Akamai Technologies, Inc. (AKAM) - Deep Dive Research Report

Prepared 2026-05-24. Concalls covered: Q1 2026 (May 7, 2026), Q4 2025 (Feb 19, 2026), Q3 2025 (Nov 6, 2025), Q2 2025 (Aug 7, 2025).

1. What the Company Does

Akamai operates a globally distributed network of roughly 4,400 points of presence (PoPs) in 700-plus cities, sitting one network hop away from most of the world's internet users. Customers pay Akamai to do three things on top of that footprint: move bits closer to end users (the original CDN business they called Delivery), inspect and block bad traffic before it reaches the customer's origin (Security), and run general-purpose virtual machines and GPU-backed inference workloads on the same edge infrastructure (Compute, branded Cloud Infrastructure Services or CIS).

The company was incorporated August 20, 1998 in Cambridge, Massachusetts. It grew out of MIT applied-math research on "consistent hashing" - a way of distributing internet content across thousands of servers without any single coordinator deciding which server holds what. Founders Tom Leighton (an MIT applied-math professor, still CEO today at age 69) and his graduate student Daniel Lewin built a prototype for the 1998 MIT $50K business plan competition. The technology let websites mirror their content at the network edge, dramatically speeding up page loads. Akamai IPO'd in October 1999 at the peak of the dotcom bubble. Daniel Lewin was killed on September 11, 2001 aboard American Airlines Flight 11; the company nearly went under in the 2001-2002 dotcom crash. It survived because the underlying technology was genuinely useful and customers stayed.

For two decades after that, Akamai was the dominant CDN. The strategic pivot started in the mid-2010s. Three things were happening at once: (1) hyperscalers (AWS CloudFront, Google Cloud CDN, Cloudflare) began commoditizing delivery, (2) cybersecurity demand exploded as enterprises moved more traffic to the public internet, and (3) Akamai realized its 4,000+ edge locations were a unique asset for compute workloads that needed to be close to users. The company spent ~$2 billion on acquisitions to pivot: Prolexic in 2014 ($390M, DDoS mitigation), Guardicore in 2021 ($600M, microsegmentation), Linode in 2022 ($900M, Infrastructure-as-a-Service), and Noname Security in 2024 (~$450M, API security).

The core value proposition today is this: Akamai owns physical infrastructure inside more telecom carrier networks, in more countries, than any other compute or security vendor. Hyperscalers run a few dozen massive data centres; Akamai runs thousands of small ones. For workloads where latency-to-end-user matters - real-time AI inference, video, web acceleration, in-line security inspection - that distribution is hard to replicate even by trillion-dollar competitors. In May 2026 this thesis was vindicated when Anthropic signed a $1.8 billion, seven-year commitment for distributed inference capacity on Akamai's network, the largest customer contract in company history.

"I do not know of a comparable time where there is this much concern about what is going to happen with security." - Tom Leighton, Q1 2026 concall

"Our architecture uniquely positions us to power and protect AI...bringing AI physically close to users." - Tom Leighton, Q4 2025 concall

2. Business Segments

Akamai reports three product categories. They are not separate legal entities - they share the same underlying global edge network, sales force, and operating overhead - but management discloses revenue and growth separately for each, and the strategic stories diverge sharply.

2.1 Security (~55% of FY2024 revenue)

This is the profit engine and Akamai's largest segment, comprising web application firewall (WAF) and DDoS protection (legacy strength built from Prolexic), bot management, API security (the Noname Security 2024 acquisition), and Zero Trust microsegmentation (the Guardicore 2021 acquisition). Security revenue was $2.04 billion in FY2024 (+16% YoY) and grew to $590M in Q1 2026 (+11% YoY).

The core capability is that Akamai sees enormous volumes of attack traffic precisely because its edge network sits in front of so many websites. That visibility trains better detection models, which makes the product more effective, which wins more customers. The newer products - API Security (Akamai's protection against attacks targeting machine-to-machine APIs) and Guardicore Segmentation (limiting lateral movement inside enterprise networks once attackers get past the perimeter) - grew "over 100% YoY" and "35-36% YoY" respectively in late 2025. API Security crossed $100M annual run rate in Q4 2025.

