Akamai’s $11.6B Anthropic contract is real; its cash return is still a forecast

Summary: Akamai’s September 24 filing disclosed a seven-year Anthropic commitment of about $11.6 billion, but its own presentation pairs it with roughly $5.5 billion of estimated contract capex, front-loaded before meaningful revenue. The pre-market reference had already risen about 21% from the prior regular close, while Anthropic receives a warrant for up to roughly 5% of Akamai’s common shares on an as-converted basis. This is a material change to the August cloud-growth thesis, but the current price does not leave a modeled +10% base case or a credible −5% downside cap. Reject / no trade.

Research timestamp: September 25, 2026, 20:42 Singapore time (12:42 UTC). The U.S. regular session has not opened. The scenario reference is the $133.70 pre-market vendor snapshot observed around 8:36 a.m. EDT; the latest completed regular close was $110.41 on September 24. Neither observation is an executable entry.

Opportunity Ranking

Rank	Candidate	Direction	Mispricing	Evidence Freshness	Catalyst Window	Positioning	10/5 Status	Tradeability	Main Rejection Risk
1	Akamai (AKAM)	Long screen	A new $11.6B seven-year service commitment could change cloud economics; the contract needs about $5.5B estimated capex and arrives with a warrant and a sharp pre-market repricing	Sep. 24 SEC 8-K, issuer capex/revenue schedule and regular close; operating balance sheet last filed June 30	First service start expected late Q2 2027; FY27 ramp; full $1.7B annual run-rate expected by end 2028	Market reaction is visible; short, fund and dealer positioning are not verified	Reject: guide-based base +0.6%, downside stress −20.9%, weighted price −2.8%	Nasdaq common; pre-market spread/depth and exit capacity unverified	Contract margin, operating cost, financing and service performance are not yet demonstrated
2	GRAIL (GRAL)	Long screen	FDA advisory support improved the Galleri PMA path, but the effectiveness vote was only 6–4 and the stock rose 15.38% on Sep. 24 after earlier event gains	Sep. 23 issuer release, FDA panel materials and Sep. 24 close	FDA decision expected in coming months; timing and label unresolved	Price/volume show event repricing, not who owns or hedges the risk	Reject: favorable regulatory odds do not establish an executable +10% base or bound a clinical/commercial downside	Nasdaq common; live post-event market structure not verified	FDA is not bound by its advisory panel; effectiveness remains contested
3	H.B. Fuller (FUL)	Long screen	Q3 margin and EPS grew, but pricing outweighed lower volume; nine-month working capital rose and net debt remained material	Sep. 23 Q3 release and Sep. 24 close	FY26 close, AMS acquisition close and next quarter’s volume/cash conversion	Current direct positioning not verified	Reject: low nominal forward P/E does not bound leverage or volume downside	NYSE common; current liquidity checks absent	Guidance still embeds low-single-digit organic volume decline and 3.0× net leverage

Selected opportunity: AKAM, for information value only, not as the best executable long.

Why this one now: The September agreement materially changes the evidence boundary from the Desk’s August 9 AKAM Watchlist. That earlier thesis tested whether reported CIS growth and more than $2.8 billion of prior signed contracts could overcome weaker cash conversion. The new SEC-filed Project Plans now disclose one customer’s $11.6 billion commitment, delivery conditions, a roughly $5.5 billion contract-capex schedule, a 2027 revenue ramp, and customer-linked warrants. The issue is no longer whether large cloud contracts exist; it is whether this unusually large commitment earns adequate returns after capital and customer claims.

What should surprise the reader: Akamai’s own schedule puts about $1.7 billion of capex in Q4 2026 with no Anthropic revenue in that period, then about $3.1 billion of capex in FY27 against only $150–$300 million of expected FY27 revenue. Full contracted run-rate is not expected until the end of 2028. The contract is valuable evidence of demand, but the headline $11.6 billion is neither near-term revenue nor profit.

Why This Is the Best Opportunity Right Now

AKAM has the clearest fresh primary state change among the candidates. The company filed an $11.6 billion agreement, published its capex and service-start phasing, and disclosed the warrant economics. That lets the market’s immediate repricing be compared with contract delivery and capital requirements rather than with an unverified AI narrative. It also changes the August cloud-growth thesis enough to justify a new screen, while remaining a Reject under current 10/5 evidence.

