Skip to main content

Select Accept all cookies to allow optional cookies, or Reject optional cookies to turn off optional advertising cookies. Analytics remains on with either choice. You can change your advertising choice at any time through Cookie settings.

Cookie Policy (opens in a new tab)Privacy Policy (opens in a new tab)

Global Markets

Anthropic's IPO: Record Demand, A Volatile Debut

A $81.34bn public offering met a 17% first-day gain. The harder question is whether revenue growth can outrun the long-term cost of compute · Post-IPO Analysis · Data cut-off 19 November U.S. close

Published 20 November 202612 min readSerrari Compliance & Strategy
Contents & display options
Global Markets

Anthropic's IPO: Record Demand, A Volatile Debut

A $81.34bn public offering met a 17% first-day gain. The harder question is whether revenue growth can outrun the long-term cost of compute · Post-IPO Analysis · Data cut-off 19 November U.S. close

Published 20 November 202612 min readSerrari Compliance & Strategy

Anthropic priced 980m Class A shares at $83.00, opened at $111.50 and closed its first session at $97.17 — a 17.1% offer-to-close gain after a 12.9% open-to-close slide. The desk works through the offer mechanics, the $1.285tn basic capitalization, 56.6% compute intensity and the $313.8bn commitment stack now facing a public valuation test.

Anthropic's IPO: Record Demand, A Volatile Debut

Report cover: Anthropic IPO — $83.00 offer price on 18 Nov, $97.17 first-session close on 19 Nov, $313.8bn future fixed / minimum commitments
Report cover: Anthropic IPO — $83.00 offer price on 18 Nov, $97.17 first-session close on 19 Nov, $313.8bn future fixed / minimum commitments

SERRARI GROUP — ANALYST DESK REPORT / U.S. EQUITIES

A $81.34bn public offering met a 17% first-day gain. The harder question is whether revenue growth can outrun the long-term cost of compute.

$83.00 — IPO offer price, 18 Nov

$97.17 — First-session close, 19 Nov

$313.8bn — Future fixed / minimum commitments

Prepared 20 November 2026 | Data cut-off: 19 November U.S. close

Research and market education; not personalised investment advice.


Executive take

Executive take: what happened, why investors care, the central question and the desk assessment of the Anthropic debut
Executive take: what happened, why investors care, the central question and the desk assessment of the Anthropic debut

The world's first listed frontier-AI developer now faces a public valuation test.

What happened

Anthropic priced 980m Class A shares at $83.00, above its marketed range. Trading started on 19 November at $111.50, then closed at $97.17. The public offering had $81.34bn of gross value.

Why investors care

It provides a liquid benchmark for frontier-AI economics. The price must ultimately reflect growth duration, compute capacity, cost discipline, corporate control and the timing of capital needs.

Central question

Can fast-growing revenue and improving operating cash receipts support the scale of compute, power and infrastructure commitments without a sustained deterioration in margins or substantial dilution?

Desk assessment

The debut demonstrated exceptional demand but not a settled valuation. Volatility from open to close shows that the market may separate long-term AI adoption from the near-term price investors will pay.


The debut: strength and sharp intraday repricing

Bar chart — ANT share price on 19 November 2026: $83.00 IPO offer, $111.50 first trade, $97.17 first close
Bar chart — ANT share price on 19 November 2026: $83.00 IPO offer, $111.50 first trade, $97.17 first close

Offer-price gains should not be confused with the return earned by investors buying at the opening trade.

ANT share price ($), 19 November 2026

IPO offer: $83.00 · First trade: $111.50 · First close: $97.17

+34.3% at the open

The opening transaction priced in a much larger premium than the final close. That is a market price, not extra capital raised by Anthropic.

-12.9% open-to-close

The first session closed at $97.17: up 17.1% versus the offer, but sharply lower than the $111.50 opening trade.


