Built by Claude Opus

The Bull Case: A 10x Internet Runs Through Akamai

Akamai already sits in front of a large share of the world's most valuable web and API traffic: about 4,300 locations in 700+ cities, protecting and delivering for banks, retailers, media companies and governments. It gets paid for that position three ways. It inspects traffic (Security, 55% of revenue), moves it (Delivery, 27%), and increasingly computes on it (Cloud Infrastructure Services, 9%).

A 10x internet means 10x more requests to classify, 10x more bytes to carry, and far more inference that wants to run close to users. The bear case says only one of those three meters scales. The bull case says all three scale at once, for the first time since the early CDN era, and the market is pricing none of it: the stock trades at 6.7x the street's own FY2030 EPS.

Why 10x is not crazy. 10x in five years is about +58% a year. Cloudflare Radar showed automated requests passing human ones on HTML traffic in June 2026 (57.5% vs 42.5%), about 18 months ahead of its CEO's forecast. HUMAN Security measured agentic AI traffic up 7,851% YoY in 2025 (off a tiny base) and automated traffic up 23.5%, against 3.1% for humans. A China Internet Conference keynote projects global IP traffic growing 60% to 90% a year to 2030, with agents about 60% of it. Cisco is more conservative (enterprise traffic up to ~9x by 2035 with agentic AI), and it found that agent-run tasks generate ~450% more traffic than the same task done by a person. 10x by 2030 is the aggressive end of that range. It is a premise, not a forecast.

The bull case starts from a weak quarter, not a strong one. Revenue grows ~5%, margins are falling as the CIS build gets paid for up front, and FCF is shrinking. What is growing is the forward book: RPO is up 77% YoY to $7.6B, before most of the $11.6B Anthropic commitment shows up.

Source: Daloopa (company filings). YoY computed from cited values.

• Security has held at +9.5% to +11% for four quarters. That is the number the bull case most needs to move. The agent-exposed pieces already grow much faster: API security grew >100% in 2025 to a >$100M run rate (Q4 25 call).
• CIS has compounded at ~40% for six straight quarters while GPU capacity sold out, and that is before any Anthropic revenue.
• The cost is real. Non-GAAP operating margin fell from 31% to 25% in three quarters, and Q2 26 capex of $346.5M was 32% of revenue. The bull case is a bet that this is the trough of an investment cycle, not a new normal.

Seven-year view: traffic grew, Delivery revenue did not

Source: Daloopa (company filings). Segment reporting starts FY22; CIS is broken out from FY24.

This table is the bear case's best evidence and the bull case's starting point. From FY22 to FY24, bandwidth fees rose from $205.3M to $233.1M (+13.6%) while Delivery revenue fell from $1,669.3M to $1,318.1M (-21%). Akamai carried more bytes and got paid less for them, because price resets at renewal beat volume growth every year. The bull case does not need that to reverse. It needs volume growth to jump from internet-normal (~20% to 30% a year) to agent-driven (~58% a year) while price resets stay as harsh as they have been.

This is the highest-confidence pillar. Every agent that hits a bank, retailer or airline has to be classified: who sent it, whether it is authorized, and whether to serve, throttle, charge or block it. That is bot management, API security and WAF, and Akamai is a Gartner leader in this market. Security is priced on the protected estate and the request load, not on bytes, so it avoids the Delivery price-reset problem.

• Attack surface scales with agents. Leighton (Q1 26) said AI "is enabling attackers to take over more devices and create enormous bot armies, we're now seeing attacks with unprecedented volumes," and that CISOs are "literally scrambling" to make sure "their applications, their agents, their APIs are protected by Akamai."
• Products aimed at agents. Agent management now sits alongside bot management. The LayerX browser-security deal covers employees' use of AI agents, and Akamai already serves agent-optimized content to authorized scrapers (Q2 26 call). That is the start of a paid "agent toll."
• Segment history supports the bull rate. Security grew +14.5% in FY23 and +15.7% in FY24 before slowing to +9.8% in FY25. The bull case asks for +17% a year to FY2030, about 1 point above the FY24 peak, driven by request volume growing several times faster than it did in FY23 and FY24.

Bull math: FY25 Security of $2,243.4M compounding at +17% a year reaches $4.92B in FY2030. Base is +10% ($3.61B), about today's rate. Bear is +5.4% ($2.92B), where Cloudflare keeps taking share.

