Context Analytics. A Bridgewise company
← All research
Case Studies·Oct 28, 2026 ·David Stolz

What Happens When You Ask Bridget™ AI to Research the Biggest IPO in History? (Part 2)

As a product manager at Context Analytics, A Bridgewise Company, I use Claude every day to write Jira tickets, run market research and run quality assurance on my products. It has become the most-used program on my laptop.

When Anthropic filed its S-1 earlier this week with a target valuation of $2 trillion (which would top the record SpaceX set earlier this year at $1.75 trillion), I wanted to dig into the filing to see if it's worth the investment. With Corporate Filings Intelligence and Bridget™, I didn't have to read 400 pages.

Bridget™ is Bridgewise's AI assistant. Corporate Filings Intelligence reads the S-1 straight from SEC EDGAR, tables included. Together they turn a filing into an afternoon of questions.

What the S-1 actually says

$16.3B
H1 2026 revenue (vs. $1.2B a year earlier)
-$80.7B
H1 2026 net loss, about 90% a non-cash charge
$112.4B
Cash and equivalents, June 30, 2026

Quarterly revenue went from $376M in Q1 2025 to $11.6B in Q2 2026, roughly 31 times larger in five quarters. The preliminary prospectus leaves Q3 revenue, the offering price and the deal size blank.

FROM THE S-1 VIA CORPORATE FILINGS INTELLIGENCE: QUARTERLY REVENUE

QuarterRevenueGrowth vs. prior quarter
Q1 2025$376MN/A
Q2 2025$787M+109%
Q3 2025$1,382M+76%
Q4 2025$2,040M+48%
Q1 2026$4,731M+132%
Q2 2026$11,555M+144%

Source: Anthropic, PBC S-1, filed October 26, 2026 (preliminary prospectus), financial tables via Corporate Filings Intelligence.

Reading the $80 billion loss

The $80.7B loss that’s stealing the headlines might not be as bad as it sounds. Before the IPO, Anthropic raised money through convertible notes and SAFEs, which work similarly to call options: holders can convert into shares at a fixed price, so the higher Anthropic's valuation climbs, the more that right is worth. Anthropic wrote those options, and accounting rules require it to mark the liability to market every quarter. As the valuation rose, the liability grew and the increase landed on the income statement as a loss. A short call on a rallying stock works the same way: a large paper loss with no cash leaving the account.

FROM THE S-1 VIA CORPORATE FILINGS INTELLIGENCE: NET LOSS VS. THE FAIR-VALUE CHARGE ($B)

PeriodNet lossFair-value charge on notes, SAFEs and derivativesCharge as % of loss
2024-$8.3B$5.4B65%
2025-$42.0B$34.3B82%
H1 2025-$13.0B$10.9B84%
H1 2026-$80.7B$72.4B90%

Strip out the charge and the operating business lost $693M in the first half of 2026, down from $2.1B a year earlier, with $1.0B of operating income in Q2 alone. Most of these instruments convert into stock at the IPO, so the charge largely stops growing (Amazon's notes carry an ownership cap, so part of them could stay outstanding). The charge reflects how much early investors' stakes have gained. The net loss isn't a measure of how fast Anthropic burns cash.

Is a $30 trillion market real?

In late August the Wall Street Journal reported that Anthropic would pitch IPO investors on a total addressable market above $30 trillion. The S-1 anchors on an even bigger pool: more than $40T in global knowledge-worker wages.

TAM CLAIMS IN CONTEXT

SourceFigureWhat it is
SpaceX S-1$28.5T93% of it AI
Anthropic, per the WSJ (via Fortune, Aug 26)More than $30TReported investor pitch, not in the S-1 text
Anthropic S-1Over $40TGlobal knowledge-worker wages in 2025 (BLS, ILO data)
U.S. GDP (per Fortune)About $32.5TFor scale

Fred Hickey of The High-Tech Strategist called the figure "nonsense," and Alex Brunicki of Backed VC expects professional investors to treat it as a mission statement. A TAM is the revenue a company would earn by capturing 100% of a market, not a forecast. Current revenue is the more useful signal.

Where the revenue actually comes from

The S-1 splits revenue geographically only into U.S. and international. In the first half of 2026 the U.S. was about 72% of revenue, and international has kept pace, growing roughly 13x year over year versus about 14x for the U.S. The filing calls international expansion "a meaningful driver of revenue growth" and cites offices and partnerships in the UK, the EU, Canada and Asia-Pacific. There is no split by industry or country.