The competitive position is mixed. In WAF and DDoS, Akamai competes head-on with Cloudflare (more developer-friendly, cheaper at the low end) and Imperva. In Zero Trust microsegmentation Guardicore is a recognized leader - Gartner named Akamai "Customers' Choice" for Network Security Microsegmentation in 2026. In API Security, the field is more fragmented (Salt Security, Traceable, Wallarm). Akamai's pitch is integration: an enterprise that already has Akamai's WAF in line can add API Security without reconfiguring DNS or routing. Management has repeatedly noted that API Security has less than 10% penetration within Akamai's existing customer base, framing it as the segment's nearest-term growth lever.

Within the group, Security is the cash cow funding the Compute capex build-out.

2.2 Cloud Infrastructure Services / Compute (~16% of FY2024 revenue, accelerating)

CIS is the growth bet. Born from the $900M Linode acquisition in February 2022, this segment offers virtual machines, managed Kubernetes, object storage, and (from late 2025) GPU-backed AI inference - sold either as hourly VM rental (the Linode legacy model, attractive to mid-market developers) or as multi-year committed-capacity contracts (the model now driving hyperscaler-scale deals like Anthropic). FY2024 revenue was $630M (+25%); CIS specifically (the strategic sub-bucket excluding legacy compute) grew 40% YoY in Q1 2026 to $95M.

The core capability is the distributed footprint. Hyperscalers concentrate GPUs in a handful of US locations; Akamai is placing NVIDIA RTX PRO 6000 Blackwell GPUs across dozens of cities, with hourly VM access in 100+ cities and full IaaS in several dozen. For real-time inference - chatbots, gaming AI, autonomous vehicle workloads, real-time translation - milliseconds matter and physical proximity to the user is the lever no amount of bandwidth can substitute for.

Competitive position: against AWS, Azure, and GCP, Akamai is roughly two orders of magnitude smaller and cannot match their managed-service breadth. But for inference workloads specifically, Leighton claimed in the Q4 2025 concall that "three major hyperscalers" are now Akamai customers for compute, using Akamai's network because their own internal infrastructure cannot match the distribution. The Anthropic deal extended that pattern to a frontier AI lab.

Strategically, this is the segment management talks about most. Capex of $700M in 2026 is being deployed specifically to fulfil the Anthropic contract.

2.3 Delivery and Other Cloud Apps (~33% of FY2024 revenue, declining)

The original CDN business, plus a handful of acquired customer contracts (Edgio, Lumen, StackPath assets bought during 2024 as those competitors exited the market). FY2024 revenue was $1.32 billion (-15% YoY); Q1 2026 was $389M (-7% YoY).

The core capability is unmatched scale at carrier interconnects, but the business is in secular decline. Hyperscalers bundle CDN into broader cloud deals at near-zero incremental cost; Cloudflare prices aggressively at the SMB end; large media customers self-build. Akamai's response has been pricing discipline rather than chasing market share.

"Pricing environment remains competitive. We still have folks selling at very low prices, which we won't do." - Tom Leighton, Q4 2025 concall

Management has signalled the decline rate will moderate to mid-single-digits through 2026 as the Edgio acquisition lap effect washes through. The segment still generates significant cash flow that funds Compute capex; it is being managed as a melting ice cube, not a growth engine.

3. Products and Business Detail

The full product surface is unified under the brand "Akamai Connected Cloud" - a marketing umbrella for the underlying edge platform, which is what the company actually sells. Underneath, the catalogue clusters around the three segments above.

Delivery and acceleration. Ion (web acceleration), Adaptive Media Delivery (live and on-demand video streaming), Download Delivery (large-file distribution like software updates and game patches), API Acceleration, Image & Video Manager (automatic format conversion and compression). Customers pay primarily on egress bandwidth, with tiered pricing that drops sharply at high volume - the structural reason hyperscalers can underprice Akamai at the very high end and Cloudflare can underprice at the low end.

Security. App and API Protector (the integrated WAF + bot + API security suite), Account Protector (credential stuffing defence), Bot Manager Premier, Prolexic (DDoS scrubbing centers globally), API Security (the Noname-derived product, deployed independently of WAF for API discovery and runtime protection), Guardicore Segmentation (server-to-server microsegmentation, sold to enterprise infrastructure teams independent of the delivery business), Enterprise Application Access (Zero Trust network access), Secure Internet Access. The newest additions, announced through 2025-2026: Firewall for AI (a WAF tuned for prompt injection and LLM-specific attacks) and AI Gateway. Pricing is mostly subscription with traffic-based components.

