How the energy hunger of artificial intelligence is changing IT procurement strategy

A brand new white paper revealed final week by the Electrical Energy Analysis Institute (EPRI) quantifies the potential for exponential progress within the electrical energy calls for of synthetic intelligence. The 35-page report, titled Augmenting Intelligence: An Evaluation of Synthetic Intelligence and Information Middle Power Consumption, predicts that whole knowledge heart power consumption by US knowledge facilities alone may greater than double to 166% by 2030.

AI-powered queries use 10 instances extra power than search

In line with EPRI, demand is essentially pushed by generative synthetic intelligence, which might require exponentially extra energy per question than conventional search. Notably, this does not embrace photographs and different wealthy content material: “At 2.9 watt-hours per ChatGPT request, AI queries are estimated to make use of 10 instances extra energy than conventional Google queries, which use about 0.3 watt-hours.” hours every; and new computational capabilities corresponding to picture, audio, and video era are unprecedented.”

EPRI Power Use Per Mannequin Picture Credit score: EPRI Report

The report examines 5 completely different use circumstances: Google Search, ChatGPT, BLOOM, and AI-powered Google Search. Amongst them, ChatGPT was the least energy-intensive of the AI-based queries. Nevertheless, the researchers anticipated the combination of Google’s AI capabilities into Google Search, noting that it may very well be greater than 3 instances increased than ChatGPT: “If Google have been to combine comparable AI into its search queries, the search energy may enhance to six.9–8.9 Wh».

A provide constraint happens

EPRI developed 4 completely different projections of potential US knowledge heart electrical energy use between 2023 and 2030 based mostly on completely different annual progress situations: low (3.7%), average (5%), excessive (10%), and better (15%). Within the increased state of affairs, by 2030 knowledge heart electrical energy consumption may rise to 403.9 TWh per yr, a 166% enhance over 2023. Even the low progress state of affairs predicts a 29% enhance to 196.3 TWh per yr.

EPRI US Information Middle Power Consumption Forecasts 2023-20230. Picture Credit score: EPRI Report

The uneven geographic distribution of this progress creates localized issues. In 2023, fifteen states accounted for 80% of the nationwide knowledge heart load, with Virginia alone accounting for 25%. Projections point out that Virginia’s knowledge heart share of whole electrical energy consumption may attain 46% by 2030 beneath a better progress state of affairs. In different states corresponding to Oregon, Iowa, Nebraska, North Dakota and Nevada, knowledge facilities are additionally projected to account for 20% or extra of whole electrical energy demand.

Several types of knowledge facilities are contributing to this progress. Company knowledge facilities, owned and operated by particular person corporations for their very own use, account for 20-30% of the entire load. Colocation facilities, the place companies hire shared area and infrastructure, and hyperscale facilities constructed by cloud giants corresponding to Amazon, Google, and Microsoft collectively account for 60-70% of the workload. Particularly, hyperscale facilities are on the forefront of power innovation, given their sheer scale, new facilities are being constructed with a capability of 100 to 1,000 megawatts, equal to the load of 80,000 to 800,000 houses.

Flipping the info heart procurement script

Because the demand for AI-powered purposes grows, companies wish to get the newest GPU-powered servers from distributors like Nvidia. Nevertheless, proudly owning these superior machines is barely half the battle. Even if you happen to handle to get the tools in time, the facility necessities for these energy-intensive methods have gotten an more and more urgent challenge. Which means that the race to undertake AI shouldn’t be solely about buying the suitable {hardware}, knowledge and fashions, but in addition about having the info heart energy. We’re again in 1999 and the dotcom growth.

On this surroundings, companies might want to begin considering extra like their hyperscale opponents. Firms corresponding to Amazon, Google and Microsoft have lengthy understood the significance of securing long-term knowledge heart capability to assist their formidable progress plans. They typically enter into multi-year contracts with energy suppliers, facility operators and contract producers to lock within the sources wanted to scale.

For enterprises, this will likely require a elementary change in the best way they strategy knowledge heart procurement. Traditionally, many corporations have relied on a three-bid-buy mannequin, issuing requests for proposals and deciding on the lowest-cost vendor for every venture. However in a world the place knowledge heart capability is more and more restricted and infrastructure {hardware} is a sizzling commodity, this strategy is probably not viable.

As a substitute, companies may have to determine long-term partnerships with knowledge facilities and {hardware} distributors, committing to a sure degree of capability over a protracted interval in trade for assured provides. This kind of provide chain association is already turning into extra widespread within the business, with some knowledge heart distributors reportedly shifting away from the normal RFP course of altogether.

“Information heart tools distributors, a lot of them, do not even reply to RFPs that usually,” an business govt instructed us on situation of anonymity. “They’re shifting to a mannequin the place they provide a specific amount of capability each month or quarter and the corporate contracts for the provision. Ten years in the past, 100% of our income was three bets and a purchase. Immediately it is 25%.”

For a lot of enterprise IT leaders, this transition to contracting could require a brand new degree of strategic considering and long-term planning. This proactive, forward-looking strategy to knowledge heart planning won’t be straightforward. It will require shut collaboration between IT, amenities and finance groups, in addition to a willingness to make important upfront infrastructure investments that won’t pay for themselves for years. However for companies critical about competing in an AI-driven future, it could be the one approach to make sure they’ve the sources they should keep forward.