GRAIL has a dated FDA catalyst and a direct regulatory vote, but the advisory recommendation is nonbinding and the effectiveness vote was divided. H.B. Fuller has an auditable Q3 operating result and raised FY26 adjusted EPS guidance, but pricing drove revenue growth while volume remained negative and net debt-to-adjusted EBITDA was 3.0×. Neither alternative has AKAM’s new, issuer-quantified contract/capex/warrant bridge.

The strongest countercase is substantial: Anthropic has committed to buy a large amount of dedicated cloud capacity; the delivery schedule is disclosed; and management says the economics should generate strong cash conversion and operating margins. If Akamai uses already planned infrastructure efficiently and hits the delivery dates, the contract could validate a new earnings engine. That upside is real. The rejected long case is narrower: at today’s pre-market reference, the near-term earnings guide and unproven contract margin do not support a +10% highest-probability base or a credible −5% adverse ceiling.

Why This Can Move More Than 5% Soon

The market has already shown a large move: AKAM closed September 24 at $110.41, down 6.78% that session before the contract announcement, then traded as high as $131.69 in the after-hours snapshot and $133.70 in a separate Sep. 25 pre-market snapshot. The latter is about 21.1% above the prior regular close. These are non-comparable session observations and the pre-market price can reverse; they demonstrate repricing, not an entry.

The next economic steps are slower than the price reaction. Akamai’s investor presentation estimates service initiation in late Q2 2027, $150–$300 million of FY27 revenue, a ramp during the second half of 2027 and the full contracted revenue run-rate at end-2028. A Q3 report can test existing guidance, but it cannot yet demonstrate sustained Anthropic service revenue. The cheapest falsification sequence is to compare reported capex, capacity delivery, service availability, recognized Anthropic revenue and customer cash receipts against that schedule as each reporting period arrives.

10/5 Asymmetry Gate

The map below is a reference-only price sensitivity using Akamai’s existing FY2026 non-GAAP diluted EPS guidance, not a forecast of Anthropic profit. The midpoint is $6.725; the guide range is $6.40–$7.05. The assumed multiples make the valuation uncertainty explicit; they are not observed consensus targets or management forecasts.

Test	Reference-only result from $133.70 pre-market snapshot
Favorable base-case move	+0.6%
Credible adverse scenario	−20.9%
Gross positive base reward / adverse risk	0.03:1
Probability-weighted price value	$129.95, or −2.8% before costs
Classification	Reject / no trade

The highest-probability base fails the +10% long hurdle; the bottom case exceeds the −5% bound. The agreement may ultimately justify a higher 2027–2030 earnings path, but the current disclosure does not give contract-level operating margin, support costs, maintenance capital, working capital or a fully reconciled funding bridge. This sensitivity cannot substitute for those missing economics. Reject precedes Watchlist even though execution evidence is also incomplete.

What Should Surprise the Reader

The useful surprise is the gap between customer commitment and cash timing. The SEC says Anthropic’s approximately $11.6 billion commitment is subject to contract termination provisions and Akamai satisfying delivery and service-availability requirements. The issuer’s presentation estimates $5.5 billion of capex for the new relationship, with about $4.8 billion scheduled through FY27, while the initial revenue phase remains small. Subtracting estimated capex from the contract commitment leaves $6.1 billion nominally before power, colocation, operations, financing, maintenance capex, taxes, timing and share claims; it is not profit or present value.

The warrant is also not an immediate five-percent share issuance. Akamai issued a warrant over 387,051 non-voting preferred shares, initially convertible into about 7.7 million common shares, at an equivalent common price of $111.33. The first 40% tranche is tied in the SEC filing to the first payment under Project Plan 3; the three remaining 20% tranches depend on each additional $3 billion of committed contract value. The issuer’s press release describes approximately 2% as expected to vest with today’s commitment and another 3% with expansion. That difference between the press summary and legal vesting trigger matters: do not count the full warrant as issued shares today, but do not treat it as free growth either.

