Built by Claude Fable

Akamai Technologies: 4.05/10 | $88.85

Deep dive pages: Overview | Business Model | Financial Trends | Thematic Exposure | Management Quality | Investor Sentiment | Concerns, Catalysts & Risks

Prepared 2026-10-19. Price $88.845 and market cap $12.77B from the FMP /stable quote (2026-10-19); EV $20.63B from FMP key-metrics TTM. Fundamentals from Daloopa (company_id 251), cited inline. Fiscal year = calendar year. Last print FY2026Q2 (2026-08-06); next print FY2026Q3 on 2026-11-05.

Composite 4.05/10. Gate flag: Below Quality Bar (Requires Exceptional Catalyst). Recommendation: AVOID for now, revisit after the 2026-11-05 print.

Bottom line: Akamai has a well-run management team that guides accurately (14 of 14 quarters at or above range), and it is turning itself into a contracted CPU/GPU capacity lessor with Anthropic as the anchor customer. It does not lead any segment. It co-leads WAAP and CDN behind a much faster-growing Cloudflare, and it has well under 1% of IaaS. Its reported financials are getting worse fast as it pays for the build up front: non-GAAP operating margin is down about 590 bps in three quarters, TTM FCF is down 31% from its peak, and convertible debt is up 84% YoY. Two of the three quality gates fail (oligopoly and positive, growing FCF), which caps the score at 5.5. The weighted math comes in lower anyway, at 4.05. AKAM trades at about 10x FY2027 EV/EBITDA vs a ~19x peer average, but that multiple ignores the build: FY2026 capex of about $3.5B turns FCF negative, so P/FCF offers no support. Management and the street really do disagree about what these contracts will earn. Under our principles, though, that makes it a catalyst trade, not a quality compounder.

Gate flag: Below Quality Bar (Requires Exceptional Catalyst). Two of the three hard gates fail: oligopoly NO and positive, growing FCF NO. Management track record is YES. The 5.5 cap does not bind because the raw score is already 4.05. The debt-funded capex and falling FCF show up in Dims 1, 3 and 5 and in the FCF gate; that is one underlying fact (the CIS build), and here it really is the thesis.

Verdict: under-appreciated as distribution, not as raw compute.

• Footprint. About 4,300 locations in 700+ cities and 130 countries. Serverless runs in all 4,300 locations, managed containers in "well over 100" cities, full IaaS in "several dozen", and Blackwell-class (RTX Pro 6000) GPUs in "a couple dozen" (Q1 and Q2 2026 calls). GPU capacity is "completely sold out" (Q2 2026 call).
• Versus neoclouds. Akamai does not disclose server count, MW or GPU count, so a per-site or per-$ comparison with CRWV, NBIS and IREN cannot be done honestly, and none is invented here. Core sites are 5 to 10 MW, expandable to 20 to 30 MW, while neocloud campuses run to hundreds of MW. On the number of sites where it can place and bill compute, Akamai beats every neocloud and is in the same range as Cloudflare (330+ cities).
• Capex. Capex jumped to $346.5M in Q2 2026 (32% of revenue), and Q3 is guided to $475M to $525M. 2026 includes ~$250M for Inference Cloud plus ~$1.7B of component pre-buys for Anthropic. FY2026 capex reconciles to about $3.5B: the August guide of ~40% of revenue (~$1.8B) plus the 8-K's ~$1.7B.

Verdict: partly a "sneaky neocloud". The contract looks like a neocloud deal sitting on a security/CDN multiple, but it is CPU capacity, not frontier GPU, and Anthropic could be 25% to 30% of revenue by 2028.