Compute. Linode-derived VMs ("Akamai Cloud Computing"), managed Kubernetes (LKE), object storage, block storage, and the headline new product Akamai Inference Cloud - GPU capacity for AI inference workloads, sold in two SKUs: hourly Blackwell VM rental and multi-year committed-capacity contracts. AI Grid is the orchestration layer that schedules inference requests across the 4,400-PoP network. Cloud Infrastructure Services was sold out in beta across the initial 20 cities through Q4 2025.

The physical infrastructure spans roughly 4,400 PoPs across 700-plus cities. Each PoP houses Akamai-owned servers running Akamai-developed software, typically co-located inside telecom carrier data centres on long-term lease arrangements that took two decades to negotiate at scale. The compute build-out adds NVIDIA RTX PRO 6000 Blackwell GPUs into a subset of those PoPs. By management's own framing, Inference Cloud now spans all 4,300 edge locations for the lightest workloads; managed containers run in 100+ cities; full IaaS runs in several dozen cities.

Akamai sells in essentially every meaningful market. International revenue was 49% of total in Q1 2026. The company has been multi-continent since the early 2000s; there is no "geographic expansion story" left to tell at the segment level.

4. Customers

Akamai serves roughly 9,000 enterprise customers. The mix skews heavily to large enterprises: banks, brokerages, streaming services, large e-commerce platforms, airlines, gaming publishers, government agencies, and increasingly the hyperscalers and AI labs themselves. Public references include Apple, Microsoft, Adobe, IBM, CBS, Turner Broadcasting, General Motors, NBC, and Paramount, though no single customer breaks the 10% concentration disclosure threshold in recent 10-Ks.

The buyer depends on segment. For Delivery, the decision sits with the head of engineering or VP of infrastructure at a digital business; the sales cycle is short (weeks) and competitive on price. For Security, the buyer is the CISO or head of application security; sales cycles run months because regulated industries require proof of concept, penetration testing, and compliance certifications (SOC 2, FedRAMP, PCI). For Compute and especially Inference Cloud, the decision involves both engineering and the CFO, sales cycles run quarters, and contracts are typically multi-year committed-capacity arrangements with hundreds of millions of dollars at stake.

Why customers buy Akamai over alternatives varies by segment. For Delivery, it is performance at scale and reliability - the largest streaming events in the world (royal weddings, Super Bowls, major league playoffs) have historically run on Akamai because it can absorb traffic surges that would crush smaller networks. For Security, it is the integrated portfolio: a bank that already trusts Akamai for delivery can layer WAF, bot defence, and API security on top without changing DNS or routing. For Compute, the pitch is geographic distribution that hyperscalers cannot match.

Switching costs vary. Delivery is relatively easy to swap (DNS change), which is part of why pricing is competitive. Security is much stickier: WAF rules are tuned per-customer over years and rebuilding them on another vendor is a major project. Microsegmentation, once deployed inside an enterprise's data centre, is essentially permanent until the next infrastructure refresh cycle. Inference Cloud contracts are committed-capacity arrangements that lock both sides in for multi-year horizons.

Concentration is moderate but rising. The Anthropic deal alone, when fully ramped, will represent roughly $257M of annual revenue - meaningful but not dominant against a ~$4.5B base. Management has highlighted recent multi-year deals with North American financial institutions, an Asian airline, gaming publishers, and three of the four major hyperscalers as Compute customers. The shift to large committed-capacity deals improves revenue predictability but increases concentration risk over time.

Contract structure: Delivery is mostly traffic-based pay-as-you-go with annual commits. Security is largely subscription with bandwidth components. CIS is splitting in two - small/mid Linode-style hourly billing on one side, and committed-capacity contracts (often $40M-$1.8B over 4-7 years) on the other.

5. Competitive Landscape

Akamai sits in three overlapping competitive arenas, and the answer to "who is the real competitor" depends entirely on which segment you mean.

In Delivery, Akamai is one of three or four giants. Cloudflare is the disruption story - founded 2009, started cheap and developer-friendly, now hosts an estimated 80%+ of CDN users by site count even though Akamai still leads on enterprise traffic and revenue. AWS CloudFront, Google Cloud CDN, and Azure Front Door bundle CDN into broader cloud commitments at marginal cost; these are the structural reason Delivery revenue is declining. Fastly serves a smaller, developer-focused slice; Edgio (formerly Limelight) effectively exited in 2024 with Akamai buying some of its customer contracts. Akamai wins on raw scale and reliability for traffic-bursting media events; loses on price-sensitive smaller customers and on bundled deals where the customer is already locked into a hyperscaler.

