Growth Equity Docket Vol. 1 No. 9 — Mega-IPOs, New Exit Paths and the Price of Growth

Anthropic put a price on itself Tuesday morning: $70 to $78 a share, 825 million shares, a raise of about $60 billion, and up to $90 billion once you count the shares its chip suppliers are buying on the side. The company plans to price next Wednesday and trade next Thursday. Two weeks ago the talk was $2 trillion. The range says about $1.3 trillion. Below I explain, slowly, what the filing says, why the number came down, and what the lock-up means for the funds that own it.

Also this week: one client, LG Technology Ventures, had five portfolio companies reach an exit by five different routes in a month, which is the best picture of the market I can give you. The Fed is almost certain to raise rates in December, and the bond market is already charging for it. California’s wealth tax lost, but by six points, not twelve. And a private defense company raised money at $29 billion to buy a radar maker for $5 billion, which tells you where the growth checks are going.

It is Veterans Day. Thank you to those who served.

Anthropic: The Range

Here is what the company filed Tuesday, in plain terms.

The offer. Anthropic will sell about 801 million new shares, and existing holders will sell about 24 million more, at $70 to $78 each. At the midpoint, the company raises about $59 billion before fees. The banks can buy another 124 million shares if demand is strong, which would push the total past $68 billion. Either way it is the largest IPO ever, by a multiple.

The side deal. On the same day, #BigTech public companies will buy about $46 billion of a separate nonvoting class of stock at the IPO price. These are the companies that sell Anthropic its chips and its computing power. They are buying stock in their customer with the money their customer pays them. Those shares are not registered, they do not vote, and they are not locked up. Count it as $90 billion of new money walking in the door on one day.

The price. The filing does not state a valuation, so the number you read depends on what the reporter counted. Count only the shares that exist today and you get about $1 trillion. Count the shares that will exist after the IPO and the side deal, about 13 billion, and you get about $1.3 trillion at the midpoint, $1.4 trillion if you add every option and promise. The last private round, in May, was $965 billion. So the IPO is a step up from the last private price, and a step down from the $1.5 to $2 trillion that was being discussed in October.

Why it came down. Three reasons, and none of them is a secret. First, the market. The Nasdaq fell 3.4 percent the week after the election and chip stocks fell 12 percent in a day; you do not price the largest IPO in history into a falling tape at the top of the range. Second, the commitments. The filing shows about $900 billion of contracts to buy computing power, and buyers want to be paid for carrying that. Third, the rating. The company’s debt is rated double-B, which is below investment grade, and the filing says so. A company that borrows at junk prices does not get the equity multiple of one that borrows at Treasury prices. Price it to work, and let it trade up. That is what the bankers did.

Control. Nothing changed from the drafts I wrote about two weeks ago. One Class F share, held by the founders, carries 50.1 percent of the vote. One Class T share, held by the long-term benefit trust, elects a majority of the board. The company calls itself a controlled company on the cover. Public holders get one vote per share and no say over who runs it.

The lock-up, and why your LPs care. Here is the part for the funds. About 15.5 billion shares are locked up, meaning holders agree not to sell for a period after the IPO. The standard period is 180 days. For the founders and anyone who owns more than half a percent of the company, it is a full year. But there is an early release: about 3.6 billion shares under the 180-day lock can come free sooner, and employees and directors can sell up to 10 percent of what they hold from the first day of trading. For a venture fund, that is the difference between a mark and a distribution. A mark is a number on a page. A distribution is cash your limited partners can spend. The funds that own this company have been showing their investors a mark since 2023. Some of them will be able to send cash in the first week.

Retail. About 20 percent of the deal is reserved for individual investors, through brokerage apps, in the United States, Britain, Canada, Japan and Australia. A normal IPO reserves 5 to 10 percent. That is roughly $13 billion of stock going to people who use the product.

Demand. Bloomberg reported the book was covered, meaning orders equal to the shares on offer, within hours of launch. That is normal for a deal this size on day one and tells you nothing about the final price. Watch for a range increase next Monday or Tuesday; that would be the signal. The roadshow runs this week. Pricing is expected Wednesday, November 18, and the first trade Thursday, November 19.

