Anthropic lists, and private AI marks meet a public price
What moved on our finance boards: Anthropic lists, Nvidia's margin test passes under a governance cloud, and tests on Rocket Lab and HeartFlow come due.
Until this week, none of the independent US frontier AI labs traded publicly: each was valued by the people who negotiated its last round. On November 19 Anthropic started trading, and the market now sets a price that its investors, its partners and its private rivals' backers can be measured against.
Anthropic's listing replaced the private mark with a market price
Anthropic priced its IPO on November 18 at $83 a share, above its $70 to $78 range, and sold 980 million Class A shares for $81.34 billion, according to its pricing term sheet. About $79.4 billion of that was the company's share at the offer price, $78.8 billion after the underwriting discount; the rest went to its seven co-founders, who sold shares. Google, Nvidia and five other companies agreed to buy another $26.0 billion of non-voting shares at the same price. The stock opened on Nasdaq at $111.50, closed its first day at $97.17, 17% above the IPO price, and fell to $88.91 on November 20.
At the IPO price the company was worth about $1.5 trillion fully diluted, about 55% above the $965 billion its Series H set in May. Our State of Frontier AI board keeps Anthropic at high conviction and says plainly why the listing did not raise it: the listing added money and a price, not new business results. At its first close Anthropic was valued at about 22 times its run-rate revenue on a fully diluted basis (run-rate revenue is its own measure, $81 billion at October 31), so much of the case is already in the price.
Amazon and Google, the two holders the prospectus puts above 5%, now hold stakes that a share price values every day. Anthropic's final prospectus gives Amazon 1.94 billion shares, 14.6% of the company, with notes that can convert into up to 1.55 billion more over the coming quarters as its 12% ownership cap allows, and Google 1.77 billion shares, or 13.4%. At the November 20 close those shares alone were worth about $172 billion for Amazon and $158 billion for Google (derived). Amazon carried its Anthropic investment, notes included, at $238.7 billion at September 30; counting the shares its notes can still convert into, the stake was worth about $310 billion at the November 20 close (derived).
| Holder | Shares after the IPO | Worth at the Nov 20 close | What its own books show |
|---|---|---|---|
| Amazon | 1.94B Class N shares, 14.6%; up to 1.55B more as its notes convert | ~$172.2B for the shares; ~$310B counting the note shares (derived) | Carried at $238.7B at Sep 30, notes included (derived from its 10-Q) |
| 1.77B shares, 13.4% | ~$157.8B (derived) | Alphabet's 10-Q does not name the company it holds |
Our call: Anthropic holds its $83 IPO price for two quarters. We expect it to close at or above $83 on May 19, 2027, six months after it listed and after the 180-day lock-up on employees and some investors has ended. We are wrong if it closes below $83 that day. Confidence: medium.
Nvidia passed its margin test, and a CEO vacancy hangs over the stock
Nvidia's results beat its own guide. Revenue for the quarter that ended October 25 was $115.8 billion, against a guide of $108.0 billion, with $107.6 billion from data centers, according to its November 17 results. Its non-GAAP gross margin of 74.9% beat the top of the 74.0% ±0.5-point guide. Our last post set the test: a margin at or above the midpoint keeps the pricing case intact. It held.
What comes next is lower, as Nvidia has said since August: its August 26 call put the fourth-quarter low at 71% to 72%, and its November 17 guide puts 71.0% to 72.0% in writing. Colette Kress, who is interim CEO and still its finance chief, told analysts that is the bottom, before 72% to 73% in fiscal 2028. The first sign of fading we watched for, a print below 73.5%, is what Nvidia has guided for the fourth quarter; no result has shown it yet.
Show as table
| Setup | % | Source |
|---|---|---|
| Q2 FY2027, reported Aug 26 | 75.0 | Nvidia Q2 FY2027 results |
| Q3 FY2027, reported Nov 17 | 74.9 | Nvidia Q3 FY2027 results |
| Q4 FY2027 guide, midpoint of 71.0–72.0% | 71.5 | Nvidia Q3 FY2027 outlook |
The stock has fallen since the results, in a week when chip stocks fell together. On November 20 it closed at $125.53, 23% below its November 13 close; it fell 14.0% that day, when AMD fell 13% and Broadcom 12%. The CEO vacancy sits on top of that: Kress signed the results filing as interim CEO, and the chair said on the call that the board has retained a search firm. The Wall Street Journal reported on November 17, citing people familiar with the matter, that the SEC has opened a civil probe examining in part whether Nvidia's disclosures misled investors. The SEC does not comment on whether an investigation exists, and Nvidia has not confirmed the report.
Our State of Nvidia, State of Frontier AI and State of Physical AI boards hold Nvidia at medium. The margin test passed and the business beat, so the rating stays where it is. The State of Nvidia board's notional position was already halved on November 12, when the CEO resigned, and stays halved until a permanent CEO is named.
Our call: Nvidia's fourth-quarter non-GAAP gross margin comes in at or above the low end of its 71.0% to 72.0% guide when it reports, in late February or early March. We are wrong if it prints below 71.0%. Even if we are right, the margin sits below the 73.5% line we called the first sign of fading. Confidence: medium.
The tests we set came due
Since our last post, our boards have recorded the results of tests we had set.
- Rocket Lab's margin stayed above our line, and its rocket slipped. We had set a test for its third quarter: cut both of our State of Space & Launch board's high ratings on Rocket Lab to medium if its GAAP gross margin fell below 30%. It came in at 33.6%, down from 36.1% a quarter earlier, computed from its November 9 release, so the ratings stand. Rocket Lab had moved Neutron's first launch into 2027 on October 14, and its November 9 release now says "no earlier than 2027". The board's own rule for a slip is to trim and not buy more, so it cut Rocket Lab from 8% to 6% of the book, the bottom of its range, and holds it there until Neutron flies.