In Security, the field is more fragmented. Cloudflare again competes across WAF, DDoS, bot, and API; F5 (BIG-IP and the NGINX-derived suite) is strong in enterprise on-premise; Imperva is a long-standing WAF rival; Zscaler is the leader in Zero Trust network access. In microsegmentation specifically, Guardicore (now Akamai) competes with Illumio. In API Security, Salt Security and Traceable are pure-play competitors. Akamai's edge is portfolio breadth and the ability to do in-line inspection at the network edge before traffic ever reaches a customer origin - a fundamentally different architecture from "agent inside the cloud" competitors.

In Compute / AI Infrastructure, Akamai's competitors are the hyperscalers themselves: AWS, Azure, Google Cloud, and (in the GPU specialist tier) CoreWeave, Lambda Labs, and Crusoe Energy. Cloudflare also has a developer-platform play (Workers AI, R2) that overlaps. Akamai cannot match hyperscaler service breadth - it has no equivalent of Bedrock, no managed databases at the AWS scale, no general-purpose ML platform. What it does have is physical distribution. The Anthropic deal validates this: a frontier AI lab with hyperscaler relationships (Google, AWS, plus xAI's Colossus) still wrote a $1.8B cheque to Akamai specifically for distributed inference capacity at the edge.

Barriers to entry are very high in the abstract: nobody is going to spend a decade negotiating carrier interconnection rights and building a 4,400-PoP global network from scratch. But Akamai's competitors don't need to - they're attacking from adjacencies (hyperscale cloud, Cloudflare's pricing model, GPU specialists' focus). The risk is not entry; it is substitution and bundling.

6. Industry

Akamai operates at the intersection of three industries, each with different dynamics.

Content delivery is a mature market estimated at roughly $25-30 billion globally in 2025, growing high single digits driven by video streaming and software downloads. Demand is structurally tied to internet traffic growth, which the major carriers and ITU peg at 25-30% per year compound. But the market is consolidating and pricing is deflating: bandwidth gets cheaper at perhaps 15-20% per year, and hyperscalers' bundling makes pure-play CDN economics ugly. That is why Akamai's Delivery line is shrinking in absolute terms even as bytes delivered keep rising.

Cybersecurity is a much larger and faster-growing pool - estimated at over $200 billion globally with double-digit growth through 2030 - and Akamai targets the application-security and network-security sub-segments worth maybe $50-60 billion combined. Demand drivers are regulatory (PCI, GDPR, DORA, NIS2, state-level US data laws), insurance (cyber insurance increasingly requires specific controls), and the threat environment itself, which deteriorates with every leap in AI capability. Bot attacks now coordinate "millions of malicious requests per second from millions of widely distributed IPs" (Leighton, Q1 2026), which is precisely the kind of attack a distributed in-line defender like Akamai can absorb better than centralised on-prem hardware.

AI inference infrastructure is the newest and most opaque category. Training accounts for the headlines and most of the disclosed GPU capex through 2025, but inference is forecast to overtake training spend by 2027 because every successful AI product eventually generates more queries than training tokens. The industry is structurally biased toward locations close to users because inference latency is felt by the end user. Akamai's central thesis: a meaningful share of inference workload will not run inside hyperscaler regional data centres because the round trip is too slow. The total addressable market estimates here are speculative - hyperscaler capex alone is running at $400B+ annually in 2026 - but even capturing fractions of a percent at the edge would be transformative for a company with Akamai's revenue base.

Regulation cuts both ways. The same regulations driving cyber demand also make Akamai's life harder: data residency rules (GDPR, China's CSL, India's DPDP) require customers to keep data within national borders, which Akamai handles well thanks to its distributed footprint but which add operational complexity. Sovereignty requirements in Europe particularly favour vendors that can demonstrably localise.

Cyclicality is moderate. Enterprise IT budgets dip during recessions but security spending is among the most defended line items. CDN traffic grows secularly. AI inference is in early innings - whether it is cyclical at all will not be clear for years.

7. Growth Triggers

Every item below is sourced from one of the four most recent concalls (Q2 2025, Q3 2025, Q4 2025, Q1 2026).

• Anthropic $1.8B / 7-year committed-capacity contract for Inference Cloud, largest deal in company history (Q1 2026 concall, May 7 2026). Revenue recognition begins Q4 2026 ($20-25M expected), then ramps ratably. Associated capex of ~$700M in 2026 and ~$125M in 2027.