Why it matters: this is the price that resets every private AI company’s last round, every venture fund’s mark, and every bank’s view of what the window will take. If it prices in the range and trades up, the ten companies behind it in line go in December and January. If it prices below the range, they wait for spring. And the lock-up terms will be copied. Every founder’s counsel will now ask for the 10 percent day-one release, and every underwriter will say Anthropic got it because it was Anthropic.

One Portfolio, One Week

Sometimes a single investor’s quarter tells the story of the market better than any index. Here is one.

LG Technology Ventures, the Silicon Valley venture arm of LG, is a client and a friend of this letter. Its chief executive, Dong-Su Kim, sent me an update this week, and I asked if I could share it. In the space of a month, five companies in its portfolio reached an exit or the door of one, by five different routes.

🔹 Anthropic, at the altar of the largest IPO in history, as described above. LG was an early corporate investor.

🔹 Forge Nano, which coats chips and battery materials one atomic layer at a time, went public on Nasdaq on October 22 through a merger with a SPAC, Archimedes Tech SPAC Partners II, at a value of about $1.3 billion. It trades as NANO. Forge Nano, Inc.

🔹 Allganize, which sells a generative AI platform to Japanese and Korean enterprises, listed on the Tokyo Stock Exchange’s Growth Market on November 4 at a value of about 24 billion yen, and closed its first day up 7 percent. Allganize

🔹 Cresta, the voice AI company, agreed to be acquired by Salesforce for about $2.7 billion. Cresta

🔹 Fiddler AI, which monitors how AI models behave in production, agreed to be acquired by CrowdStrike on November 3. Fiddler AI

Why it matters: an IPO on Nasdaq, a SPAC, a listing in Tokyo, and two sales to strategics, all from one fund in one month. That is what a working exit market looks like, and it is the first time in three years I can point to one portfolio and say so. It also says something about where the exits are. Two of the five went to a strategic buyer, one went public outside the United States, one used a SPAC, and only one, the largest, used the traditional American IPO. The window is open, but it has five doors, and a well-run corporate venture fund walks its companies through whichever one fits.

We are proud to work alongside LG Technology Ventures in the growth equity lane, where corporate investors like LG, Samsung, TDK and Hanwha have become some of the most patient and strategic capital in the Valley. If your fund had a month like this one, write to me. I would like to tell the story.

The Fed Is Going to Raise Rates

Tuesday’s inflation report was worse than expected: prices rose 0.4 percent in October, and 0.3 percent without food and energy. The futures market now puts the odds of a rate increase at the December meeting at 89 percent, and some of that is for a half-point move. A Fed governor said Tuesday that raising only a quarter point “would put our credibility at risk.” Two members already voted to raise in October. The 10-year Treasury yield closed at 4.98 percent and touched 5 percent during the day.

What that means for a deal. The interest rate on most buyout debt is set off a short-term rate that follows the Fed. A quarter-point increase on a $1 billion loan is $2.5 million a year, before the market adds its own margin for risk. And the margin is widening: a $1.3 billion buyout loan for a large corporate borrower led by a bulge bracket bank could not find enough buyers among the banks’ usual investors last week and is being moved to private credit funds, who charge more. Meanwhile the ratings agency KBRA counted a record 112 defaults among private credit borrowers in the third quarter, up from 87, and one in ten loans held by the public lending funds is marked below 90 cents on the dollar.

Why it matters: the election did not change any of this and was never going to. If you are refinancing in 2027, price it off 5 percent, not 4. If you are buying, assume your debt costs more than the model from September says, and that some of it comes from a private lender rather than a bank. And if you own a company that borrowed from private credit in 2021, read the covenants this week, because the lenders are.