Offer mechanics: public float, private capital

Offer mechanics: 980m Class A shares offered, 956.35m new issuer shares, 23.65m selling-holder shares, $26.0bn Class N private placement, ~$104.8bn net proceeds
Offer mechanics: 980m Class A shares offered, 956.35m new issuer shares, 23.65m selling-holder shares, $26.0bn Class N private placement, ~$104.8bn net proceeds

Different components of the transaction do not all provide the same cash to the company.

980m — Class A shares offered

956.35m — New issuer shares

23.65m — Selling-holder shares

$26.0bn — Class N private placement

Gross public issue value ($bn)

$79.38bn issuer primary — issuer cash: gross, before fees. Secondary seller proceeds: $1.96bn; no issuer benefit.

~$104.8bn net

Issuer-estimated combined net proceeds from the public primary sale and the $26bn private placement. The $81.34bn public issue value includes secondary shares. The underwriters' option is separate and not assumed exercised.

The private placement was conditional on completion of the public offering; closing was scheduled for 20 November 2026.


Valuation: specify the share-count basis

Valuation bars — market cap on basic share count: $1.097tn at offer, $1.474tn at open, $1.285tn at close; ~$2tn reported fully diluted value at opening
Valuation bars — market cap on basic share count: $1.097tn at offer, $1.474tn at open, $1.285tn at close; ~$2tn reported fully diluted value at opening

Headline "market cap" figures differ when analysts include outstanding options, RSUs or other contingent shares.

13.220bn — Basic issued shares after offer / placement

$1.285tn — Basic capitalization at $97.17 close

~$2tn — Reported fully diluted value at opening

Market cap on basic share count ($tn)

At offer: $1.097tn · At open: $1.474tn · At close: $1.285tn

Two valuation lenses

At the first close, the basic-capitalization valuation is about 29.0x annualized January–September revenue or 18.9x four times third-quarter revenue. Both are mechanical annualizations, not management forecasts or standard forward multiples.


Revenue acceleration: exceptional, but not yet normalized

Bar chart — quarterly recognized revenue ($bn): Q1'25 0.4, Q2'25 0.8, Q3'25 1.4, Q4'25 2.0, Q1'26 4.7, Q2'26 11.6, Q3'26 17.0
Bar chart — quarterly recognized revenue ($bn): Q1'25 0.4, Q2'25 0.8, Q3'25 1.4, Q4'25 2.0, Q1'26 4.7, Q2'26 11.6, Q3'26 17.0

Quarterly revenue jumped sharply in 2026; such rates are a poor proxy for permanent growth.

Quarterly recognized revenue ($bn)

Q1'25: 0.4 · Q2'25: 0.8 · Q3'25: 1.4 · Q4'25: 2.0 · Q1'26: 4.7 · Q2'26: 11.6 · Q3'26: 17.0

$33.24bn

Recognized revenue for the nine months to September 2026, compared with $2.545bn for the same nine-month period of 2025.

What changes next?

The investment case depends on whether customer cohorts continue expanding while pricing, inference efficiency and market competition change.


Operating economics: compute remains the hinge

Bar chart — nine months to September 2026 ($bn): revenue 33.242, compute 18.808, personnel 4.863, sales/marketing 5.529, other admin 5.302
Bar chart — nine months to September 2026 ($bn): revenue 33.242, compute 18.808, personnel 4.863, sales/marketing 5.529, other admin 5.302

Income statement expenses show a 3.8% nine-month operating loss margin even after rapid expansion.

Nine months to September 2026 ($bn)

Revenue: 33.242 · Compute: 18.808 · Personnel: 4.863 · Sales / marketing: 5.529 · Other admin: 5.302

56.6% compute intensity

Compute and infrastructure expenses absorbed 56.6% of recognized revenue during the nine-month period. This is an expense-to-revenue ratio, not a disclosed gross margin.

-$1.26bn operating result

After $34.50bn of total operating expenses on $33.24bn of revenue, the period ended with an operating loss. Q2 2026 was profitable; Q3 returned to an operating loss.