The Setup

Fact: Akamai entered Project Plans 2 and 3 with Anthropic on September 18 under the existing May 5 Master Services Agreement. The SEC-filed 8-K says Anthropic committed approximately $11.6 billion in aggregate over initial seven-year terms, subject to delivery/service availability requirements and termination rights. It can expand by another $9 billion at mutually agreed terms; expansion is not committed revenue today.

Fact: The September 24 company presentation estimates about $5.5 billion in contract capex: $1.7 billion in Q4 2026, $3.1 billion in FY27 and $0.7 billion in FY28. It estimates no FY26 Anthropic revenue, $150–$300 million in FY27, the full approximately $1.7 billion annual run-rate by end-2028 and ratable revenue thereafter. These are management estimates and can change.

Fact: The same presentation says fulfillment of prior large CIS contracts plus Anthropic requires approximately 95–105 megawatts and estimates about $22 million of annual revenue per megawatt. It combines approximately $2.8 billion of earlier commitments with the new $11.6 billion agreement for about $14.4 billion total contract value. The earlier commitments and the Anthropic plan must not be double-counted as incremental Anthropic sales.

Fact: Akamai’s most recent Form 10-Q is for June 30. It showed $4.616 billion of cash, cash equivalents and marketable securities; $7.640 billion of convertible-note principal; first-half operating cash flow of $638.8 million; cash capex of $246.6 million; and Q2 non-GAAP operating margin of 25%, down from 30% a year earlier. These balances predate the September agreement and do not prove funding sufficiency or a cash shortfall.

Inference: The signed commitment de-risks demand relative to the prior thesis, but moves the underwriting bottleneck to execution and incremental returns. The near-term funding draw is front-loaded, while most revenue is later. The prior thesis asked whether a generic cloud pipeline would convert; the new thesis asks whether a large committed customer contract can generate return on capital after build-out, operating cost and equity consideration.

Unknown: Contract-level contribution margin; power, colocation and maintenance costs; financing cost; precise customer-payment schedule; cancellation remedies for each service failure; service utilization and acceptance criteria; and the full cost/return on the initial compute build. I do not have sufficient reliable data to quantify this accurately.

The Market Price

The latest completed regular-session close was $110.41 on September 24 at 3:59:56 p.m. EDT, down 6.78%. ChartExchange’s page also showed an after-hours $131.69 snapshot at 4:59:30 p.m. EDT and a pre-market $129.51 snapshot at 8:12:30 a.m. EDT on September 25. StockAnalysis’ pre-market movers snapshot observed around 8:36 a.m. EDT showed $133.70. These references differ by session and time; only the $133.70 snapshot is used for the scenario map, and none is an executable quote.

ChartExchange’s page showed a $110.43 summary quote and a $110.41 historical close for Sep. 24; I use its timestamped $110.41 row. Displayed volume also varied by snapshot. I do not rely on volume to infer who traded. The event repricing and exact current pre-market depth, spread, venue quality, volume quality and exit liquidity were not independently verified.

Using the latest share count read in Akamai’s June 30 filing, approximately 143.7 million shares, the $133.70 reference implies roughly $19.2 billion of equity value. At the $6.725 FY2026 non-GAAP EPS midpoint it is about 19.9×. The report and the new contract do not provide a current filed share count after warrant issuance, nor a revised FY2026 EPS bridge that includes Anthropic revenue. This market-cap/EPS comparison excludes net debt and is not enterprise value.

At the current pre-market reference, a +10% base return requires about 21.9× the $6.725 midpoint EPS, or about 20.9× the $7.05 upper-end EPS. The contract’s main revenue ramp is expected in FY27 and 2028, not the current FY26 EPS period; the market is therefore paying for anticipated execution before reported service revenue.

The Mispricing

The price disagreement is between a newly signed, large-volume compute customer and the value retained by Akamai shareholders after the capacity build. A $1.7 billion annualized service run-rate is material beside Akamai’s prior $4.4–$4.5 billion FY26 revenue guide, but it takes time to reach, and revenue is not contribution profit. The planned $5.5 billion capex is around 29% of the approximately $19.2 billion pre-market equity value in gross dollars, though it is a multi-year investment and should not be treated as an immediate equity loss.