Sacramento, Washington, Delaware

The wealth tax lost by six, not twelve. The AP called Prop 40 on Monday night: 53 percent no, 47 percent yes. On election night it was 56 to 44. Late-counted ballots in California lean young and lean yes on tax measures, and this year they moved the result by four points. The union behind the measure says it will return in 2028. Prop 42, which bars this kind of tax without another vote of the people, is at 52 percent yes and still not called. Why it matters: the measure lost, and Prop 42 makes a return hard, but the final margin is a reminder that a wealth tax in California is a 47 percent idea, not a fringe one. Finish the estate plan.

Washington. Democrats hold at least 54 Senate seats and about 253 House seats. Alaska goes to a ranked-choice count on November 18 and Mississippi certifies Friday; both could add a Democratic seat. Nothing in that changes the arithmetic from last week: no veto-proof majority, the filibuster intact, the agencies unchanged. Carried interest and qualified small business stock stay where they are. The one date that matters is still December 11, when government funding runs out. Congress returned Monday and has passed none of the twelve spending bills in the Senate. If funding lapses, the SEC stops reviewing IPO filings. Anthropic will have priced by then. The companies behind it may not.

Data centers. Voters decided 41 local measures on data centers last Tuesday, and 40 of them went against the developers: five outright bans, twelve partial bans, three moratoria in Maine. The incoming governor of Texas has promised a statewide moratorium until the legislature acts, and the incoming senator from Texas wants to repeal the state’s sales tax exemption for data center equipment. Why it matters: if your project has its permits, its power contract and its site, it is worth more today, because fewer competitors will get built. If it does not, the timeline just moved from months to a legislative session.

Delaware. Two updates on the cases I wrote about last month. In the Meta case, the court agreed on November 4 to move quickly and set a hearing on the shareholders’ request to block the Class C dividend for early January, before the January 19 effective date; it declined to hold a full trial before then. In Cloudflare, the judge agreed on November 3 to let the company appeal his injunction now, in the middle of the case, on the question of what “good faith” means under the new statute. The Delaware Supreme Court has not yet said whether it will hear it. Why it matters: by February we may have the state’s highest court telling us how the controller safe harbor actually works. Every board with a dual-class recap or a nonvoting share plan on its agenda should wait for that opinion before it signs anything.

And the count. Companies leaving Delaware this year now stand at 30 for Nevada and 28 for Texas, with DoorDash, Datadog and Samsara among the recent Nevada moves. Texas Pacific Land moves to Texas on November 17.The Legal Bench

A quieter week for the big companies and a busy one for the venture-backed, which is the pattern I keep pointing to.

🔹 OneTrust, the privacy software company, named Kimberly Paine Chief Legal Officer from TRG Screen. LogicGate, the compliance software company in Chicago, named Amanda Meglemre General Counsel; she was general counsel of Avalara through its take-private.

🔹 Covera Health, the AI radiology company, named Jeremy Livianu Chief Legal Officer; he ran legal at Homeward through its combination with Cityblock. DTN, the weather and energy data company, named Heather Randall Chief Legal Officer from Trintech. DataSpring, the health care data company run by CAQH, promoted Andrew Ruiz de Gamboa to Chief Legal Officer.

🔹 ARS Pharmaceuticals named John Slebir Chief Legal Officer; he was general counsel of Dynavax through its sale to Sanofi. Two companies in a row hiring the lawyer who just sold a company is not a coincidence.

🔹 Green Plains, the biofuels company, made Allan Williams General Counsel after three months as interim; he was a partner at Husch Blackwell. Dutch Bros named Kira Druyan, Starbucks’s deputy general counsel, as its next Chief Legal Officer, starting in January, as Victoria Tullett retires.

🔹 Departures. Enovix, the battery company, lost its Chief Legal Officer, Arthi Chakravarthy, at the end of October with no successor named and an interim chief executive in the seat. Indivior’s Chief Legal Officer left at the close of its merger with Supernus, one of four executives out on the same day.

Why it matters: four of this week’s hires went to companies that sell software or data to regulated industries, and two of them hired a lawyer who had just run a sale. When the first senior lawyer arrives at a company with institutional capital, a transaction usually follows within two years. If you have a move to announce, send it to me.