FROM THE S-1 VIA CORPORATE FILINGS INTELLIGENCE: REVENUE BY GEOGRAPHY ($M)

PeriodUnited StatesInternationalInternational share
2024$263$12332%
2025$3,362$1,22327%
H1 2025$820$34329%
H1 2026$11,763$4,52328%

Usage-based revenue made up 80% of first-half 2026 revenue ($13.1B), with subscriptions the other 20% ($3.2B). Agentic coding leads the way: Claude's success rate on hard, open-ended engineering problems reached about 90% in September 2026, and Claude wrote over 90% of Anthropic's own September code contributions. Customers spending $100K or more grew from about 20 at the end of 2023 to more than 11,900 by September 30, 2026.

The most promising number is net revenue retention of about 900%: customers who spent anything in Q3 2025 (excluding Pro and Max subscribers) spent about nine times as much in Q3 2026. One argument against Anthropic is that it has priced Claude cheaply, even at a loss, to get customers in the door. Retention this high suggests those customers are willing to pay far more once Claude is part of their work. Two cautions: the base is small (Q3 2025 revenue was $1.4B), and the metric is company-defined and not independently verified.

The S-1 lists seven vertical solutions (Financial Services, Financial Advisors, Healthcare, Life Sciences, Legal, Education and Government), a push toward physical work such as lab and manufacturing equipment, and a note that Fortune 500 customers spend under 0.1% of their expenses on Claude. In finance, it describes agents that draft pitch books, review earnings and run financial analysis on top of the market data firms already use. Agents like that are only as good as the data behind them, which is the problem Bridget™ was built to solve.

How does Anthropic stack up against OpenAI?

The S-1 names OpenAI and Google as its primary competitors and is blunt about it: both have "substantially greater financial, technical, and marketing resources," more compute and greater brand recognition. Anthropic's pitch for why it wins is a "primary focus on enterprises," frontier research and compute efficiency, aimed at customers who prioritize "model quality, safety, and reliability."

The tell: a $637 billion compute bill

In the SpaceX S-1, the sleeper detail was orbital real estate, a finite resource one company was quietly claiming. In Anthropic's S-1, it's compute. The filing names seven large agreements and leaves the aggregate blank. The named deals add up to about $637B, though some may overlap, so treat it as an estimate. First-half revenue was $16.3B.

FROM THE S-1 VIA CORPORATE FILINGS INTELLIGENCE: NAMED COMPUTE COMMITMENTS

CounterpartyCommitmentTerm and notes
BroadcomAbout $161.2BFive-year leases, non-cancelable
SpaceXAIUp to $120.5BFull capacity expected December 2026
GoogleAt least $114.5BApril 2026 to July 2033
AmazonAt least $110BMay 2026 to April 2036
MicrosoftAt least $44.9BNovember 2026 to May 2033
NscaleUp to about $44.6BInto 2033
LambdaAbout $41.3BInto 2034

Source: Anthropic S-1 via Corporate Filings Intelligence. Off the table: an additional $125B to $150B of TPU compute for 2028 to 2031. AMD has no fixed dollar commitment.

SpaceX shows up on this list, and the deal has grown since Part 1. In my SpaceX post I wrote that Anthropic would pay about $1.25 billion a month for compute at the Colossus 1 data center under the agreement signed in May. Amendments since then added the much larger Colossus II, and once all capacity is delivered (expected by December 2026), payments rise to up to $3.1B a month, or up to $120.5B in total. After the non-cancelable periods end in early 2027, Anthropic can cancel on 90 days' notice.

The money is also circular. Amazon, Google and Microsoft are each an investor in Anthropic, a supplier to it and a customer of it. They invest billions, Anthropic spends a large share of it back on their cloud and chips, that spending shows up as their revenue and backlog, and as Anthropic's valuation rises they book gains on their stakes. The same markup that creates Anthropic's paper loss creates their paper gains. Amazon's convertible notes carry a 12% ownership cap, and Google can buy up to $10B of Class N shares in a concurrent private placement.