  "This is a powerful validation of the Akamai value proposition in the age of AI and a clear indicator of the scale at which we can operate." - Ed McGowan, Q1 2026 concall

• $200M / 4-year AI Inference Cloud contract with a major US tech company (Q4 2025 concall, Feb 19 2026; reiterated Q1 2026). Revenue begins Q4 2026.

• Three major US hyperscalers now buying CIS from Akamai for compute workloads (Q3 2025 and Q4 2025 concalls). Management treats this as validation that the distributed-edge architecture wins on technical merits even when the buyer has its own hyperscale infrastructure.

• CIS growth expected to accelerate to 45-50% YoY in 2026 (Q4 2025 concall, Feb 19 2026), driven by Inference Cloud scaling on NVIDIA Blackwell GPUs.

• Inference Cloud sold out in beta across initial 20 cities (Q4 2025 concall). Management ordering additional GPU capacity; further orders contemplated in H2 2026 not yet baked into guidance (Q1 2026 concall).

• Double-digit total revenue growth expected starting 2027 (Q1 2026 concall). Implicit inflection driven by Anthropic and other large CIS deals ramping.

• API Security to scale from <10% existing-customer penetration toward majority (Q4 2025 concall). Exited 2025 at >$100M run rate, growing >100% YoY.

  "API Security following the closure of the NoName acquisition in June is experiencing its first quarter of full organic year-on-year growth." - Tom Leighton, Q3 2025 concall

• Guardicore Segmentation pipeline weighted to new (non-Akamai) customers (Q4 2025 concall). Implies the broader Akamai sales engine has not yet been brought to bear; cross-sell upside remains.

• Firewall for AI and AI Gateway product launches (Q2 2025 and Q4 2025 concalls). New product surface attached to the broader enterprise AI rollout.

• Pricing discipline in Delivery moderating the decline (Q4 2025 concall). Management guides Delivery decline to slow to mid-single-digits in 2026 as Edgio lap effect washes through; segment then stabilises.

• $47M / 5-year contract with major hardware company combining API Security and CIS (Q4 2025 concall). Cross-segment bundling example.

• $40M / 4-year Guardicore Segmentation deal with North American financial institution (Q4 2025 concall).

• $45M / 3-year security migration from hyperscaler at large financial institution (Q4 2025 concall). Notable as a competitive win against hyperscaler-native security.

8. Key Risks

Capex burden if AI inference adoption disappoints. Akamai is spending $800-825M over the next 12 months specifically to fulfil committed-capacity contracts, particularly Anthropic. The Anthropic deal is take-or-pay from the customer's side, so direct counterparty risk is contained. But if generalised inference demand outside Anthropic does not materialise, Akamai is left with a stock of NVIDIA Blackwell GPUs depreciating on accelerated schedules. Capex jumps to 40-42% of revenue in 2026 from the historical low-20s. This is a one-time bet, but a large one.

Delivery decline reaccelerating. The segment is now -7% YoY in Q1 2026 and structurally exposed to hyperscaler bundling. Management's guidance that decline moderates to mid-single-digits depends on the Edgio lap effect rolling off and pricing discipline holding. If a major media customer self-builds or migrates to a hyperscaler, the segment could surprise to the downside again. Each percentage point of Delivery decline is hard for CIS growth to backfill.

Customer concentration in Compute. A handful of mega-deals (Anthropic, the $200M US tech contract, several hyperscalers) will increasingly drive CIS revenue. If Anthropic struggles commercially (the AI model lab business is winner-take-most and capital-intensive), Akamai becomes exposed to a single counterparty that did not exist as a meaningful customer two years ago. Anthropic has multiple infrastructure partners (Google Cloud, AWS, SpaceX), which is both reassuring (they have alternatives, suggesting Akamai is providing something specific) and concerning (they could reallocate spend).

Hyperscaler competitive escalation. AWS, Google, and Microsoft can in principle build out their own distributed edge networks - they have capital, customer relationships, and existing CDN products. Cloudflare has already built much of this and has a much bigger developer footprint. Akamai's defence is twenty years of carrier relationships and physical PoPs, but if a hyperscaler decides edge inference is strategically critical, the gap can be closed faster than Akamai can build deeper moats.

Security competitive pressure from Cloudflare. Cloudflare has grown faster than Akamai across most security categories for several years. Akamai's security growth is decelerating to high-single-digits in 2026 guidance. If Cloudflare continues to win the modernisation cycle, Akamai's profit engine slows just as Compute capex peaks.