• Terms (8-K filed 2026-09-24, verified). Master Services Agreement with Anthropic dated May 5, 2026. Project Plans 2 and 3, signed Sept 18, 2026, total about $11.6B over 7 years for dedicated, mainly CPU, capacity, and can expand by up to $9B more (~$20B).
• Build cost. About $5.5B of related capex plus ~$1.7B of 2026 memory pre-buys (through Jabil). The $1.7B is additive to the August 2026 capex guide. Whether it also counts toward the $5.5B build is not stated in the 8-K, and that swings the deal return materially.
• Warrant. Anthropic gets a warrant on 7.74M as-converted shares (~5%) at $111.33; about 2% vests with the base commitment.
• Revenue timing. No 2026 revenue impact per the 8-K. Press and Morgan Stanley put revenue at $150M to $300M in 2H27, rising to a ~$1.7B run rate by end 2028 (unverified). Anthropic can terminate a plan on a material outage, subject to conditions.
• The May 2026 deal. A $1.8B, 7-year "frontier model company" contract, with $20M to $25M of revenue starting Q4 2026. The 8-K confirms Anthropic as the counterparty under an MSA dated May 5, 2026, two days before this deal was announced, and Plans 2 and 3 imply a Plan 1 that matches it. The 8-K does not size Plan 1.
• Sizing. $11.6B / 7 = ~$1.66B a year, about 5x FY2025 CIS revenue of $313.9M.

The open question is the return, not the demand. On management's own margin framework (~$1.66B/yr revenue, mid-60s to mid-70s cash gross margin), the PM Review estimates roughly a 12% pre-tax unlevered IRR on $5.5B of capex, falling to roughly 4% to 5% if the $1.7B pre-buy is additive, before the back-loaded ramp and the warrant. Independent work puts project IRRs at 7.4% to 10.7%. Until the IRR is shown above the cost of capital, the street's skepticism on the economics looks earned.

Verdict: Akamai is relevant through Security (55% of revenue), not through the CDN. Delivery is billed on committed traffic volume at unit prices that reset lower at renewal; Security is sized to the protected estate and threat surface, so bot, API and WAF demand grows with agent traffic. Leighton himself said agent traffic "does not make a huge swing in bits delivered" unless it is video (Q1 2026 call). Cloudflare's developer funnel is better placed to capture new apps.

Illustrative sensitivity, not a forecast. This is the single version used on the Overview, Thematics and Valuation pages: a five-year horizon from FY2025, calibrated to Akamai's own history. Akamai does not disclose traffic, so revenue per unit of traffic cannot be measured directly. The anchor is that Delivery revenue fell 25% from FY2022 to FY2025 (-9%/yr) while traffic grew, so even 10x traffic is unlikely to lift Delivery; the Security cases bracket its own recent growth rates of 5% to 15% a year. See Thematics and Valuation for the full build.

Two NO answers set the maximum composite at 5.5. The raw weighted composite of 4.05 is already below that, so the cap does not bind, but the flag still applies.

What would change the score:

• Up: Q3 RPO steps up by ~$11B+, Q4 2026 CIS revenue shows the guided hockey stick (FY2026 CIS growth of at least 50%), and 2027 guidance shows margins steady at 25%+ with a credible path back to positive FCF in 2028. That would lift Financial Trends and Risks, but it would not clear the oligopoly gate.
• Down: the CIS ramp slips again, Anthropic's service start moves past 2H27, or more dilutive financing.

