Akamai Technologies closed an $11.6 billion multi-year AI infrastructure contract, converting two decades of content delivery network assets into inference compute capacity. The deal represents 47% of Akamai's current market capitalization and marks the largest single customer commitment in the company's 25-year history.
The contract commits Akamai's globally distributed edge servers — originally built to cache Netflix streams and accelerate web pages — to AI model inference workloads. The customer remains undisclosed, but the deal structure suggests a hyperscaler or large model provider seeking geographic distribution that centralized GPU clusters cannot provide. Akamai operates 4,100 edge locations across 135 countries, positioning inference closer to end users than AWS, Azure, or Google Cloud's availability zones. The revenue will recognize over 36 to 48 months, based on capacity delivery milestones tied to GPU deployment and network throughput commitments.
This reconfigures the economics of edge infrastructure. Akamai's legacy CDN business generated $3.6 billion in trailing twelve-month revenue at 18% EBITDA margins. AI inference contracts carry 45-60% gross margins when GPU utilization exceeds 70%, according to CoreWeave and Lambda Labs disclosures. If Akamai executes the buildout without cost overruns, the deal could lift company-wide EBITDA margins by 8-12 percentage points by fiscal 2026. The risk is execution: Akamai must deploy 15,000 to 25,000 H100 or H200 GPUs across distributed data centers, navigate Nvidia allocation constraints, and maintain sub-50ms latency SLAs that centralized inference providers ignore.
The deal also pressure-tests the thesis that AI workloads require centralized, liquid-cooled megaclusters. Distributed inference makes sense for latency-sensitive applications — voice agents, real-time translation, autonomous vehicle decision layers — where round-trip time to a Virginia data center kills product experience. Akamai's edge footprint becomes valuable precisely because it is NOT a single building in Iowa. If the contract performs, expect hyperscalers to acquire smaller edge providers or accelerate their own edge AI product rollouts. Cloudflare, Fastly, and Lumen Technologies hold similar global footprints but lack Akamai's security product suite, which likely sweetened this deal.
Watch Akamai's CapEx guidance in the next two earnings calls. The company will need $2.8 billion to $4.1 billion in hardware spend to fulfill this contract, far exceeding its historical $400-$500 million annual CapEx. If management announces debt raises or equity dilution to fund GPU procurement, the stock reprices lower. If they secure vendor financing from Nvidia or Dell, margins stay intact. Also watch for contract termination clauses: if the customer can exit after 12 months with minimal penalty, this is a pilot, not a platform shift. Akamai reports Q1 fiscal 2025 earnings on April 29; guidance on GPU delivery timelines and margin impact will clarify whether this deal is transformational or a high-risk buildout with single-customer concentration risk.
The undisclosed customer probably cannot get equivalent capacity from the Big Three clouds in the required timeframe. That tells you something about 2025 inference demand.