Execution risk on the AI Grid. Distributed AI inference at this scale - across 4,400 PoPs with intelligent orchestration - has not been done before. Operational glitches that hyperscalers can absorb in centralised data centres are much harder to debug across thousands of small sites. Outages would be public, damaging, and could push Anthropic and similar customers back to centralised hyperscalers.

Acquisition integration debt. Akamai has spent ~$2 billion on four major acquisitions in the past decade. Each requires sales force integration, product rationalisation, and platform unification. Noname Security is still integrating; Guardicore took years; Linode is still operationally distinct. Goodwill and intangibles are large.

"I do not know of a comparable time where there is this much concern about what is going to happen with security." - Leighton, Q1 2026.

Management frames this as opportunity. It is also a warning: the same AI capabilities making attackers more dangerous are reshaping the security tooling stack rapidly, and incumbent vendors that miss a generational transition (Symantec, McAfee, Cisco's security business) become structurally impaired within a few years.

9. Walk the Talk

Concalls reviewed: Q2 2025 (Aug 7, 2025), Q3 2025 (Nov 6, 2025), Q4 2025 (Feb 19, 2026), Q1 2026 (May 7, 2026).

The clearest test of Akamai management credibility over these four quarters is the Cloud Infrastructure Services growth story. In the Q2 2025 concall Tom Leighton told investors "Our Cloud Infrastructure Services grew 30% year-over-year. And we expect that rate to accelerate through the remainder of the year." This was a specific, datable commitment. In Q3 2025 CIS growth came in at 39%; in Q4 2025 at 45%; in Q1 2026 at 40%. The trajectory accelerated quarter-on-quarter through the back half of 2025 exactly as promised, then held in the 40s through Q1 2026 - in line with the higher 40-45% target McGowan set in mid-2025 and the 45-50% range Leighton extended at the Q4 2025 call.

The second clear delivery was guidance: in the Q2 2025 concall Akamai raised full-year revenue guidance to $4.135-$4.205B. Q3 2025 ($1.055B) and Q4 2025 ($1.095B) actuals tracked at the high end of that band. Q4 2025 EPS came in at $1.84, ahead of guidance. Management has been consistently conservative-to-on-track on the revenue and earnings numbers across these four quarters.

The third tracking item is delivery decline. Q2 2025 reported -3% YoY in Delivery and management said the pricing environment was stabilising. Q3 2025 was -4%, Q4 2025 stepped up slightly, and Q1 2026 came in at -7% YoY. Here management's tone has been more cautious than precise: they have not promised stabilisation, but the -7% in Q1 2026 was at the worse end of "mid-single-digit decline" framings. McGowan in Q1 2026 noted "we expect this effect in the rate of decline to moderate throughout the remainder of the year" - a softer commitment than earlier in the year.

The most striking item is what management did NOT pre-announce. The Anthropic $1.8B deal landed in the Q1 2026 concall (May 7 2026) as a surprise. In the Q4 2025 concall ten weeks earlier, management announced a separate $200M / 4-year AI infrastructure deal with a "major US tech company" but gave no hint that a vastly larger contract was already in negotiation. This is appropriate confidentiality - one does not pre-announce un-signed commercial contracts - but it does mean management's pipeline disclosures are deliberately conservative. CIS pipeline has been described as "significantly exceeding existing and projected inventory" (Q1 2026), suggesting more such surprises are possible.

The capex story is on plan but escalating. Q2 2025 capex was modest; Q3 2025 capex was ~16% of revenue; Q4 2025 introduced the $250M incremental investment for Inference Cloud; Q1 2026 escalated dramatically to support the Anthropic deal with full-year guidance of 40-42% of revenue. Each step was disclosed when it was committed; nothing has been quietly walked back.

Net assessment: Tom Leighton (CEO since 2013, co-founder) and Ed McGowan (CFO since 2013) are among the longest-tenured exec pairs in large-cap tech. The track record across these four quarters is consistent: they guide conservatively, they deliver, they pre-announce capex when they commit it, and they let large commercial wins land as surprises rather than tease them. This is management that does what they say. The one open question is whether the magnitude of the AI bet - $700M of capex for one contract - is being adequately stress-tested for downside scenarios.

10. Shareholder Friendliness Index

Akamai has never paid a cash dividend on its common stock and has explicitly stated it does not intend to. The Investor FAQ on the company's IR site confirms this; there is no dividend history on stockanalysis.com or any dividend tracker. There is no special dividend, no IPO-era policy that was later suspended, no signalling at all on the dividend side - capital return is one-channel-only.

