Akamai lands $11.6 billion Anthropic deal, shares soar

Akamai shares rose as much as 20% after securing a $11.6 billion, seven-year cloud-infrastructure contract with Anthropic to support the AI company's CPU workloads. The commitment adds to more than $2.8 billion in previously announced multi-year cloud-services commitments, although Akamai expects roughly $5.5 billion of related capex and a $1.7 billion increase in 2026 capital expenditures to pre-purchase supply-chain components such as memory. The deal does not change Akamai's 2026 revenue guidance, while Anthropic can expand its commitment by up to an additional $9 billion, linked to vesting of warrants for up to approximately 5% of Akamai's shares.

Analysis

The market should separate headline total contract value from near-term economics: AKAM must fund a material portion of infrastructure before revenue meaningfully enters guidance, creating a 2026 free-cash-flow trough and likely requiring either incremental leverage or reduced buybacks. The key underwriting variable is not the nominal commitment but whether it is take-or-pay, includes inflation/pass-through protection for memory and power, and earns returns above AKAM's cost of capital after network operations and depreciation. Until those terms are disclosed, the post-close move is a rerating of backlog quality rather than a demonstrated earnings upgrade.

The warrant structure makes Anthropic both customer and economically advantaged counterparty. If the initial equity tranche vests, dilution is manageable but real; further dilution is tied to expansion that may be strategically attractive yet lowers AKAM's effective revenue yield. The $111.33 exercise price is also a useful technical reference: sustained trading materially above it makes the dilution increasingly economic, while a decline below it would signal that investors are assigning little value to the future expansion component.

Over the next 1-3 months, focus on financing, 2027 revenue cadence, gross-margin guidance, and customer-concentration disclosures rather than reported bookings. Memory procurement creates a second-order beneficiary in MU, although the order is unlikely to alter industry supply/demand without evidence that AKAM is locking meaningful HBM or server-memory capacity. Over 6-18 months, successful execution could establish AKAM as a differentiated distributed inference/CPU platform; failure would leave it with stranded, rapidly depreciating AI infrastructure and a structurally lower FCF multiple.

Contrarian view: the announcement is not necessarily negative for hyperscalers or GPU-cloud peers because CPU-heavy workloads are complementary to, rather than substitutes for, premium GPU training capacity. The more immediate competitive pressure falls on smaller edge/cloud providers that cannot pre-purchase components or finance dedicated capacity, but AKAM's advantage remains unproven until utilization and pricing are visible.

Key Decisions for Investors

• Do not chase the initial AKAM gap. Place a 1-3 month watch to initiate long exposure only after management discloses minimum-spend/termination provisions and shows that projected cloud gross profit covers incremental depreciation, power, and financing costs; target a 15-20% upside rerating if 2027 revenue conversion is firm, with thesis invalidated by lower FCF guidance or debt-funded capex beyond the stated plan.
• For existing AKAM longs, trim 25-33% into strength and retain a core position rather than exiting outright. Use a sustained break below the $111.33 warrant exercise price, combined with no 2027 revenue-framework disclosure at the next earnings call, as a signal to reduce further because the market would be discounting execution risk and prospective dilution.
• Monitor MU as a supply-chain read-through, not a standalone trade from this announcement. Upgrade to a tactical long only if AKAM quantifies material memory commitments or MU cites broad AI-infrastructure demand tightening; the falsifier is continued DRAM pricing softness or evidence that purchases are conventional server memory rather than constrained high-value products.
• Consider a small AKAM-long / NET-short relative-value position only after AKAM's post-announcement valuation premium is quantified and contract protections are verified. The thesis is that dedicated AI infrastructure can narrow AKAM's growth discount to CDN peers; stop out if AKAM's cloud margin outlook deteriorates or NET demonstrates superior AI-related revenue acceleration.