Akamai

Akamai announced a seven-year $11.6 billion cloud computing deal with Anthropic, with an option to grow to $20.6 billion. As part of the deal, Akamai issued Anthropic warrants for about 7.7 million shares, 40% of which are vested, with the remaining 60% vesting if Anthropic adds workloads.

Why it matters: Akamai continues to diversify beyond its core content delivery and security offerings into cloud computing and services. This latest deal represents, at minimum, a quadrupling of the $2.8 billion in commitments the firm has signed. Increased artificial intelligence usage is driving demand for all three of Akamai's segments, but none greater than cloud compute. As this Anthropic deal ramps across 2027, 2028, and 2029, we expect average annual compute segment growth of 93%, lifting firmwide growth to 22%. When Akamai announced the first Anthropic deal in May, we expected there would be additional deals with frontier model labs, but we worried the firm would offer favorable and even margin-dilutive terms to secure the largest customers. However, this latest deal looks at worst margin-neutral.

The bottom line: After incorporating the revenue, capital, and margin implications of this new deal, we raise our fair value estimate for no-moat Akamai to $150 from $135. While the firm's distributed points of presence are attractive to customers, we believe Anthropic is securing whatever compute it can. We expect this deal to earn roughly 30%-35% gross margins. While lower than the security margin, we expect fewer incremental operating expenses. With future deals done with a wider range of customers with less bargaining power, we expect operating margins to continue to improve. To support this deal, Akamai expects to deploy $5.5 billion in CPU capital expenses. The annual recurring revenue/capex ratio in this agreement was 1/3, lower than the more common 1/2, but the size of this deal likely required Akamai to offer a discounted rate, though still value-accretive.

We are accounting only for $11.6 billion of a possible $20.6 billion in revenue for Akamai. CFO Ed McGowan said the $9 billion option is for as-yet-unknown workloads and would require its own capex. In our view, this makes the $9 billion option too uncertain to include in our model, rather than treat it as guaranteed incremental revenue.

Our $150 per share fair value estimate implies a 24 P/E ratio based on our 2026 EPS estimate. After signing another deal with Anthropic worth $11.6 billion in September 2026, we assume Akamai will deliver high-teens to low-double-digit annual revenue growth in 2027, 2028, and 2029. This will likely be followed by mid- to high-single-digit growth and modestly expand margins as the firm continues to offer competitive security products and uses its distributed network to offer cloud and computing services. Weighing on our valuation, we assume heightened capital intensity as the firm invests in scaling its computing infrastructure, which could reduce near-term free cash flow conversion.

In content delivery, we expect traffic increases from trends such as the Internet of Things, autonomous vehicles, and edge computing to begin offsetting long-standing declines in transmission pricing, resulting in roughly flat revenue over our forecast. Qualitatively, we still believe content delivery plays a key role in enhancing Akamai’s security products and offers value beyond revenue.

Security will remain the backbone of Akamai’s offering and revenue generation, with higher growth from API security and Guardicore segmentation solutions, building on more mature denial-of-service protection and zero-trust application services. We forecast average security revenue growth of 10% annually through 2035, slightly below that of the major security players, but improved cross-selling across the compute and security segments could narrow the gap. While successful cross-selling represents upside to our fair value estimate, we exclude it from our base case because compute buyers often opt for single-purpose infrastructure rather than integrated security bundles.

Compute and cloud offerings should continue to grow as demand for AI inference leverages Akamai’s distributed networking capabilities. We still think the GPU-as-a-service market is highly competitive and commoditized, but we believe the appeal of Akamai’s network should make the compute and cloud segments margin-accretive. We expect compute revenue to grow at an average rate of 36% annually, highlighted by 93% annual growth across 2027-29, with the potential for further growth if Akamai enters into additional deals with large language model providers.

In total, we believe consolidated revenue will grow 12% on average, with adjusted operating margin expanding 10% over our 10-year forecast, from 17% to 27%. This margin expansion will be driven by a shift in the revenue mix toward higher-margin security and by compute workloads becoming margin-positive. Free cash flow conversion will initially lag operating income due to elevated capital expenditures, but we expect the gap to narrow as compute investments mature.

We expect Akamai to raise its capital spending to support a buildout of its compute segment, averaging $2.56+ billion per year over the next decade, or 2.7 times 2025 levels. We believe such spending levels will likely serve as a baseline, with the potential for significantly higher capital expenditures in some years if Akamai pursues large compute deals with other frontier model providers.

While we believe that Akamai’s security business exhibits network effects and switching costs, in part derived from the content delivery segment, we also believe management’s focus on turning the cloud and compute business into an asset-heavy, “neocloud” distributed GPU platform is inherently risky and offers little in the way of competitive advantages. While it helps to break the business into three segments, as Akamai does, these categories are largely interconnected and represent alternative ways to monetize existing infrastructure. With such heavy investment in a no-moat, highly commoditized compute business, we believe the security business is not sufficient to offset the risk of value destruction. Thus, we assign Akamai a no-moat Morningstar Economic Moat Rating.