The market may be right to re-rate the company: a seven-year commitment supports utilization visibility and a real infrastructure business rather than a speculative pipeline. It may also be overpaying for gross contract value if capacity cost, power, support, financing or dilution absorbs the economics. Current FY26 guidance implies little base-case price upside under a 20× midpoint sensitivity, and the disclosed service schedule offers no near-term reported revenue proof.

The Positioning

Observed: the regular close fell 6.78% on September 24 before the evening announcement; pre-market prices subsequently rose sharply. This shows a state change in price, not forced covering, dealer gamma or fund flows.

Known corporate claim: Anthropic received a warrant for a 5% maximum as-converted common-equivalent stake, with the first 40% vesting only on first payment under Project Plan 3 in the SEC filing and later tranches linked to added commitments. The initial warrant has an equivalent strike of $111.33. At the $133.70 pre-market reference that strike is below market, but vesting, exercise and conversion remain conditional; do not describe the entire warrant as current float.

Missing: current short interest, borrow, options open interest, dealer exposure, holder flows, complete regular-session depth, pre-market spread and exit-size analysis. The positioning score is capped at 3/5.

The Catalyst

Contract delivery and first service: Akamai’s issuer presentation expects service initiation in late Q2 2027 and FY27 revenue of $150–$300 million. Verify acceptance, revenue recognition, collections, and actual utilization, rather than treating the initial commitment as immediate sales.
Build-out and capital use: compare quarterly capex with the $1.7B/$3.1B/$0.7B 2026–2028 schedule, and reconcile CFO, working capital, debt and liquidity. Delay, cost overruns, supply limits or unavailable capacity can impair returns even if Anthropic demand persists.
Warrant vesting and expansion: distinguish the first-payment trigger in the filed agreement from the issuer’s “expected to vest” press-release wording; later warrant tranches require each additional $3B of contract value at mutually agreed terms.
Next operating reports: evaluate FY26 guide delivery and later CIS margin, non-GAAP/GAAP cost bridge and conversion. The new contract’s full run-rate is expected only by end-2028, so near-term reports can validate readiness but not the entire long-run economics.

Cheapest falsification test: the next filed quarter’s capex, cash flow and CIS revenue against management’s disclosed deployment timetable. If capex or cash needs materially exceed the schedule, while service starts or customer receipts slip, the long inference weakens. Conversely, timely delivery, high utilization and a reconciled positive incremental cash contribution would justify a new valuation, not automatic entry.

The Payoff

This is a six-month, price-only sensitivity around the next two reporting periods. Top and base use the FY2026 non-GAAP EPS guide’s upper and midpoint, respectively; bottom uses the lower end. The multiples are analyst assumptions, not historical downside floors or market consensus. The map intentionally assigns no value to the long-dated Anthropic revenue before evidence establishes its margin and capital return. It is therefore a near-term rejection screen, not a full DCF of the seven-year contract.

Subjective probabilities are 20% / 50% / 30%. The top needs guide delivery plus a 22× re-rating as the contracted cloud opportunity becomes a credible earnings stream. Base is most likely because it holds the current multiple near 20× the midpoint, with no material Anthropic revenue yet recognized. Bottom allows 16.5× on the low-end guidance if 2026 guide/cash conversion weakens while shareholders fund capacity and absorb warrant-related claims. The probabilities are estimates, not observed frequencies.

Price Target and Probability Map

Scenario	Probability	Target / Level	Return / Payoff	Horizon	Conditions	Evidence Quality
Top Case	20%	$155.10	+16.0%	Six months, through next two reporting periods	FY26 EPS reaches $7.05 upper guide and market applies 22× as capacity delivery and customer commitment support a higher cloud value	Low: EPS guide is issuer-stated; 22× is an assumption; contract margin is unknown
Base Case	50%	$134.50	+0.6%	Same six-month window	$6.725 guide midpoint and 20× reference multiple; no material recognized Anthropic service revenue yet	Medium-low: guide is current but multiple is subjective and new contract cash economics remain unreported
Bottom Case	30%	$105.60	−21.0%	Same six-month window	FY26 EPS reaches $6.40 guide low and multiple compresses to 16.5× as investment, timing or legacy segment concerns dominate	Low: explicit stress rather than issuer forecast
Invalidation	n/a	No numeric entry threshold	N/A	Next results or material service update	A formal contract amendment or failure to meet service availability, a material FY26 guide change, or a major capex/funding deviation requires rebuilding the map; it is not an automatic order	Medium

Probability-weighted expected value: $129.95, or −2.8% from the $133.70 pre-market reference before costs.

