From Data Location to GPU and Power Control: The Sovereign AI Infrastructure Landscape Is Changing
IT DAILY ·
✦ AI Summary
Gartner said in its report that the core change in sovereign AI is moving from a focus on data storage location to direct control over infrastructure as a whole, including GPU and power.
Gartner analyzed that global geopolitical fragmentation and the spread of generative AI are exposing the limits of the single public cloud and high-performance GPU clusters' power and cooling capacity.
Gartner proposed a practical sovereignty strategy and a hybrid stack configuration, and recommended verifying region-specific Neocloud providers, building local infrastructure, and introducing technical verification frameworks such as Confidential Computing.
Gartner said in its report, "Sovereign AI Infrastructure Strategies in a Multipolar World," that the landscape for sovereign AI infrastructure is undergoing a major shift. Gartner said the core change in sovereign AI is moving from a focus on data storage location to direct control over infrastructure as a whole, including GPU and power. This assessment and the limits it points to are based on Gartner's analysis.
Gartner cited global geopolitical fragmentation as the backdrop to this shift. It also analyzed that the single public cloud model has reached its limits amid this fragmentation. Gartner presented this background and analysis together in the report.
In a survey Gartner conducted among global infrastructure and operations (I&O) leaders, 77% of respondents said they viewed data sovereignty as an essential element based on achieving organizational goals, 75% said the same of operational sovereignty, and 54% said the same of technical sovereignty. Gartner also pointed to the spread of generative AI. As a result, it said power and cooling constraints are emerging in high-performance GPU clusters.
Against this backdrop, Gartner said securing "Compute Sovereignty" is urgent. It cited the changing nature of large-scale GPU infrastructure, or "AI factories," as the reason. Gartner said "AI factories" have taken on a status that goes beyond simple IT facilities and are being treated as national strategic assets.
Countries pursuing their own sovereign AI stacks will need to invest at least 1% of GDP in AI infrastructure through 2029, the report said. Global annual additional power consumption needed to operate AI data centers is also estimated to rise from 74 TWh in 2022 to 500 TWh in 2027.
These infrastructure constraints and complex regulations are presented as factors driving the geopatriation of corporate workloads. Gartner forecast that by 2030, more than 75% of companies in Europe and the Middle East will relocate infrastructure in order to reduce geopolitical risk.
In this process, Neocloud providers that can quickly secure scarce GPUs and offer full-stack solutions are expected to rise rapidly. The AI cloud market is forecast to reach a total of USD 267 billion in 2030, of which Neocloud providers are expected to account for 20%, or about USD 53 billion.
Gartner said the end-to-end AI stack is concentrated in the U.S. and China, and argued that organizations should adopt a "practical sovereignty strategy" instead of "complete technological independence." Gartner said strategic dependence needs to be reduced, and suggested that a hybrid stack configuration could be a realistic alternative.
Gartner also warned that marketing alone from cloud providers makes it difficult to avoid the risk of extraterritorial jurisdiction, and that the risk cannot be avoided by contractual promises alone. It therefore recommended verifying region-specific Neocloud providers and building local infrastructure to diversify the supply chain, and suggested budgeting for the "Sovereignty Premium" and presenting it as essential insurance. It also recommended introducing a technical verification framework based on Confidential Computing, external key management, and TEE.
Source: IT DAILY · Kwon Young-seok
Original: https://www.itdaily.kr/news/articleView.html?idxno=242091
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.