The Tape

🔹 Shield AI buys Echodyne, and raises $2.5 billion at $28.8 billion. Shield AI makes software that flies drones without a pilot. Echodyne makes small radars. Shield is paying $4.8 billion at closing and up to $500 million more if targets are hit, and on the same day it raised $1 billion of new equity, $1 billion of structured equity from Blackstone and $500 million of debt led by JPMorgan. Why it matters: a private company just did a public-company deal, with an earnout, a preferred instrument and a term loan, all at once. That is what the growth market looks like now. The round was not to grow; it was to buy. Expect more of this from the defense and space companies, which have the valuations to use as currency and buyers who want to be paid in it.

🔹 Marvell, week two. Marvell’s stock fell 28 percent in the week after it agreed to pay $115 billion in stock for Lumentum, and is now down about a third from before the leak. Bloomberg put the odds the deal closes at about even. The chief executive says the large holders he has met are supportive. No one has filed anything against it, and the merger proxy is not yet out. Why it matters: a stock deal with no collar lives or dies on the buyer’s share price, and this one has eighteen months to closing. Watch for a shareholder to ask for a renegotiated ratio before the proxy is filed. That is the only lever Lumentum has.

🔹 The other AI money. OpenAI took in an $11 billion first tranche at a $1.25 trillion valuation on Friday, down from the $30 billion at $1.4 trillion it had been seeking, with its IPO on hold. Zipline, the drone delivery company, raised $1.5 billion at $21.5 billion. Fal, which runs AI models for other companies, raised $1.3 billion at $16 billion. CesiumAstro raised $730 million. Path Robotics raised $300 million on Tuesday. Why it matters: the week Anthropic launched its IPO, five other companies raised more than $5 billion of private money between them. The IPO is not draining the private market. It is pricing it.

🔹 Seats are the floor; usage is the growth. A note from Joel Fishbein at Guggenheim landed Wednesday morning on the shift from charging per seat to charging for what gets used, and I went looking for the data. It is one-sided. In High Alpha’s survey of 800 software companies, the ones pricing by usage grew a median 43 percent last year and the ones pricing by seat grew 34; the ones pricing by outcome grew 65. Among public companies, the consumption names (Palantir, DigitalOcean, Cloudflare, Snowflake, Datadog, MongoDB) are growing about 35 percent and the seat names (Salesforce, Workday, Adobe, HubSpot, Zoom) about 15. Anthropic’s filing settles the question for AI: 80 percent of its first-half revenue was metered by the token, and only 20 percent was subscription. Microsoft said the same thing on its last call, with Azure up 48 percent and paid Office seats up 5. And Zendesk dropped seat limits altogether this week and now bills per resolved ticket.

Why it matters: consumption revenue grows about twice as fast as seat revenue, and it is worth less per dollar. It carries 30 points less gross margin, by Bessemer’s count, and it cannot be forecast the way a seat contract can, which is Joel’s point: the winners will be the companies that learn to guide a consumption business, not the first to price one. For a growth investor, that changes the diligence. Ask what share of revenue is metered, what the gross margin is on that share, and how the company forecasts it. Seats are not dead. They are becoming the floor, and usage is becoming the growth.

🔹 ADT’s debt. Apollo has committed $9.6 billion of debt for the ADT take-private and will try to sell it to bond and loan investors. That is the largest single-sponsor financing commitment of the year. Whether it sells, and at what price, is the best test we will get this month of what the market will lend to a buyout.

🔹 Qualcomm and SpaceX. Qualcomm will supply the chips and software for Starlink’s mobile network, tens of thousands of sites, starting in 2027. No price disclosed. Qualcomm rose 5 percent on the news; SpaceX fell 4 percent, because it is now spending that money.

🔹 Private credit. Blackstone closed its fourth European direct lending fund with about 13 billion euros of capacity. PGIM and LGT closed a $680 million private credit continuation vehicle, the second one I have seen this month. Why it matters: the continuation fund, invented to let buyout sponsors hold a company longer, has now reached credit. The lenders are keeping their loans the way the sponsors kept their companies.