Cross-dimension reconciliation notes

• Anthropic is the confirmed counterparty per the Sept 24 8-K (MSA dated May 5, 2026), and the May $1.8B deal is Plan 1 by date. Plans 2 and 3 ($11.6B) are separate project plans, so the $1.8B is additive: ~$14.2B itemized, ~$14.4B on the company's own framing ($11.6B plus "more than $2.8B" of other 2026 CIS commitments).
• FY2026 capex is about $3.5B on every page: the August guide of ~40% of revenue (~$1.8B) plus the 8-K's ~$1.7B memory pre-buy. The 8-K does not say whether the $1.7B counts toward the $5.5B Anthropic build.
• Insider activity is one figure on every page: $17.1M of open-market sales (Form 4 code S) by 11 insiders from 2025-10-19 to 2026-10-19, $9.9M under 10b5-1 plans and $7.2M discretionary, with no purchases. $19.0M of tax-withholding dispositions (code F) are excluded. CEO buying in 2025 was $6.6M ($6.8M with the Chair).
• Debt is $7.64B of convertible notes principal ($7.56B carrying), about $9.4B including $1.78B of operating leases, against $4.62B of cash and securities (Daloopa, June 30, 2026). The unsourced "~$10B gross debt" figure was removed.
• The 10x traffic scenario uses one five-year table on the Overview, Thematics and Valuation pages, calibrated to FY2022 to FY2025 segment growth.
• Delivery YoY differs slightly across dims. Dims 1 and 2 use the standalone Delivery series (-7.8% Q1 26, -6.1% Q2 26), while Dim 5 uses Delivery plus other cloud apps (-6.6%, -5.8%). Both show the same mid single digit decline.
• No extra Daloopa pulls were needed. All five dimensions already cover 8 quarters and 5+ fiscal years, with src citations for every segment line (Security, Delivery, CIS, other cloud apps).

Reviewed 2026-10-19 against pm-review and investing-principles.

Verdict: AVOID stands, and the 4.05 composite is about right. AKAM is not the leader in any segment (Cloudflare grows ~6x faster in the same markets), so the principles rule it out as a core holding whatever the catalysts. The work is thorough, well cited, and strong on the guidance track record. It has three real weaknesses. (1) It never tests the deal economics that the whole sentiment and catalyst case depends on. (2) The dimensions disagree with each other on capex, insiders and the 10x traffic scenario. (3) The Concerns score rests on a valuation argument that ignores negative FCF and leaves out the true leader as a peer. Dims 4 and 5 are each about 1 point too generous. Dim 3 is mechanically harsh but the right call. The net effect on the composite is small (roughly 3.8 to 4.0), but the reasoning needs fixing before this goes out as a deep dive.

1. Gaps vs the pm-review checklist

YoY on every metric chart: not met in several places.

• Dim 4 "Daloopa trajectory" table: Total revenue, Delivery + other, RPO and Capex have no YoY rows. RPO is the most important forward metric in the whole document, and its growth is never stated. $7.6B vs $4.3B a year earlier is about +77% YoY. cRPO is never shown at all.
• Dim 5 quarterly concerns table: Revenue, Adj. EBITDA margin (no bps change YoY), Capex and FCF have no YoY. The Dim 5 annual table has no growth rows at all.
• Dim 5 peer table: multiples only, no peer growth. A 10x multiple next to +16% growth means nothing without the peers' FY2027 growth and margin beside it.
• Dim 3 guidance tables are fine. Dim 1 is the model to copy.

Trajectory in bps: mostly good, with one gap. Dim 1 does this well (-590 bps non-GAAP OM, -270 bps cash GM). The acceleration and deceleration of Security growth is never quantified in bps against the theme. Security grows +10% against a stated theme rate of +10% to +15%, which means AKAM is flat to losing share. No dimension says that outright.

5 years annual + 8 quarters: met at company level, not at segment level. Segment annuals start in FY2022, which gives only 3 years of segment YoY. Dim 3 cites FY2021 Security growth of 24.6%, but that number is not in any table. Rebuild the Security and Compute annual history back to FY2020 (the pre-2022 segment basis is in Daloopa) so the 5-year deceleration from ~25% to ~10% is shown, not just mentioned.

Explaining why trends move: good in Dim 1, weak on Security. The margin and FCF explanations are tied to the CIS build. Nobody explains why Security slowed from +14% to +10%. Is it WAF saturation, Cloudflare share gains, or a tough Noname/Guardicore comp? That deceleration matters more for the 18 to 36 month thesis than Delivery does, because Security is 55% of revenue.

Street consensus on key metrics: partial.

• Revenue and EPS consensus are shown. Consensus for FCF, capex, EBITDA margin and CIS/Security segment revenue is missing. Those are the lines the bull/bear debate is actually about.
• There is no estimate revision trend. Has FY2027 and FY2028 EPS consensus moved since the 2026-09-24 8-K? That is the most direct test of whether the street "does not believe" management. Without it, the Dim 4 divergence claim rests on a single snapshot.
• Consensus comes from FMP. The principles name Visible Alpha as the primary consensus source.

