The Deal, By the Numbers

Snowflake and Amazon Web Services announced Wednesday a $6 billion, five-year spending commitment — Snowflake's largest infrastructure deal ever.

To understand the scale: according to TechCrunch, Snowflake has generated roughly $7 billion total in AWS Marketplace revenue since the company was founded in 2012. This single new contract is worth nearly that entire 14-year history.

Snowflake's AWS spending has been accelerating fast. According to TechCrunch, it doubled in 2025 to $2 billion for that calendar year alone. The new agreement implies an average annual spend of $1.2 billion, according to CNBC.

Snowflake's Earnings Were the Real Story

The AWS deal didn't land in a vacuum. Snowflake dropped its fiscal Q1 results the same day, and they were strong.

According to CNBC, Snowflake reported $0.39 in adjusted EPS on $1.39 billion in revenue — up 33% year over year. Analysts polled by LSEG had expected $0.32 per share and $1.32 billion in revenue. Both beats.

Guidance also came in above expectations. Snowflake projected $1.415 billion to $1.420 billion in Q2 product revenue with a 12.5% adjusted operating margin. Analysts surveyed by StreetAccount had penciled in $1.37 billion and an 11.9% margin.

The market responded accordingly. Snowflake shares jumped as much as 36% in extended trading, according to CNBC.

What's Driving This: AI and the CPU Boom

Every tech story right now has "AI" in the headline, but this one has a specific angle.

The $6 billion deal includes a major commitment to AWS's Graviton chips — Amazon's homegrown ARM-based processors. According to TechCrunch, as AI moves from model training toward everyday usage and automation via agents, CPU demand is skyrocketing. GPUs dominate training and reasoning workloads. CPUs handle the rest — including the grunt work of running AI agents at scale.

Everyone is focused on Nvidia's GPU dominance, but the battle for CPU workloads is intensifying. Amazon CEO Andy Jassy claimed last month that Amazon's homegrown chips offer "better price-performance than Nvidia's offerings," according to TechCrunch — though AWS still runs Nvidia hardware too.

Amazon Is Building a Graviton Client List

The Snowflake deal isn't a one-off. AWS is assembling a serious roster of Graviton customers.

Last month, according to TechCrunch, AWS signed a deal to supply millions of Graviton chips to Meta for its AI compute needs — a notable win given Meta had signed a $10 billion deal with Google Cloud just months earlier.

Add in Anthropic's commitment to spend over $100 billion on AWS over a decade and Amazon's separate deal with OpenAI, and AWS is making a case that it's the default home for serious AI workloads. According to CNBC, both the Anthropic and OpenAI deals include equity investments from Amazon — the Snowflake deal does NOT.

What Mainstream Coverage Is Missing

Both TechCrunch and CNBC covered this accurately, but neither pressed on a bigger strategic question: what does a $6 billion commitment tell us about vendor lock-in?

Snowflake positions itself as a multi-cloud platform — available on AWS, Microsoft Azure, and Google Cloud. But a $6 billion, five-year commitment to a single cloud provider is not a neutral multi-cloud posture. Snowflake is effectively doubling down on AWS as its primary infrastructure backbone. That's a business reality its enterprise customers should factor in.

Also underreported: the historical context of Snowflake's AWS agreements. According to CNBC, at Snowflake's 2020 IPO, it had a $1.2 billion, five-year deal with Amazon. By 2023, that climbed to $2.5 billion. Now it's $6 billion. That's a 5x increase in committed spend in roughly six years. Either Snowflake's business is exploding — which the earnings confirm — or AWS has serious leverage over a company that built its entire infrastructure on Amazon's cloud. Probably both.

The Nvidia Subtext

Cloud giants — Amazon, Google, and now Microsoft with its Maia AI chip — are pouring billions into custom silicon specifically to reduce dependence on Nvidia. According to TechCrunch, Google has been building its own AI chips for years, and Microsoft just launched Maia.

Nvidia's dominance in GPU training workloads is real and isn't disappearing tomorrow. But the hyperscalers are methodically carving out the CPU-heavy, agent-driven AI workloads where homegrown chips can compete on cost. Every Graviton deal chips away at the assumption that Nvidia owns all AI compute.

Looking Ahead

Snowflake had a strong quarter and made a large infrastructure bet. Six billion dollars over five years is serious money. The Graviton focus signals where enterprise AI spending is headed — away from pure GPU spend and toward the operational compute that actually runs AI day-to-day.

Regular people won't see this directly. But if your company runs on Snowflake — and a lot of them do — your data infrastructure just got a $6 billion vote of confidence in Amazon's cloud. Whether that's smart strategic commitment or expensive lock-in is a question every CTO should be asking.