Net loss is not a cash-burn estimate

Net loss decomposition: -$104.0bn reported nine-month net loss, -$94.3bn fair-value changes largely non-cash, -$1.26bn operating loss separate from marks
Net loss decomposition: -$104.0bn reported nine-month net loss, -$94.3bn fair-value changes largely non-cash, -$1.26bn operating loss separate from marks

The $104bn reported nine-month net loss was dominated by fair-value changes in financing instruments.

-$104.0bn — Reported nine-month net loss

-$94.3bn — Fair-value changes, largely non-cash

-$1.26bn — Operating loss, separate from marks

Why the reported loss is so large

The company remeasured convertible notes, derivative liabilities and related instruments as its private valuation increased. These fair-value expenses drove the bulk of the reported net loss, rather than customer billings or direct operating cash outlays.

These charges are not proof that $94.3bn of cash left the company in the period.

Investor interpretation

The company still carried genuine operating costs and a complex financing structure. The right comparison is among recognized revenue, operating margin, operating cash flow, investment cash flow and post-IPO capital requirements—not the net loss headline in isolation.


Liquidity is substantial; obligations are larger

Illustrative liquidity bridge ($bn): cash + investments 131.9, net offering + placement 104.8, illustrative combined 236.7
Illustrative liquidity bridge ($bn): cash + investments 131.9, net offering + placement 104.8, illustrative combined 236.7

Cash balances and IPO proceeds provide financing capacity but do not erase contractual commitments.

$95.6bn — Cash / cash equivalents at 30 Sep

$36.3bn — Short-term investments at 30 Sep

$6.57bn — 9M 2026 operating cash flow

Illustrative liquidity bridge ($bn)

Cash + investments: 131.9 · Net offering + placement: 104.8 · Illustrative combined: 236.7

Not an audited pro forma

The $236.7bn is simple arithmetic before post-September movements, working-capital needs, transaction settlement, restrictions and investment commitments. Historical investing cash outflow of $36.55bn also reflects items beyond conventional capital expenditure.


The $313.8bn compute test

The $313.8bn compute test: supply-risk chain from chips (GPU/memory availability) to power (grid access and cooling) to capacity (data-centre delivery) to margins (revenue/cost recovery)
The $313.8bn compute test: supply-risk chain from chips (GPU/memory availability) to power (grid access and cooling) to capacity (data-centre delivery) to margins (revenue/cost recovery)

Contractual infrastructure exposure runs across years and must be matched by future demand and efficiency.

$313.8bn — Future fixed / minimum guaranteed commitments

1–20 yrs — Stated range of contractual terms

Beyond this — Variable commitments excluded from the figure

How supply risk becomes an equity valuation problem

CHIPS (GPU / memory availability) › POWER (Grid access and cooling) › CAPACITY (Data-centre delivery) › MARGINS (Revenue / cost recovery)

The $313.8bn is a future contractual exposure, not an immediate payable, recognized expense or outstanding debt balance.


Revenue mix and concentration

Donut chart — nine-month 2026 revenue mix: consumption 79.1% ($26.31bn), subscription 20.9% ($6.94bn); 44% of recognized revenue via cloud-service-provider marketplaces
Donut chart — nine-month 2026 revenue mix: consumption 79.1% ($26.31bn), subscription 20.9% ($6.94bn); 44% of recognized revenue via cloud-service-provider marketplaces

Strong usage demand coexists with distribution and customer concentration risks.

Nine-month revenue mix, 2026 ($bn)

Consumption 79.1% · Subscription 20.9%

Consumption-based fees: $26.31bn. Subscription-based: $6.94bn. Revenue exposure therefore reflects both enterprise usage intensity and subscription retention.

44% of recognized revenue came through cloud-service-provider marketplaces.

Top-five accounts: ~25%

No single customer exceeded 10% of revenue in the nine months to September, but the five largest represented about one quarter of sales. Contracts and usage can change quickly.