Sources after big claims: mostly met. These are not:

• "~$10B of gross debt" (Dim 5, used three times, including in the bear case). No source, and the Daloopa converts figure is $7.56B. Show the bridge from converts to total debt (leases plus anything else).
• The "~$18.5B EV if all cash is netted" arithmetic does not reconcile. $12.77B market cap + ~$7.6B converts + leases - $4.6B cash comes out nearer $16.5B to $17B unless leases are about $1.5B to $2B. Show the bridge.
• "FY2026 capex lands near $3.5B" (Dim 5) is the analyst's estimate and conflicts with Dim 1 (see below).

Price talk: violates the "don't talk about price" rule.

• The bottom line says "the stock is cheap."
• Dim 4.2 argues from price targets vs the current price ("every published target sits well above the current price").
• Dim 4.4 relies on Stocktwits and Wall Street Zen.
• Recast these as fundamental statements (e.g. multiple vs growth vs FCF yield) or cut them.

2. Gaps vs investing-principles

Leaders remain leaders (the gate that decides this name).

• The oligopoly NO is correct, but it is argued from weak data. WAAP share "~20% to 30%, unverified", a $17B CDN TAM from a source the analyst calls low quality, and W3Techs usage share.
• The one hard data point in the document is decisive and under-used. In Q2 2026, Cloudflare's total revenue was $696M growing ~30%, and Akamai's Security was $604M growing ~10%. Make this the headline gate evidence. Show Cloudflare vs Akamai revenue in the overlapping segments, YoY, for 8 quarters. If Akamai is ceding share in its best segment, that ends the "co-leader" framing.
• The principles say "don't settle for the 2nd in a market." Treat Akamai as the #2 that is losing ground, and say so plainly in the Bottom line.

Trajectory over absolutes: there is a contradiction across dimensions.

• Dim 1 frames FY2026 capex at "~40% of revenue" and says Q3 FCF "could be" negative. That uses the August guide.
• The 8-K adds ~$1.7B to 2026 capex. Dim 5 picks this up and estimates ~$3.5B of FY2026 capex, which is about 78% of revenue, with FY2026 FCF "sharply negative."
• These cannot both stand. Dim 1 understates the FCF trajectory it scores. The score is already at the 1/10 floor, so the number does not change, but the headline "TTM FCF down 31%" badly understates where FY2026 lands.
• It is also unresolved whether the $1.7B memory pre-buy is inside the $5.5B Anthropic capex or on top of it. Dim 2 says "plus," and the Bottom line says "plus." That one fact swings the deal's return materially (next point).

You are not contrarian, only management is. Dim 4 is over-scored at 7.

• The divergence is real, but the analyst never checks whether management or the street is arithmetically right. Here is a rough check on management's own terms: ~$1.66B/yr average revenue, 65% to 75% cash gross margin and low-20s to low-30s operating margin, which implies ~70% EBITDA-like cash margin, or ~$1.2B/yr of pre-tax cash.
  • On $5.5B of capex over 7 years, that is roughly a 12% pre-tax unlevered IRR.
  • If the $1.7B pre-buy is additive ($7.2B), it falls to roughly 4% to 5%.
  • The back-loaded ramp ($150M to $300M in 2027) and the ~5% warrant given to Anthropic push both numbers lower.
• That range brackets the independent 7.4% to 10.7% IRR the analyst cites but does not engage with. If the street's skepticism is supported by management's own numbers, this is not an NVDA-style "street doesn't believe" setup. It is a street that has done the math.
• The NVDA template needs management to be right on economics the street dismisses. Until the IRR is shown above the cost of capital, the divergence sub-score should be 6, not 8. That gives Dim 4 roughly 6.

Sentiment inversion (worse to better): not addressed.