Four channels could carry that into their stocks (analysis, not a prediction):

  • Earnings quality. Reported net income can include large non-cash gains. Press reports put Alphabet's net gains last quarter at $99B and Amazon's marketable securities up $50.5B in Q2, mostly its Anthropic stake. A weak aftermarket could flip those gains into losses with no change in the core business.
  • Revenue quality. Part of their cloud backlog is one customer's commitment, funded partly by their own capital. The S-1 expects Anthropic to use more cash than it generates for the foreseeable future, so that backlog depends on Anthropic's ability to keep raising money.
  • Supply overhang. Class N shares convert to Class A when sold to an unaffiliated party, and press reports say partners could sell stakes after the lockup to fund data centers.
  • Price discovery. Their stakes are carried near the last private round ($965B, per press). An IPO price sets a public mark that can move their results either way.

Bridget™, as of the October 27 close, shows no change: ratings and targets held for all five partners, and none of the reports mention Anthropic. The clue is in the margins. Alphabet's Q2 net margin was 93.65% against a 34% EBIT margin, and Amazon's net income rose 107% from Q1 on a 13.7% EBIT margin. Both gaps fit with large investment gains, though the data doesn't name the source.

BRIDGET™: ANTHROPIC'S BIGGEST PARTNERS

TickerRatingTarget vs. closeMarket cap
AMZNHold$230 vs. $253.10$2.73T
GOOGLHold$350 vs. $320.58$3.90T
MSFTHold$400 vs. $531.33$3.95T
AVGOBuy$510 vs. $342.33$1.63T
NVDAOutperform$245 vs. $217.94$5.26T

Source: Bridget™ ratings and price targets, close of October 27, 2026.

What the filing says could go wrong

The risk factors get specific. They say the most capable models could provide meaningful assistance to chemical and biological weapons actors, that safeguards "may be incomplete," that self-preserving model behaviors have been observed, and that "no single insurer" may have the capacity to cover the risk.

Safety also shows up in revenue. About 5% to 7% of nine-month revenue came from customers later found to be violating Anthropic's usage policy. That revenue is recognized before the violation is detected, and the company says it tightened enforcement in August and September.

If Anthropic has so much cash, why IPO?

Cash plus short-term investments totaled $131.7B at June 30, 2026. Against roughly $637B of named compute commitments, that isn't enough, and the filing expects significant net cash used in operating and investing activities for the foreseeable future.

The S-1 doesn't state a reason for the timing, though the filing points to several. The notes and SAFEs convert at the IPO, which cleans up the balance sheet and ends most of the paper charge. Selling stockholders can cash out. Partner deals are tied to the offering, including Google's option to buy up to $10B of stock. Press reports add that the window is open: SpaceX listed in June and OpenAI has postponed its own IPO.

So, is it a good investment?

I asked Bridget™ the obvious question.

BRIDGET™ AIQ: Is Anthropic a good investment?Deciding whether to invest in a specific company is a personal choice that depends on your unique financial goals. Since Anthropic is currently a private company, it is not available for public trading on our platform.

Bridget™ gave the same answer on SpaceX before it traded. It answers from verified data, so when a company has no trading history or coverage, it says so instead of inventing a view. With my own money, that’s the answer I want.

The bottom line

The growth is real, the $80.7B loss is mostly an accounting mark, and the $30 trillion TAM is aspirational. The tell is a compute bill several times larger than anything Anthropic has earned, owed largely to partners who are also its investors and customers. Whether that justifies $2 trillion depends on how long revenue keeps compounding, and the final prospectus still has to fill in the price, the size and Q3.

Read the filing, check the data, and do your homework.

This post is for informational purposes only and does not constitute investment advice. Context Analytics and Bridget™ AI do not issue buy/sell recommendations on securities without established price history. All S-1 data sourced via Corporate Filings Intelligence from SEC EDGAR (preliminary prospectus filed October 26, 2026; price, size and Q3 revenue are blank). Figures attributed to press reports were not confirmed in the filing text. Net revenue retention and customer metrics are company-defined and not independently verified. Bridget™ data as of the October 27, 2026 close.

More research
Nov 11, 2026 The Calm Before the Crash? What Twitter Sentiment Revealed About the 2026 Midterms Sep 30, 2026 How Tariff Refunds Are Boosting Headline Earnings Sep 16, 2026 AI Is Entering the Annual Report Through the Risk Door
Subscribe

New research, in your inbox.

One note when we publish something worth your time, written by the team that built the data, not marketing.

No spam. Unsubscribe anytime.
CadenceWeekly, at most
Written byThe research team, not marketing
Opt-outOne click, anytime

Test the data behind the research.

Request trial data