🔹 The window. One bank priced last week. This week three small deals try: Rezera, Windward Bio at about $1 billion, and vVardis. Next week is Anthropic, and nothing else of size is on the calendar, because no banker wants to compete with it for attention. December is open if Anthropic trades well and the government stays funded.

🔹 The Loan Desk. October added 165,000 jobs and unemployment is 4.1 percent. CPI rose 0.4 percent for the month. The Fed holds 3.75 to 4.00 percent and the market expects 4.00 to 4.25 or higher after December 9. The S&P 500 closed Tuesday at 7,521, down 0.4 percent; the Nasdaq at 25,066; the VIX at 19.6. The 10-year closed at 4.98 percent. Oil is $86. Why it matters: a 5 percent 10-year is the number at which every leveraged model written this year stops working as written. Rerun yours.

From My Desk

🔹 CFO Executive Forum, Wednesday, November 18, hosted by Lockton, with Foley, Protiviti and First Citizens Innovation Banking, produced by Open Future Forum. The agenda is the one in this letter: what Anthropic’s filing means for every CFO with an S-1 in a drawer, and what a 5 percent 10-year does to a 2027 plan. Apply for a seat here.

🔹 GC Executive Forum lunch, Friday, December 4, Hotel Nia, Menlo Park. Approval-only, Chatham House rules. Reply for a seat.

🔹 Three pieces this week on foley.com. Hard Things: Hardware Is Hot Again. Bet on the Boring Machine, from our evening at Helm.ai on November 2, on why the investor’s edge in hardware is the machine that automates one paying job, not the humanoid. Biotech Is Showing What an Open Window Really Looks Like, on why the one sector pricing IPOs this fall is the one with the strategic buyers behind it. And the write-up of our GC lunch on governing the AI agents we build and defending against the ones we do not, with Kelly Boyd, Kelly Kriegshauser, Andrew Stephens and Eric Chiu.

🔹 In the Daily Journal on November 2, I said I am cautiously optimistic about 2027: a healthier deal market if rates and financing settle, with strong underlying demand for strategic deals even if volume does not jump. A week later the strategic deals arrived and the rates did not settle. Half right is the usual score.

Calendar

1. CFO Executive Forum, Wednesday, November 18, Lockton. Reply for a seat.
2. GC Executive Forum lunch, Friday, December 4, Hotel Nia, Menlo Park. Reply for a seat.
3. The Investors’ Christmas Mixer, December 16, Palo Alto. A private reception for venture and growth investors, corporate VCs and family offices, with Open Future Forum and RSM. Champagne, a chef flambéing pancakes, no panels, no pitches; $100 million minimum fund size, host approval required. One seat left as of this morning. Request to join.
4. CFO Holiday Mixer, December 17, 5:30 to 8 pm, downtown Palo Alto. For CFOs, heads of finance and chief accounting officers at venture-backed, PE-backed and public Bay Area companies, with Lockton, Protiviti and First Citizens Innovation Banking. Off the record; pick the sticker with your top question for 2027 and the hosts will introduce you to the people asking the same one. Approval required. Request to join.
5. JPMorgan Healthcare Conference reception, Monday, January 11, San Francisco.
6. GC Summit, wine country, early 2027. Save the date to follow.

Closing Argument

🔹 Founders: Anthropic got a 10 percent day-one release inside a 180-day lock-up. Ask for it. You may not get it, but the answer tells you how much the banks want your deal.

🔹 Funds: a mark is a number; a distribution is cash. Next Thursday, for the first time in three years, some of the AI marks become cash. Tell your LPs which of yours will.

🔹 Boards: the Fed is raising in December and the 10-year is at 5 percent. Rerun the model before the budget meeting, not after.

🔹 Sellers: Marvell is down a third and the Lumentum ratio has not moved. If you take stock, take a collar.

A question for you. Will Anthropic price above the range, in it, or below it? Write back with your number; I will publish the spread, no names, next week.

The largest IPO in history launched into a falling market and a rising rate. That is not bad timing. That is the market telling you what it is willing to pay. And one fund showed there are five doors out, not one. Build accordingly.