AWS plans to outspend Microsoft and Google on AI infrastructure

Amazon Web Services (AWS) intends to invest $100 billion in ramping up infrastructure for AI cloud services this year, surpassing the spending plans of rivals Microsoft and Google.

Amazon’s capital expenditure in the fourth quarter totaled $26.3 billion, the vast majority of it on AI for AWS, CEO Andy Jassy said during the company’s earnings call on Thursday. “That is reasonably representative of what you could expect in annualized capex rate in 2025,” he said, suggesting the company plans to invest over $100 billion this year.

That’s also a sign of how hungry AWS thinks the market is for access to data centers, processors, networking gear, and other hardware for AI and generative AI workloads, according to the CEO.

“We don’t procure it unless we see significant signals of demand. And so, when AWS is expanding its capex, particularly in what we think is one of these once-in-a-lifetime type of business opportunities like AI represents, I think it’s actually quite a good sign, medium to long term, for the AWS business,” Jassy said, adding that he expects every enterprise application will be reinvented with AI inside with inferencing being a core building block, just like compute, storage, and databases.

Demand for AI is so huge that market research firm IDC expects that global spending on AI-supporting technologies will surpass $749 billion by 2028 — and nearly 67% of the projected $227 billion AI spending in 2025 will come from enterprises embedding AI capabilities into their core business operations.

AI data centers don’t come cheap

While AWS’ $100 billion spending target might sound like a lot for just one year of spending on AI infrastructure improvements, AI-enabled data centers don’t come cheap.

AI data centers must support much higher power densities than traditional data centers: Nvidia’s GB200 NVL72 systems are estimated to consume up to 120kW per rack, for instance, with classic computing infrastructure consuming perhaps one-tenth of that.

On top of that, the AI-enabled data center will need liquid cooling, advanced networking infrastructure, and advanced infrastructure management software.

And AWS isn’t the only cloud service provider that is ramping up its investments into AI-enabled data centers. Rival cloud service providers are all investing in either upgrading or opening new data centers to capture a larger chunk of business from developers and users of large language models (LLMs).

Earlier this year, Microsoft President Brad Smith said the company is on track to invest nearly $80 billion to build out AI-enabled data centers this fiscal year.

The majority of the $75 billion in capital expenditure Google will make this year will go toward technical infrastructure including servers and data centers, Google CFO Anat Ashkenazi said in the company’s earnings call this week.

Based on the investment numbers provided by the three major cloud service providers over the last week, AWS is leading the pack by around $20 to $25 billion — and is ahead, too, of analysts’ forecasts for cloud infrastructure spending.

A report published by Bloomberg Intelligence in October 2024 had estimated that demand for generative AI would push Microsoft, AWS, Google, Oracle, Meta, and Apple between them devote $200 billion to capex in 2025, up from $110 billion in 2023.

Separately, several large technology firms including OpenAI, SoftBank, Oracle, Nvidia, and MGX have partnered to set up a new company in the US, named Project Stargate, to ramp up AI infrastructure in the country.

Project Stargate, which will be financially led by SoftBank and operationally by OpenAI, will see an immediate investment of $100 billion although an additional $400 billion is expected in the next four years.

Supply chain and energy constraints may dampen plans

While cloud service providers are boosting their capital expenditure to scale up infrastructure to support even more AI workloads, supply chain and energy constraints may dampen their plans.

Amazon’s Jassy said the company’s cloud business could be growing faster, but faced constraints on capacity, including a slow-down in deliveries of chips from its suppliers, difficulties in getting working chips, and power supply constraints”

The supply chain constraints also affect the Tranium 2 chips it developed in-house, and server motherboards, both of which the company expects to start receiving in larger numbers over the next couple of quarters. The situation is similar with Microsoft and Google, which are seeing more demand for AI workloads than they can supply, executives said in their most recent earnings calls.