From Cloud Customer to Mega-Buyer: Inside Anthropic’s Compute Commitments
Anthropic is rapidly evolving from a typical cloud customer into one of the AI sector’s most aggressive compute buyers. Reports indicate the company may pay Google USD 200 billion (approx. RM920 billion) over five years for cloud and chip access, an amount that would rank among the largest forward commitments in AI infrastructure. This figure sits on top of already public capacity announcements, including a TPU expansion with Google and Broadcom expected to come online in 2027, underscoring how far Anthropic is planning ahead for Claude’s growth. In parallel, Anthropic has reportedly signed a USD 1.8 billion (approx. RM8.3 billion) compute deal with Akamai Technologies, opening another channel for Claude workloads. Together, these Anthropic compute deals signal a strategic shift: infrastructure is not just an operational cost, but a central weapon in the race to deploy larger models, support more enterprise traffic, and deliver new AI-native products.
Why Anthropic Is Locking In AI Infrastructure Investment Years Ahead
Anthropic’s cloud computing agreements are designed to solve a looming supply problem before it becomes a product problem. Training frontier models, scaling enterprise inference, and accelerating release cycles all depend on reliable access to specialized hardware and data center capacity. The reported Google deal, whose purchase window begins in 2027, gives both companies time to align chip procurement, power, networking, and construction schedules around a known demand profile rather than relying on short-term spot capacity. Physical constraints—land, cooling, grid connections, transmission, and installation timelines—make late reservations risky when launch calendars are fixed. For Anthropic, securing this capacity is essential to sustaining the Claude scaling strategy, where larger and more frequent model upgrades must be backed by stable infrastructure. For Google, a multiyear, high-confidence customer helps justify massive capital investments in data centers and custom accelerators long before every rack is installed.
Akamai’s Edge Advantage: Powering Claude Agents and Inference
The Akamai deal highlights a different side of Anthropic’s infrastructure play: low-latency, globally distributed inference for Claude. Akamai, long known for content delivery and cybersecurity, now pitches a distributed GPU and edge architecture aimed at millisecond-level inference for agentic applications. That aligns closely with Claude’s evolving workload profile. Managed agents, hosted automations, and longer-running coding sessions keep sessions active well beyond simple one-off chat replies, frequently calling tools, external systems, and code execution. These workloads need fast responses, proximity to end users, and the ability to spill over smoothly across multiple locations when demand spikes. By tapping Akamai, Anthropic gains another path to spread inference traffic instead of routing everything through a single, distant cluster. This reduces latency for enterprise deployments and strengthens resilience as Claude usage grows, especially in scenarios where uptime guarantees and responsive agents are more critical than raw training throughput.
Multi-Partner Cloud Strategy: Reducing Risk and Shaping Market Dynamics
Anthropic’s deals with Google and Akamai illustrate a deliberate multi-partner strategy rather than dependence on a single cloud. Earlier arrangements with Google, plus separate capacity paths through other suppliers, show how the company is reserving compute across multiple ecosystems to hedge against shortages and vendor risk. For Anthropic, this diversification supports both training and inference needs: large, planned clusters for frontier model development, and flexible, distributed capacity for real-time Claude traffic and agents. However, such long-term reservations may have broader ripple effects. As major AI labs pre-book future capacity, smaller buyers could face longer waits, weaker pricing leverage, and reduced flexibility when negotiating their own cloud terms. Anthropic’s aggressive forward planning effectively shifts bargaining power toward hyperscale providers and top-tier labs, reinforcing the idea that in the new AI economy, access to compute—not just algorithms—will define who can compete at the frontier.