The dual role of partners

Cloud platforms can supply infrastructure, distribute models, bill customers and hold equity. That creates strategic reach alongside dependency and potential conflicts.


Control is not proportional to public ownership

Share classes: Class A publicly traded with one vote per share, Class N generally non-voting, Class F fixed 50.1% aggregate voting power
Share classes: Class A publicly traded with one vote per share, Class N generally non-voting, Class F fixed 50.1% aggregate voting power

The capital structure gives mission and founders powers unavailable to ordinary public investors.

Class A — Publicly traded / one vote per share

Class N — Generally non-voting stock

Class F — Fixed 50.1% aggregate voting power

Founder control

A single Class F share held by an entity associated with the seven co-founders carries fixed majority voting power on matters generally submitted to shareholders. Further Class A issuance does not dilute this voting percentage.

Long-Term Benefit Trust

A special Class T share gives the trust authority over a majority of board-director positions under the prospectus structure. Holders of publicly traded Class A stock cannot assume standard shareholder influence over strategy.


Risk map: where a premium valuation can break

Risk map table: opening-price volatility and compute economics high/high; energy delivery and regulatory medium-high/high; control, concentration medium/high; lock-up medium/medium-high
Risk map table: opening-price volatility and compute economics high/high; energy delivery and regulatory medium-high/high; control, concentration medium/high; lock-up medium/medium-high
Risk / exposureLikelihoodImpactMonitoring signal
Opening-price volatilityHighHighIntraday spread / volumes
Compute economicsHighHighCompute share of sales
Energy / capacity deliveryMedium-highHighInfrastructure delivery
Control and dilutionMediumHighGovernance / new issuance
Customer / CSP concentrationMediumHighMarketplace exposure
Regulatory and model safetyMedium-highHighLegal / policy outcomes
Lock-up and float supplyMediumMedium-highUnlock disclosures

Likelihood and impact are qualitative Serrari desk judgments, not modelled probabilities.

Figure: Risk map — where a premium valuation can break.


Scenario map: value depends on duration and dilution

Basic-share valuation sensitivity table from $70.00 ($0.925tn, 20.9x) to $125.00 ($1.652tn, 37.3x), with upside, base and downside paths
Basic-share valuation sensitivity table from $70.00 ($0.925tn, 20.9x) to $125.00 ($1.652tn, 37.3x), with upside, base and downside paths
Hypothetical share priceMarket cap (tn)Annualized 9M salesQ3 x 4
$70.00$0.925tn20.9x13.6x
$83.00$1.097tn24.8x16.2x
$97.17$1.285tn29.0x18.9x
$111.50$1.474tn33.3x21.7x
$125.00$1.652tn37.3x24.4x

Simple price sensitivities are illustrative mark-to-market arithmetic, not targets or recommendations.

Basic-share valuation sensitivity

Upside path — Demand and efficiency improve, supporting richer multiples.

Base path — Sales grow, but compute spending keeps margins volatile.

Downside path — Compute pressures and rising float weaken equity returns.


Conclusion and evidence register

Conclusion and evidence register: first-day premium is an observation, the long-run investment thesis remains conditional; eight supporting references
Conclusion and evidence register: first-day premium is an observation, the long-run investment thesis remains conditional; eight supporting references

The first-day premium is an observation; the long-run investment thesis remains conditional.

The first day confirms extraordinary IPO demand, not an equilibrium valuation. Anthropic's revenue and operating cash flow improvements are material, but the public market will now test whether its compute-heavy model scales economically. Investors should track operating cost intensity, capex and contractual capacity deployment, concentration and future share supply alongside revenue—not just the trading premium.

Evidence and supporting references

Method: financial data are from the final offering prospectus, mostly as of 30 September 2026. First-day trading figures are for 19 November. Share-count-based valuations and revenue multiples are Serrari calculations, not issuer guidance. Scenarios are illustrative; no stock recommendation is made.