We see Akamai’s content delivery network working in concert with its security solutions offerings, producing a network effect that has kept CDN traffic volume rising amid pricing pressure. A CDN is a distributed network of servers that acts as a highway, allowing customers to reduce latency by keeping content closer to consumers. This infrastructure has inherent network effects. With 20%-30% of global internet traffic passing through Akamai’s network, the firm can use data generated by the content it carries and the websites it hosts to improve web performance and to offer security products. This data often includes multiple petabytes of telemetry per day, fueling the firm's machine learning models for web application and API protection, bot mitigation, and distributed denial-of-service, or DDoS, detection, creating an advantage that is difficult for smaller peers to replicate.

These security services often attract customers that transmit increasing amounts of data through the Akamai infrastructure. As more data flows through Akamai’s network, the security offerings improve. The security improvements then attract more traffic, which, in turn, improves security offerings. Although the CDN provides users that fuel the network effect, the benefit accrues to the cloud-based security business. Security now accounts for 53% of revenue, up from 12% a decade ago.

Despite fueling this network effect, the CDN market has suffered from commoditization and persistent price deflation. While increasing traffic volume has helped slow the decline in CDN revenue, intense competition from newer, developer-friendly networks like Cloudflare and Fastly, combined with major media companies like Netflix building their own in-house networks, has steadily eroded Akamai's pricing power. Additionally, Amazon.com Web Services, Microsoft Azure, and Google Cloud natively bundle CDN, security, and compute functions into their broader cloud ecosystems. This bundling power makes it easier for enterprises to bypass stand-alone providers like Akamai. A network effect will likely persist in some form for Akamai, but security offerings will likely need to expand margins to offset the ongoing downward pressure on content delivery.

Aside from benefiting from network effects in the CDN business, the security business also imposes switching costs. Security solutions are among the most mission-critical business functions, with customers seeking to reduce risk. Security products like Guardicore and Zero Trust Network Access require deep integration of Akamai's platform into a customer's core security policies and IT architecture, making it incredibly complex and risky to rip out.

We have previously viewed Akamai’s compute and cloud infrastructure segment as no-moat but not a major drag on firmwide profitability. However, in May 2026, Akamai announced a $1.8 billion seven-year deal to provide compute to Anthropic. To support this deal, Akamai will invest $800 million in compute capacity, likely Nvidia GPUs, in 2026. Instead of making steady, incremental compute investments that could serve all Akamai customers, Akamai is investing substantial capital in GPUs that will be used exclusively by one customer. This marks the start of a strategic shift toward offering compute as a service.

In that vein, in September 2026, Akamai announced an additional seven-year deal, worth $11.6 billion with an option for an additional $9 billion. While this agreement will require $5.5 billion in capex, the overall operating margin structure is more accretive than we expected, though not enough to materially improve the firm’s returns on invested capital or challenge Akamai's no-moat.

Compute demand has outstripped supply, allowing current entrants to enjoy elevated pricing and attractive headline unit economics. However, this gap is a function of a supply/response lag, not of a maintainable market structure. As supply catches up to demand, per-unit compute pricing should compress materially, and the providers most exposed will be the marginal, subscale operators without differentiated offerings or locked-in customer bases.

While Akamai’s contract with Anthropic is set, we think it will be difficult to prevent customers from switching compute providers in the future. Compute and inference workloads are largely commoditized at the infrastructure layer, leaving little to differentiate on in the core offering. Customers can—and increasingly do—design workloads to be portable across providers, often multi-homing to preserve negotiating leverage. Unlike the hyperscalers, which embed customers through proprietary data services, managed databases, identity systems, and sprawling ecosystems, Akamai is entering as a pure capacity provider, with only modest opportunities to cross-sell into security. While Akamai’s distributed points of presence will be attractive for customers that need inference at the network’s edge, we think the major customers are still looking for heavy localized compute.

Wireless network enhancements and Internet of Things deployments will increase data traffic, adding demand to house content and services as close to customers as possible.

Data generated by media and cloud customers will continue to make Akamai’s security offerings more robust versus competitors that don’t have access to these types of flows.

Akamai’s existing CDN can be adapted to support edge computing and AI inference, which frontier model providers are increasingly demanding.

Media customers of CDNs will continue to insource the transmission of their content, making a content segment turnaround more difficult.

Hyperscalers, which compete with Akamai, are increasingly offering additional security functions with their cloud services, making Akamai less differentiated.

Akamai’s neocloud shift dedicates a majority of capital to a commoditized business, which will only get more competitive as compute supply increases.