage, with selling LPs taking cash and buying LPs (sometimes the same LPs, sometimes new ones the GP brings in) writing the new check. The asset doesn't trade in any clearing market. The price is what the GP and its rolodex agree it is.

Swap "GP" for "company" and "LPs" for "shareholders" and you have a tender offer. Stripe's February tender at $159 billion was led by Thrive, Coatue, and a16z, the same firms that already own large Stripe positions. Anthropic's $350 billion tender ran in lockstep with its $30 billion primary at the same price, with overlapping investor lists. SpaceX repriced from $400 billion in summer 2025 to $800 billion in December via two consecutive tenders, with the same insider syndicate on both sides. OpenAI's $6.6 billion October 2025 tender at roughly $500 billion drew on the same investor base that had been buying primaries for years. Vercel ran a $300 million tender alongside a $300 million Series F at $9.3 billion. Intercom did a $100 million tender at a $2 billion-plus valuation while explicitly noting it had no capital need. Decagon ran its first employee tender at the same $4.5 billion mark as its Series D, led by the Series D investors. The pattern across all of them: same actors, fresh marks, no asset actually leaving the orbit of the parties who set the price.

Anthropic, in three valuations

The most honest evidence that these aren't clearing prices is the Anthropic shortfall itself. If $350 billion were a market-cleared price, you would expect tender supply to roughly match demand at that price. Instead, demand exceeded supply because employees declined to sell. They were betting the IPO would clear higher than what the company's preferred investor syndicate was willing to pay. They didn't have to wait for the IPO. The market settled the question in less than a month.

The discount to anonymous secondary marks is the same signal in continuous form, and Anthropic now provides the cleanest example anyone has produced. The February primary closed at a $380 billion post-money valuation. The April tender priced off the same $350 billion pre-investment mark. Within weeks, Bloomberg reported the company was being offered $800 billion by institutional investors operating outside the tender's curated buyer set. The implied mark on Forge Global has since crossed $1 trillion. Three valuations, three groups of buyers, three different views of the same shares. The lowest of the three was the one the company picked for the tender. Stripe's $159 billion tender mark sits roughly in line with its Forge and Hiive prints, which is the more typical case; most tender marks are directionally right. The information lives in the spread. When tender marks sit well above where the same shares trade in anonymous markets, the company's preferred number is doing work that the market disagrees with. When they sit well below, employees pull back. Both gaps are tells. The tender mark is one data point. It is not the price.

Are 409As broken?

There's a third number running alongside tender marks and secondary marks, and it explains why the $350 billion looked coherent inside Anthropic's own books. The 409A valuation. Required by the IRS for any private company that issues stock options, the 409A sets the fair market value of common stock so that option strike prices don't trigger immediate tax events for employees. Independent appraisers calculate it using a basket of methods: the company's primary round price, comparable public companies, discounted cash flow models, recent transactions in the company's shares. They produce a defensible number for the board to adopt. They redo it annually, or sooner if a "material event" forces a refresh.

Here's what 409A methodology typically excludes: anonymous secondary market trades. The reasoning is procedural rather than absurd. Anonymous trades are minority positions with limited information. They're often illiquid, restricted, or otherwise non-arm's length. And they aren't curated by the company's existing investor base, which the 409A process implicitly treats as the group whose views are qualified to set fair value. So the same mechanism that supplies the company with a defensible number for tax purposes is, by construction, deaf to the $1 trillion print on Forge.

This is why the tender mark looks coherent from inside the company even when it's $650 billion below the secondary mark. The 409A process, the tender pricing process, and the primary round all draw from the same pool of approved data points. The most important of those data points is the primary round itself, which is set by the same investor syndicate that runs the tender. Three different mechanisms ostensibly produce three independent estimates of fair value. In practice, they're one mechanism producing one number.

Are 409As broken? Not for their intended purpose, which is giving the IRS a number that survives an audit. The harder question is whether that number tells anyone outside the IRS anything about what the asset is actually worth. The answer is mostly no, and the people who hold the most pre-IPO equity already know it. But the marks get used as if they did, in fund reports, board materials, employee comp planning, audit financials, and every other place where a private valuation has to look like a public one. 409As and tender marks aren't broken. They're irrelevant for the thing the language around them keeps implying they're doing.

For pre-IPO holders: the practical implication is to stop treating tender marks as marks-to-market. They're marks-to-the-table the company decided to set. A position that looks fully marked at the tender price may be worth more or less than that depending on which other tables you can access. For employees: the question isn't whether the tender price is fair; it's whether it's the only liquidity available. Often it is, and the discount to a hypothetical anonymous bid is the price you pay for the option to actually transact. That cost is real. For the broader ecosystem: the more $100 billion-plus companies that reprice via tender, the more institutional holders mistake those marks for prices. Mutual funds write up positions, endowments report performance, none of it has been disciplined by a market that could disagree. PE has been living with this dynamic in CVs for several years. Venture is now learning the same trick.

The strongest counter is simple and partly correct. Tender offers aren't actually like CVs in one important respect: cash does leave the building. New investors put in fresh capital, employees get actual money, the transaction is real. The investor base writing tender checks — Thrive, Coatue, a16z, Iconiq, Sequoia — has plenty of skin in the game and no incentive to overpay. There's a market discipline of sorts; just not a public one. Fair. But the counter doesn't address the architectural problem, which is that the company curates the buyer set, the lead investor sits on both sides, and the shareholders who think the price is wrong have no recourse. They can't accept a different bid because there isn't one. A real market is one where any buyer can bid against any seller. A tender, by construction, isn't.

What makes the Anthropic shortfall so useful is that it briefly turned the tender into a one-sided market — employees said no — and the company's preferred number couldn't find clearing volume. That's information. The thirty days since proved those employees right faster than any IPO ever could have. Investor offers at $800 billion. Secondary trades clearing toward $1 trillion. Run-rate revenue tripling. None of it disciplined the tender mark in real time, because tender marks aren't disciplined by anything except the syndicate that sets them. The structural point is what the shortfall reveals about what tenders are and aren't, and why the gap between tender marks and market prices is the most important number nobody publishes.

📈 Data Point of the Day

$240 billion

The implied 2026 run rate for non-clearing repricing transactions across PE and venture combined: $115 billion-plus in PE GP-led secondaries, with another $50 billion-plus in venture late-stage tenders just from Stripe, Anthropic, OpenAI, and SpaceX activity, plus a long tail of smaller names. For comparison, U.S. IPO proceeds totaled roughly $47 billion in 2025, per EY's market data. The repricing market, the one where assets are marked rather than sold, is now running at five times the pace of the actual exit market.

🎓 Manual

GP-Led Secondary

A transaction in private equity where a fund's general partner, the same firm that originally invested, orchestrates a sale of one or more portfolio assets out of the existing fund and into a new vehicle, called a continuation fund. Selling LPs receive cash; new (or recommitting) LPs get exposure to assets the GP wants to keep managing past the original fund's term. Because the same firm sits on both sides of the deal, GP-led secondaries have become simultaneously the fastest-growing and most-scrutinized corner of the secondaries market.

‍