That channel is share repurchases, and Akamai has been an aggressive net buyer of its own stock. In FY2024 Akamai spent $557M to buy back 5.6M shares at an average $99.14. In H1 2025 alone the company spent $800M ($500M in Q1 2025 at avg $81.19, $300M in Q2 2025 at avg $77.51) buying back 10.1M shares - notably during a period when the stock was trading in the high-$70s to low-$80s, before the Anthropic deal lifted the price. The Q1 2026 disclosure shows another ~2M shares retired and $975M of authorisation remaining under a $2B program effective through June 2027. Across the trailing 3 years the buyback has been large enough to offset stock-based compensation dilution and reduce shares outstanding modestly; the share count is moving in the right direction.

Verdict: Returns Capital. Buyback-only, no dividends, but the buyback program has been consistent, large, and well-timed (heavy repurchases in 2025 below $80 ahead of the May 2026 Anthropic-driven rerating). For a company funding a multi-hundred-million-dollar capex build-out, the buyback pace is notable - management has been comfortable returning capital and investing in growth simultaneously.

11. Insider Activities

Akamai files Form 4s on the SEC EDGAR system; below are the material transactions from the trailing 12 months (roughly May 2025 through May 2026). I am including one transaction from outside that window - the CEO's February 2025 open-market buy - because it is the dominant signal and would distort the picture if omitted.

Recent material transactions:

Buys. The single most important data point in this entire section is Tom Leighton's February 27, 2025 open-market purchase of 37,670 shares for approximately $3.0 million at an average price around $79.85. For a CEO who is already a major shareholder (he holds well over 2 million shares through various entities), a personal open-market buy of $3M is a very bullish signal. The buy was made during a period of broad pressure on the name following weak Delivery numbers and before the CIS acceleration story became consensus. The price he paid was within 10% of the 52-week low. Roughly 14 months later, the stock traded well above $100 following the Anthropic announcement. Leighton has not bought again in the trailing 12 months - the February 2025 buy was a one-shot conviction signal, not a pattern - but its timing was prescient.

No other officer has made open-market purchases in the window. There is no cluster-buying signal.

Sells. The sells are concentrated in a single window: mid-March 2026, in the immediate wake of the strong Q4 2025 report (announced Feb 19, 2026) which lifted the stock from the high-$80s into the $100-$107 range. Five named executive officers (CFO, COO, EVP CHRO, EVP Sales, EVP General Counsel) all sold in roughly the same fortnight. Every one of those sales was conducted under Rule 10b5-1 trading plans adopted in August or September 2025, well before the Q4 2025 results were known. The plans were structured to execute on price triggers or date triggers; the price spike post-Q4 results activated the price-triggered tranches. The volumes are modest in context - each executive retained most of their shares, and the dollar values represent diversification rather than retreat. The earlier (March 2026 vintage) sales are documented as pre-planned diversification; there is no evidence of information asymmetry trading. The fact that none of the officers sold ahead of the May 2026 Anthropic announcement, when many would have known of the negotiation, further suggests the plans were genuinely arms-length.

Net assessment. The signal is mixed but skews neutral to mildly positive. The CEO's $3M open-market buy 14 months ago was a high-conviction insider purchase and has been vindicated by subsequent business performance. The 2026 sells were scheduled, modest relative to holdings, and consistent with standard executive diversification rather than fear. No officer is dumping; no officer is loading up beyond Leighton's solitary 2025 buy. Read this as: insiders are comfortable, the CEO has skin in the game, and the recent diversification by the broader exec team is what one would expect from senior officers after a multi-quarter run.

12. Scenarios

Bull case. The Inference Cloud thesis works. Distributed AI inference becomes the dominant deployment model because user-facing AI products demand low latency, and centralised hyperscaler regions cannot match Akamai's network proximity. Anthropic's $1.8B contract is followed by similar deals from OpenAI, xAI, Meta, and second-tier AI labs through 2027-2028. CIS revenue tips into multi-billion-dollar territory and CIS becomes the largest segment by 2028. Three of the four hyperscalers continue to use Akamai for distribution; one quietly becomes a top-five customer. Meanwhile API Security and Guardicore Segmentation cross-sell into the existing 9,000-enterprise base lifts overall Security growth back into the mid-teens. Delivery stabilises at -3 to -5% per year. Akamai exits 2028 as a credible third infrastructure pillar of the AI economy, behind only hyperscalers and NVIDIA, with operating margins held in the high-20s through capex digestion.