- Medicare kept AI software on interim terms. Our State of AI for Healthcare board held a small short on HeartFlow into Medicare's 2027 hospital outpatient rule. The final rule, released November 4, says CMS is "finalizing the following interim payment policies". It kept HeartFlow's code in the new-technology payment class it proposed and refused a higher one; the proposed rate in that class, $850.50, was about 3% below the 2026 rate of $877.34. The short's cover trigger did not fire. But the date it was built around has passed, and HeartFlow's volume is outrunning the rate: third-quarter revenue rose 57% to $72.9 million and it raised its 2026 guide to $262 million to $264 million. The board cut the short to low conviction and kept its 2% size; its price targets, all below the $46.69 close, are under review.
- Two other names reported in line with our ratings. Oklo, which our State of Nuclear Energy board rates avoid, now targets its first powerhouse for 2028 to 2029 rather than 2028, and has sold $88.8 million of stock under its second $1 billion program, per its quarterly filing. Constellation raised the midpoint of its 2026 adjusted operating earnings guidance and narrowed it to $12.50 to $12.80 a share.
| The test we set | What happened | Verdict |
|---|---|---|
| Nvidia's Q3 gross margin at or above the 74.0% midpoint | 74.9% non-GAAP, 74.8% GAAP (Nov 17) | Held |
| Rocket Lab: cut both high ratings to medium if Q3 GAAP gross margin fell below 30% | 33.6%, computed from its Nov 9 release | Did not fire |
| Rocket Lab: trim, and do not buy more, if Neutron slips | Moved into 2027 on Oct 14; "no earlier than 2027" in the Nov 9 release | Fired: trimmed from 8% to 6% of the book; no adds |
| HeartFlow short: cover if Medicare's 2027 rule moves its code to a clinical payment class or pays above the 2026 level | The final rule (Nov 4) kept its code in the new-technology class it proposed and refused a higher one | Did not fire; conviction cut to low, size kept |
Our call: HeartFlow meets its raised 2026 revenue guide. We are wrong if its 2026 revenue comes in below $262 million when it reports in early 2027 (date not announced). Confidence: medium.
Ratings and positions we reviewed
| Company | Board | Result | Why |
|---|---|---|---|
| HeartFlow (short) | State of AI for Healthcare | Medium → Low | The rule the short was dated to has landed; volume is outrunning the rate |
| Rocket Lab | State of Space & Launch | High, held; trimmed from 8% to 6% | The board's own slip rule fired; the margin test did not |
| Anthropic | State of Frontier AI | High, re-affirmed | The listing added a price, not new business results |
| Nvidia | State of Nvidia; Frontier AI; Physical AI | Medium, held | The margin test passed; no permanent CEO yet |
What we're watching
| When | What | Why it matters |
|---|---|---|
| Mon Nov 23 | Symbotic's Q4 FY2026 results | Revenue against its $760–780M guide, and whether free cash flow turned positive |
| Pending | Nvidia names a permanent CEO | What our medium rating on Nvidia waits for |
| ~Late Feb or early Mar 2027 (not announced) | Nvidia's Q4 FY2027 results | Gross margin against the 71.0–72.0% guide |
| By May 17 2027 | Anthropic's 180-day lock-up ends for about 3.0B shares | Staged releases began Nov 19 2026; holders of about 8.0B shares, and the co-founders, stay locked up to Nov 2027 |
Our calls, in one place
We grade every call when its date arrives, and the wrong ones stay up with their outcome.
| Call | By | Confidence | We are wrong if | Status |
|---|---|---|---|---|
| Anthropic’s shares are at or above the $83 IPO price two quarters after listing | May 19 2027 | Medium | It closes below $83 that day | Open |
| Nvidia's Q4 FY2027 non-GAAP gross margin at or above the low end of its 71.0–72.0% guide | Q4 results, ~Feb–Mar 2027 | Medium | It prints below 71.0% | Open |
| HeartFlow meets its raised 2026 revenue guide of $262–264M | Q4 results, early 2027 | Medium | 2026 revenue comes in below $262M | Open |
| Nvidia's Q3 gross margin at or above the 74.0% midpoint (Nov 14 post) | Nov 17 2026 | Test | A print below 74.0% | Held: 74.9% |
| Another private AI lab or platform from our Nov 14 table files publicly for an IPO | Jun 30 2027 | Medium | None has filed by then | Open |
| AMD delivers its ~$17B Q4 guide at a non-GAAP gross margin near 55% (Nov 14 post) | Q4 2026 results | Medium | Q4 revenue misses the guide, or that margin falls below 55% | Open |
| AI cloud misses stay timing slips; no hyperscaler cuts its capex guide (Nov 14 post) | Mar 31 2027 | Medium | A hyperscaler cuts its capex guide, or a neocloud defaults | Open |
| Hesai's lidar revenue per unit keeps falling as volume rises (Nov 14 post) | Q4 2026 results | Medium | A quarter in which lidar revenue grows faster than units shipped | Open |
Every rating here comes from our live boards; the calls are our own. Open State of Frontier AI for the full read.
Corrections: Amazon's fund page counted all $53.4 billion of its second-quarter non-operating income as its Anthropic gain; the filed Anthropic figure is $50.5 billion. Situational Awareness's fund page now says its 13F for June 30 lists none of the chip-stock puts it held at March 31, so those listed puts were gone before July, not closed in July.
Not financial advice. Ratings and calls are QAI research views, not recommendations or solicitations, and board positions are illustrative. Conflicts of interest: none.