Can Meta Really Beat Amazon At Cloud? A $10B Anthropic Bet Says Yes
Meta’s Cloud Push accelerated on July 17 when the company was reported to be recruiting a senior Amazon Web Services compute executive while simultaneously negotiating a deal worth roughly $10 billion to sell cloud computing capacity to Anthropic.
The moves together signal that Meta Platforms is no longer content to build AI infrastructure solely for internal use, and instead wants to compete directly with Amazon, Microsoft, and Alphabet as an external cloud provider.
Meta’s Cloud Push Targets Anthropic as a First Anchor Tenant
The Wall Street Journal reported the incoming executive as Dave Brown, who led AWS’s compute division before Amazon disclosed his departure this week. Meta Platforms (META) would deploy Brown’s expertise to focus on data center buildout and cloud infrastructure as part of its broader Cloud Push strategy.
GeekWire confirmed Brown’s move separately. Amazon (AMZN), Microsoft (MSFT), and Alphabet (GOOGL) are the incumbent hyperscalers Meta now aims to challenge.
Anthropic, the AI safety company backed by Amazon and Google, would serve as a major early customer if the deal closes. A $10 billion compute arrangement would be one of the largest single cloud contracts in the AI era, comparable in scale to the multi-year agreements hyperscalers sign with sovereign wealth funds or governments.
The Anthropic angle carries its own irony.
Amazon has spent billions building Anthropic into its cloud ecosystem through AWS partnerships. If Meta’s Cloud Push wins Anthropic’s compute business, it would be poaching a flagship AI customer from the very company whose executive it is hiring.
Why the AI Compute Market Just Got a New Competitor
Cloud computing, at its core, is the business of renting access to data center hardware.
Servers, networking, and storage are pooled into vast facilities and sold to customers who pay by the hour or by the unit of work completed.
What makes this moment different is that AI has transformed the economics. A single large-language model training run can consume thousands of Nvidia GPUs for months.
Inference, the ongoing process of running a trained model to answer user queries, requires sustained GPU capacity at massive scale. Demand has outpaced supply since 2023, and prices for premium compute remain elevated.
Meta has spent more than $50 billion building out its own GPU fleet, primarily to train its Llama family of open-weight models.
That infrastructure sits largely underutilized between training runs. Selling spare capacity to external AI companies converts a sunk cost into recurring revenue — a core rationale behind the Cloud Push.
From Internal Tool to Revenue Engine
Meta’s infrastructure journey began as a support function for Facebook, Instagram, and WhatsApp.
The company built its own server hardware through the Open Compute Project starting in 2011, which let it design racks and cooling systems optimized for its own workloads rather than buying off-the-shelf from established vendors.
That philosophy lowered costs and gave Meta deep engineering expertise. It also produced data centers that, by design, are harder to monetize externally than AWS or Azure facilities, because they were never built with multi-tenant access controls or the billing and compliance layers enterprise customers expect.
Building a true Cloud Push means adding those layers, which is precisely the kind of operational work a veteran AWS executive would know how to execute.
A Sharp Tech podcast discussion from July 10 touched on the structural difficulties Meta faces as a cloud provider, noting that the company’s culture and product orientation make external cloud services an awkward fit. Brown’s hire suggests Meta views those challenges as solvable given the scale of the AI compute opportunity and the momentum behind its Cloud Push.
What a $10B Deal Would Mean for the Compute Market
A signed Anthropic contract at that scale would validate Meta’s Cloud Push ambitions and almost certainly accelerate its buildout timeline.
Anthropic would gain a third major compute source alongside AWS and Google Cloud, giving it negotiating leverage on pricing and reducing single-provider risk.
For the broader market, a fourth hyperscale-class provider entering the AI compute business would increase supply at the top of the market. That is directionally good for AI startups and research labs that have struggled to secure GPU allocations.
It may also put modest downward pressure on inference pricing over a 12-to-18 month horizon.
The deal remains in negotiation and may not close in its reported form. Meta has not made a public statement confirming either the Anthropic talks or Brown’s hire as of this reporting.
Read Next: Morgan Stanley Opens Spot Crypto To 8.6 Million E*TRADE Clients, Coinbase Just Got A New Rival