This is the most contrarian pillar, and the one with the most upside if it works. Delivery revenue = traffic x unit price. Akamai does not disclose traffic, so the unit-price decline has to be inferred. If traffic grew 20% to 30% a year from FY22 to FY25 while Delivery revenue fell ~9% a year, unit prices fell ~24% to 30% a year. Hold that price decline constant and lift traffic growth to +58%:

Illustrative. The historical traffic rates are assumptions, not disclosures.

The math says Delivery grows 11% to 20% a year in a true 10x world. The bull case uses only +8% a year ($1,256.7M to $1.85B), because multi-CDN customers move bytes to the cheapest vendor and Cloudflare will win much of the new volume. Leighton's own caveat shows where the volume would come from. Text agents do not move many bits, but agents that generate video do. His example is a retail site generating a video of a shopper wearing the sweater they are looking at, which he said "generates a lot of traffic." Akamai's media and commerce customers are exactly where that traffic would land.

Delivery is also where the operating leverage is. The network is already built, so every extra byte that fits on existing capacity carries a very high incremental margin. Even flat-to-up Delivery revenue lifts the consolidated margin.

In agent workflows, most of the traffic is inference: Cisco puts it at ~70%. Agents acting for a user in real time (checkout, voice, video, robotics) want compute close to that user, not in a distant hyperscaler region. Akamai's pitch is to run "agents and models within milliseconds of their end users" on a footprint no neocloud matches in site count. Customers are buying it: a US consumer AI platform runs "live personalized agents" on Akamai cloud (Q1 26 call).

The contracted base is real: ~$14.2B of signed multi-year CIS commitments, led by Anthropic's $11.6B, 7-year CPU deal (8-K, 2026-09-24), with an option for up to $9B more.

FY25 CIS was $313.9M. The base case assumes only the contracted floor ramps (~$2.3B). The bear case assumes Anthropic slips or partly terminates (~$1.2B). Management guides these contracts to low-20s to low-30s non-GAAP operating margins at maturity, and the bull case takes the top of that range.

Author's estimates, not company guidance. Street FY2030 from FMP (/stable, 2026-10-19), 6 analysts. EPS assumes a 19% non-GAAP tax rate and ~$150M of after-tax net interest. The share count rises to ~160M (from 148.3M today) for the Anthropic warrant and convert dilution, so buybacks only offset dilution. Value is the average of the P/E and EV/EBITDA methods. EV/EBITDA uses adj. EBITDA margins of 36% (bear), 42% (base) and 47% (bull), with ~$8B to $9B of net debt. Present value discounts from end-2030 at 10%.

Probability-weighted, the stock is roughly fairly priced. The payoff is lopsided. Weighting 20/50/30 gives a present value of about $94, ~5% above $89.63, so the market is pricing something close to the base case. The bull case is worth +142% in present-value terms and the bear case -68%. This is a bet on the premise, not on the multiple. The bull case gets only 20% because it needs all three pillars at once, and two of them (Security reaccelerating and Delivery turning positive) have no support in the reported numbers yet.

Where the bull case is weakest: cash. Even in the bull case, FY2030 capex runs about 25% of revenue (~$2.8B), so FCF is roughly $1.65B, or ~$10 a share. That puts the $325 value at ~31x FCF. The EPS is earned on a capital-heavy base, and anyone underwriting the 20x P/E should know that.

1. Security growth rate. Each extra point of Security CAGR adds ~$0.2B of FY2030 revenue at the company's highest margin. This matters more than anything else.
2. Anthropic delivery and the expansion option. About $2.9B of the bull case's $4.0B CIS line depends on one counterparty that can terminate a plan over a material outage.
3. The Delivery price-reset rate. If unit prices fall 35% a year instead of 27%, the +58% volume case only gets Delivery to flat. That decides whether pillar 2 adds value or adds nothing.
4. Margin recovery timing. Non-GAAP operating margin has to get from 25% back above 30% by FY2028, or the EPS bridge breaks before the multiple matters.
5. Cloudflare's share of new agent traffic. Cloudflare's developer funnel is better placed for new apps. Akamai's edge is its Global 2000 installed base. The bull case needs Akamai to hold its share of agent traffic at its existing customers.

The honest summary: the Security and contracted-compute legs are plausible, and the Delivery and re-rating legs are long shots. If traffic really goes 10x, Akamai is one of the few companies paid on all three meters (inspect, move and compute), which is why the upside is ~3.6x. The reported numbers do not show it yet. That is why the scored view stays AVOID, and why the first real test is the 2026-11-05 print.