• The document never asks whether this is the trough, or when the reported numbers inflect.
• Build the quarter in which, on consensus, non-GAAP operating margin stops falling YoY, and the year FCF turns positive again. Without that date, "revisit after 11/5" has no test attached. 11/5 can only show bookings and capex, not the inflection.

Management.

• The track-record work is the best part of the document.
• There is a data conflict on insiders. Dim 3 shows ~$17.1M of open-market sales by 11 insiders, including CFO McGowan at $2.0M, and fires a red flag on it. Dim 4 says the CEO and CFO had only vest and tax-withholding dispositions, and the other sales were 6 small 10b5-1 trades with the largest ~$0.88M. One of these is wrong.
  • If tax withholding was counted as selling, the insider flag may not fire. Dim 3 would still have 2 flags, and the cap and the score of 4 hold.
  • Even so, the document cannot publish two different insider stories.
• CEO buying is given as ~$6.6M in one place and ~$6.8M in another.
• Dim 1 and Dim 3 body text still say Anthropic is "reported, not confirmed" after the 8-K confirmed the counterparty. The consolidation note flags this, but the body text was never updated. Fix the text, not just the footnote.

Red flags: the same fact is counted four times.

• Debt-funded capex and falling FCF are penalized in Dim 1 (two penalties), Dim 3 (two red flags), Dim 5 (-2 drag), and the FCF gate.
• Here the conclusion survives because the underlying fact really is the thesis. Still, say this explicitly in the composite section, so a reader does not think the 4.05 is five independent negatives.

Valuation framework. Dim 5 is over-scored at 6.

• P/FCF (next year) is one of the four required forward multiples and is missing. On the analyst's own numbers FY2026 FCF is sharply negative, and FY2027 is likely negative too. So "cheap on every lens" is false on the lens that matters most for a capex-heavy lessor.
• NET is excluded as an "outlier," but NET is the market leader and the closest comp. Taking out the leader to show AKAM is cheap is circular. Show NET, then explain the gap with growth and FCF.
• CRWV, the levered GPU lessor, trades at 9.5x, below AKAM's 10.1x. The document treats that as an exception, but it is evidence for the bear case: the market is pricing AKAM as a lessor.
• The EBITDA caveat (FMP FY2026 consensus $1.94B vs a company-basis ~$1.75B) means the 10.1x is understated by 1 to 2 turns, which the analyst admits. Use company-basis EBITDA.
• Valuation sub-score: 6 to 7, not 8 to 9. That gives Dim 5 roughly 5.

Data sources.

• No Priority 1 internal sources were checked: no broker research on the internal drive, no Outlook broker color, no analyst models, no Teams.
• No alt data (Snowflake web traffic or developer adoption for Linode/Inference Cloud).
• No Daloopa expert-call transcripts. Those are the best available check on Anthropic's SLA and termination terms and on Akamai vs Cloudflare WAAP win rates.
• Market data and transcripts came from FMP rather than S&P Global. Web search (SA, Quiver, GuruFocus, Stocktwits) carries too much weight on the most important open questions.

10x traffic scenario: inconsistent and uncalibrated.

• Dim 2 says Delivery is "flat to +30% over 5 years" and Security "+30% to 60%." Dim 5 says Delivery is 0.7x to 1.5x and Security +30% / +60% / +100%, with no time horizon. The Bottom line quotes Dim 5.
• Neither is calibrated to Akamai's own history. Delivery revenue fell ~25% from FY2022 to FY2025 while traffic grew. Use that elasticity (revenue change per unit of traffic growth) to anchor the bear/base/bull cases, and pick one version for the whole document.