Current market level and timestamp: $133.70 pre-market vendor snapshot observed around Sep. 25, 2026, 8:36 a.m. EDT; separate timestamped $129.51 pre-market quote at 8:12:30 a.m. EDT; latest completed regular close $110.41 at 3:59:56 p.m. EDT Sep. 24.

Primary instrument: AKAM unlevered common stock, reference analysis only.

10/5 favorable base move: +0.6%.

10/5 credible adverse move: −20.9%.

10/5 measurement basis: reference-only from the $133.70 pre-market snapshot.

10/5 status: Reject.

Confidence: High on disclosed contract, capex schedule and warrant terms; low on contribution margin, 2027–2030 valuation and execution probabilities.

The Kill Shot

The mature bull case is not just an AI label: Anthropic has agreed to spend approximately $11.6 billion over seven years, Akamai’s distributed network can serve CPU workloads at scale, and the schedule offers visibility not present in the old pipeline. The company also retains an established security business. If incremental compute is high-margin after the one-time capacity build, cash conversion could be attractive and a 20× FY26 price sensitivity could understate the long-run value.

The fragile assumption in the bear/reject case is that investors should wait for service and margin evidence before recognizing much of the contract. The market may rationally capitalize part of the signed demand now. The fragile assumption in the bull case is more consequential for a long at this price: that contract value converts to shareholder earnings at margins high enough to justify the front-loaded $5.5 billion capacity investment and the customer equity option. No contract-level contribution-margin schedule is disclosed.

The $6.1 billion difference between contract price and estimated capex is not a margin estimate. It omits power, support, network operations, financing, taxes and maintenance. Without that bridge, direction may be right while the stock still de-rates as cash is deployed or the ramp slips. That is why the evidence supports research interest but not a 10/5 long.

What Could Go Wrong

Capacity delivery, colocation or memory procurement is delayed or costs more than the disclosed schedule; service availability conditions defer payment.
Anthropic workload demand, purchasing patterns or service economics change before the run-rate arrives; the maximum additional $9 billion is optional expansion at mutually agreed terms, not guaranteed revenue.
The $5.5 billion capex forecast excludes some power, network, maintenance or financing demands relevant to total return.
Legacy Delivery revenue continues to decline, masking new CIS growth or pressuring consolidated margins.
The warrant vests on first payment and later commercial milestones, creating an additional equity claim; the exercise strike does not eliminate dilution if in the money.
Rising debt/cost of capital, customer concentration, security incidents, AI hardware oversupply or competition could lower multiples even if the contract remains active.
Pre-market prints are not regular-session liquidity; gap reversal, spread, depth, venue, volume quality, halt risk, slippage and exit capacity remain untested.

What Would Prove This Wrong

The Reject would be too conservative if Akamai begins service on the stated schedule, delivers the planned capacity without materially exceeding the $5.5 billion investment estimate, records customer receipts and CIS revenue as scheduled, and reports a transparent positive contribution margin after energy, support, maintenance capex and financing. It would then need a share/debt bridge showing the contract’s net value accrues to common shareholders after warrant and other claims. Those facts could support a new long model; a headline commitment or stock-price jump alone would not.

Risk Audit

The main risks are contract execution, front-loaded capital, margin visibility, power/supply-chain availability, customer concentration, financing, warrant vesting, legacy delivery erosion, and a valuation already capitalizing future services. The June 30 cash, debt and share data predate the September deal. The contract’s terms are subject to termination and service conditions. A failed test, late delivery or margin shortfall can create a gap beyond 5%; a stop cannot insure that path. No live pre-market spread, depth, venue, market-quality or exit-liquidity evidence was reviewed.

Best Trade Strategy

No trade. Set entry.price=null and execution.can_execute=false. Keep $133.70 separate as a pre-market reference. Reassess after a fresh regular-session quote and after filed reports show service start, recognized Anthropic revenue, capex, cash flow and contract contribution economics. Reconcile warrant vesting/exercise with the fully diluted share count. Common stock only for any future analysis; no options, leverage, margin, market orders or price-floor language.

