A $200 Billion Bet on Google Cloud Infrastructure
Anthropic is reportedly preparing one of the largest AI compute deals ever: a potential commitment to pay Google USD 200 billion (approx. RM920 billion) over five years for cloud and chip access. While Google and Anthropic have not attached a confirmed price tag to their previously announced TPU capacity expansion coming online from 2027, this reported figure would dramatically scale up what had already been a substantial partnership. The arrangement goes beyond routine cloud consumption, effectively reserving future AI compute deals years in advance. For Google, a customer willing to pre-book demand at this magnitude makes it easier to justify aggressive data centre construction, chip procurement and network upgrades. For Anthropic, it is about solving a looming supply problem before it becomes a product bottleneck, securing the training and inference capacity needed to keep rolling out larger models and handling heavier enterprise workloads.
Locking In Chips and Capacity Amid an AI Chip Shortage
The reported Google agreement highlights how leading labs are responding to the AI chip shortage by locking in capacity long before workloads fully materialise. Anthropic’s purchase window, expected to begin in 2027, shifts the heaviest commitments into the next infrastructure buildout cycle rather than relying on spot-market cloud capacity. This gives both Anthropic and Google time to coordinate land acquisition, power, cooling, networking and grid connections—constraints that often move slower than software demand. For Anthropic’s customers, such forward planning is critical to ensuring low latency and stable uptime as usage of Claude-based services grows. The structure of these multiyear AI compute deals also matters: they can influence queue priority and chip allocation, shaping who gets served first when supply tightens. In effect, large pre-reservations become a strategic hedge against both construction bottlenecks and volatile cloud computing pricing.
Multi-Cloud Strategy: Google, AWS and Now Akamai
Anthropic’s Google Cloud infrastructure commitments do not stand alone. The company has already agreed to spend over USD 100 billion (approx. RM460 billion) on AWS over 10 years, tied to up to 5 GW of computing capacity, reinforcing a deliberate multi-cloud strategy. Google’s role is not only as a supplier but also reportedly as a financier, with plans to invest USD 10 billion (approx. RM46 billion) immediately and more subject to performance. On top of these hyperscale relationships, Anthropic has reportedly signed a separate USD 1.8 billion (approx. RM8.28 billion) seven-year cloud computing deal with Akamai, which the latter described as the largest in its history. Together, these layered deals show Anthropic trading long-duration spending for redundancy and resilience: if chip output, pricing or deployment schedules shift at one provider, workloads can be shifted to another without derailing product timelines.
Implications for Smaller AI Buyers and Market Competition
As frontier AI labs and cloud providers strike ever larger exclusive or semi-exclusive agreements, smaller AI buyers may find themselves squeezed. Infrastructure vendors and data centre operators naturally prioritise contracts that promise the biggest and longest demand streams, giving hyperscale commitments an outsized influence on who gets access to scarce hardware. With Anthropic reserving future capacity across Google, AWS and Akamai, more of the high-performance compute pipeline is effectively pre-sold years ahead. That can lengthen wait times and weaken pricing leverage for emerging AI startups, enterprises and public-sector buyers that lack comparable balance sheets. They may be forced to diversify across more niche providers, accept less favourable cloud computing pricing, or build smaller-scale infrastructure themselves. The broader competitive risk is consolidation: as AI chip shortage pressures persist, long-term AI compute deals between major labs and cloud giants could harden market power and raise the barrier to entry for new players.