3. Action Items

1. Build the Anthropic unit-economics model. Use the ramp by year (2027 $150M to $300M, end-2028 $1.7B run rate), capex timing, management's margin framework, a 7-year depreciation life, and the warrant cost. Report IRR under two cases: $5.5B capex, and $5.5B + $1.7B. Re-score Dim 4 on the result. Over 10% supports management, so keep the divergence at 8. Under 8% means the street is right, so cut it to about 5.
2. Resolve FY2026 capex in one number. Confirm whether the $1.7B pre-buy is additive (8-K Ex. 99.1 text, or the 11/5 call). Update Dim 1's FCF narrative and gate evidence to the post-8-K figure. Add consensus FY2026 to FY2028 capex and FCF rows.
3. Add YoY rows everywhere they are missing: Dim 4 trajectory table (Revenue, RPO, cRPO, Capex), the Dim 5 quarterly and annual tables (Revenue, EBITDA margin bps, Capex, FCF), and peer FY2027 revenue growth and EBITDA margin in the peer table.
4. Do the leadership test with hard data. Put Cloudflare vs Akamai Security/WAAP revenue and YoY side by side for 8 quarters, add Gartner/IDC vendor share if available on internal research, and state whether Akamai is gaining or losing share. Explain the Security slowdown from +14% to +10% with management commentary and citations.
5. Fix the valuation. Add P/FCF and EV/(EBITDA - capex) for FY2027 and FY2028. Put NET back in the peer set. Use company-basis adjusted EBITDA. Show the EV bridge (converts, leases, cash) with sources, and replace the "~$10B gross debt" figure with a sourced number.
6. Reconcile insiders. Pull Form 4s directly from SEC EDGAR. Separate open-market sales from 10b5-1 sales and from tax withholding. Publish one table and re-test the >$10M flag.
7. Clean the stale text. Update the Anthropic language in Dims 1 and 3 to "confirmed counterparty per 8-K." Make the 10x traffic scenario consistent across Dims 2 and 5 and calibrate it to FY2022 to FY2025 Delivery elasticity.
8. Remove price-based reasoning: price targets vs price, "the stock is cheap", and Stocktwits. Replace with fundamental framing.
9. Add consensus revision tracking. Show FY2027 and FY2028 revenue, EPS and FCF consensus before vs after 2026-09-24 (Visible Alpha). This is the direct evidence for or against "the street doesn't believe management."
10. Define the inflection test for 11/5. Name the specific numbers that would move the score: Q3 RPO including the $11.6B, a Q4 CIS guide implying 50%+ FY growth, an FY2027 framework with operating margin of 25% or more, and a stated date for FCF to turn positive. Also name the quarter on consensus when operating margin stops compressing YoY.
11. Check internal sources first: broker research on the drive, Outlook broker color, Daloopa expert calls on Anthropic SLAs and Akamai vs Cloudflare win rates. Then use Snowflake alt data on Linode/Inference Cloud adoption. Then web search.

Indicative re-score if nothing else changes:

AVOID either way. The upgrade path runs through the IRR test and share data, not through the 11/5 headline numbers.

PM Review follow-ups applied (2026-10-19 dashboard pass)

• FY2026 capex reconciled to about $3.5B (August ~40%-of-revenue guide plus the 8-K's ~$1.7B pre-buy) and used on every page. Whether the pre-buy sits inside the $5.5B build is still not disclosed.
• Insiders reconciled from Form 4s: $17.1M of open-market sales by 11 insiders, $9.9M under 10b5-1 plans, no purchases; tax withholding excluded. The CFO sold $2.0M under a 10b5-1 plan, so the earlier "vest and withholding only" read was wrong. The Dim 3 red flag stands.
• Anthropic wording updated to confirmed counterparty per the 8-K on every page; the stale update banners were removed.
• RPO and cRPO YoY added (Business Model, Sentiment, Valuation): RPO +77% YoY, cRPO about +18%.
• P/FCF added: 20.3x TTM, not meaningful for FY2026 on negative FCF.
• Debt sourced: $7.64B of converts, about $9.4B with leases; the EV bridge is about $17.5B.
• 10x traffic scenario unified across pages and calibrated to Akamai's own segment growth history.
• Leadership test: Cloudflare's total revenue passed Akamai Security in Q4 2025 and grows about 3.8x faster (8-quarter table on Business Model).
• Not done in this pass: the full Anthropic IRR model, Visible Alpha consensus revisions, and a re-score. The composite stays at 4.05.