Sources

Akamai September 24 $11.6B Anthropic agreement, issuer release; contract value, expansion option and warrant summary.
Akamai September 24 Form 8-K, SEC filing; Project Plans, service/termination terms, warrant tranches and exercise conditions.
Expanded strategic relationship with Anthropic presentation, issuer presentation dated Sep. 24; capex, service start, revenue ramp and capacity estimates.
Akamai Q2 2026 Form 10-Q, SEC filing for June 30; cash/securities, debt, shares, operating cash flow, capex and segment performance.
Akamai Q2 2026 results, issuer 8-K and exhibit; FY26 guidance and prior-quarter segment results.
AKAM Sep. 24 historical market data, timestamped close, after-hours and pre-market observations; market-data feed discrepancies are disclosed.
AKAM Sep. 25 price and company overview, market-data snapshot observed at about 8:36 a.m. EDT; $133.70 reference only.
GRAIL Sep. 23 FDA panel release, issuer release; candidate comparison and advisory votes.
GRAIL Sep. 24 close, historical market data; candidate comparison.
H.B. Fuller Q3 FY2026 results, issuer release; candidate comparison.
H.B. Fuller Sep. 24 close and market statistics, historical market data; candidate comparison.

Research Quality Scorecard

Criterion	Score	Evidence and deduction
Market disagreement	4/5	New contract and abrupt repricing against capex, service timing and warrant terms
Evidence base	4/5	Fresh SEC 8-K and issuer schedule; contract contribution margins and service terms remain incomplete
Positioning and flows	3/5	Regular-close/pre-market repricing observed; current holder, borrow and dealer flow are unknown
Catalyst path	4/5	First service timing and revenue/capex phases are stated; commercial proof is mostly 2027–2028
Payoff architecture	2/5	Guide-based sensitivity fails the long hurdle; contract-level earnings/FCF cannot yet be modeled
Invalidation discipline	4/5	Delivery, cash, capex, revenue, margins and warrant state are monitorable
Differentiated insight	4/5	Separates committed gross revenue, capex schedule, service ramp and legal warrant trigger
Client value	4/5	Explains why a major signed deal can be real and still not support this entry after the gap
Total	29/40	No-trade screen; failed 10/5 economics take precedence over the Watchlist score band

Bottom Line

The new Anthropic agreement is not merely a pipeline slide: it is a large SEC-disclosed commitment with a detailed infrastructure schedule. But the market’s pre-market repricing arrives before Akamai spends the estimated $5.5 billion, starts service, and reports Anthropic revenue at scale. The $133.70 reference is already about 19.9× FY26 midpoint adjusted EPS, while this six-month base case is essentially flat and downside remains wider than 5%. The contract deserves a new model when delivery and margin data arrive. Today it does not clear the Desk’s long hurdle. Reject / no trade.

AI Illustration Prompt

Create a restrained institutional editorial illustration for The Mispricing Desk: a seven-year cloud-service contract ledger stamped “ANTHROPIC COMMITMENT: $11.6B” feeds into a distributed Akamai data-center network. In front of the network show a construction/capital ledger reading “CAPEX: Q4’26 $1.7B / FY27 $3.1B / FY28 $0.7B / TOTAL ~$5.5B,” with a delayed revenue curve marked “FY27 $150–300M / FULL ~$1.7B RUN-RATE BY END-2028.” Add a separate paper instrument labeled “WARRANT: UP TO ~5% AS-CONVERTED / $111.33 EQUIVALENT STRIKE / TRANCHE CONDITIONS,” avoiding any implication that all shares are issued today. A muted price panel shows “PRE-MARKET REF $133.70 / REGULAR CLOSE $110.41 / +21.1% GAP” and a small scenario strip “TOP $155.10 / BASE $134.50 / BOTTOM $105.60 / REJECT.” Use graphite, cream, restrained cobalt and amber, precise document-like labels, no rockets, chips, generic candlesticks or logos. Add a subtle readable “The Mispricing Desk” watermark. Wide 16:9 editorial composition.