Base case. Anthropic ramps as guided. CIS grows 45-50% in 2026 and re-accelerates further in 2027 as new committed-capacity deals layer on top. Security continues high-single-digit growth, with API Security and Guardicore offsetting maturity in WAF and DDoS. Delivery declines mid-single-digit through 2026 and stabilises by 2027 as competitive intensity moderates. Total revenue growth re-accelerates to double-digits in 2027 as Leighton guided. Operating margins compress slightly in 2026-2027 as the Anthropic capex flows through depreciation, then recover from 2028 as the contract revenue scales against a now-built infrastructure base. The company continues to repurchase shares at scale, pivoting some of the cash flow toward capex but maintaining a buyback floor. Akamai becomes a clearer growth story rather than a value-trap CDN.

Bear case. Distributed inference fails to materialise as a workload pattern. Anthropic ramps to its committed minimums but does not generate growth beyond contract size; other AI labs decline to commit. The $700M of 2026 capex sits with sub-target utilisation, depreciating against revenue that does not arrive. Delivery decline accelerates again as a major media customer self-builds. Cloudflare takes another 200-300 bps of Security share by undercutting on price across WAF and API. Operating margins compress into the low-20s through 2027 as capex flows through D&A without matching revenue. The buyback pace slows as cash flow tightens. The market reframes Akamai as a structurally challenged CDN with an expensive AI side-bet that did not work. Management's strategic credibility, built over a decade of consistent execution, takes a major hit.

13. Further Reading

• Akamai Buys Linode, Akamai's Strategy, Cloudflare and Disruption - Stratechery, 2022-02-16 [paid]

Sources:
- [Akamai Q1 2026 Earnings Transcript - The Motley Fool](https://www.fool.com/earnings/call-transcripts/2026/05/07/akamai-akam-q1-2026-earnings-transcript/)
- [Akamai Q1 2026 Earnings Call - Investing.com](https://www.investing.com/news/transcripts/earnings-call-transcript-akamai-q1-2026-sees-strong-growth-stock-surges-21-93CH-4675571)
- [Akamai Q4 2025 Earnings Call Transcript - Insider Monkey](https://www.insidermonkey.com/blog/akamai-technologies-inc-nasdaqakam-q4-2025-earnings-call-transcript-1699765/)
- [Akamai Q3 2025 Earnings Transcript - Seeking Alpha](https://seekingalpha.com/article/4839797-akamai-technologies-inc-akam-q3-2025-earnings-call-transcript)
- [Akamai Q2 2025 Earnings Transcript - The Motley Fool](https://www.fool.com/earnings/call-transcripts/2025/08/07/akamai-akam-q2-2025-earnings-call-transcript/)
- [Akamai Wikipedia](https://en.wikipedia.org/wiki/Akamai_Technologies)
- [Akamai 10-K FY2024 - SEC](https://www.sec.gov/Archives/edgar/data/0001086222/000108622225000028/akam-20241231.htm)
- [Akamai CEO buys $3M in stock - Investing.com](https://www.investing.com/news/stock-market-news/akamai-technologies-stock-gains-on-ceos-3m-purchase-93CH-3898147)
- [AKAM Insider Trading - SecForm4](https://www.secform4.com/insider-trading/1086222.htm)
- [Anthropic-Akamai $1.8B Deal - Bloomberg](https://www.bloomberg.com/news/articles/2026-05-08/anthropic-inks-1-8-billion-computing-deal-with-akamai)
- [Akamai-Anthropic Deal Analysis - The Next Web](https://thenextweb.com/news/akamai-anthropic-cloud-deal-ai-infrastructure)
- [Akamai Inference Cloud with NVIDIA - Akamai Press Release](https://www.akamai.com/newsroom/press-release/akamai-inference-cloud-transforms-ai-from-core-to-edge-with-nvidia)
- [CDN Market Share 2025 - BlazingCDN](https://blog.blazingcdn.com/en-us/biggest-cdn-providers-market-share-pop-count-2025)
- [Akamai-Linode Acquisition - TechCrunch](https://techcrunch.com/2022/02/15/akamai-acquires-linode-for-900m/)

Report saved as deliverable. Four concalls confirmed and used (Q2 2025, Q3 2025, Q4 2025, Q1 2026 - the latter within 17 days of today). Key surprise: the $1.8B Anthropic deal landing in Q1 2026 reframes the entire investment story, and CEO Leighton's Feb 2025 $3M open-market buy at ~$80 (vs current ~$100+) is the single most important insider signal.