Loading story…
Google warned that power-grid interconnection delays are becoming the primary bottleneck to expanding data center capacity for surging AI workloads, with some US utilities reportedly quoting 4–10 years for new data center connections and, in extreme cases, up to 12 years even to review interconnection requests. The company framed the issue as a transmission and interconnection constraint that is increasingly limiting cloud capacity expansion as data center electricity demand accelerates.
Microsoft, responding to community and regulatory concerns about higher residential electricity rates and local resource impacts tied to large AI data centers, said it will push for utility rate structures that ensure data centers pay the full cost of the electricity and required grid infrastructure upgrades needed to serve them. The company also highlighted efforts to reduce environmental externalities—targeting a 40% improvement in water-use intensity by 2030 and deploying a closed-loop cooling design (already used in Wisconsin and Georgia) intended to eliminate the need for potable water for cooling—underscoring that grid capacity, pricing, and resource constraints are becoming central operational risks for AI-scale infrastructure.
Mallory correlates global threat intelligence with your attack surface — know if you’re exposed before adversaries strike.
3 events from the most recent confirmed update back to the earliest known activity.
At an AEI event on January 15, 2026, Google's head of global sustainability and climate policy said delays in connecting new loads to the power grid had become the biggest obstacle to expanding data-center capacity. Google said utilities in some regions were quoting 4–10 years for new connections and up to 12 years just to review interconnection requests.
In a Microsoft blog post reported on January 13, 2026, the company said it would ask utilities and regulators to set data-center electricity rates high enough to cover full costs, support rate structures charging very large customers their cost to serve, and avoid seeking local property-tax reductions. Microsoft also said it aims to improve data-center water-use intensity by 40% by 2030 and highlighted closed-loop cooling deployments in Wisconsin and Georgia.
A Lawrence Berkeley National Laboratory study published in December 2025 projected U.S. data-center electricity consumption would grow from 176 TWh in 2023 to 325–580 TWh by 2028, underscoring mounting grid pressure from AI workloads.
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
2 references tracked. Mallory keeps watching after this page renders.
Map indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
A growing **data center capacity crunch** is putting enterprise AI, cloud migration, and broader digital transformation roadmaps at risk as power-grid limitations, permitting, and construction constraints slow new builds. Reporting citing a Sightline Climate analysis says 2025 saw nearly **6 GW** of capacity deployed, but **26% of projects were delayed**; for 2026, roughly **16 GW** is slated to come online, yet only about **5 GW** is reportedly under construction so far, with projections that **30%–50%** of planned projects could slip due to power availability and related bottlenecks. Separately, OpenAI’s planned *Stargate* AI data center expansion with **Oracle** and **SoftBank** has reportedly been delayed by governance and control disputes among partners, complicating efforts to rapidly secure large-scale compute. The reporting indicates OpenAI initially sought to own the data centers outright to reduce long-term dependence on third-party cloud providers, but investor concerns about upfront capital requirements and projections that OpenAI could run out of cash by mid-2027 pushed it back into negotiations; the delays reportedly affect plans to secure up to **10 GW** of compute capacity over three years, alongside previously reported commitments including Oracle’s proposed facility targeting **~2 million chips** and OpenAI’s stated intent to purchase **$300B** in compute over five years.
Jun 29, 2026
President Trump used the State of the Union to propose a “**ratepayer protection pledge**” aimed at curbing electricity price increases attributed to hyperscale AI data center growth, arguing the grid cannot sustain rapidly rising demand. The plan would require large technology companies to **build, bring, or buy dedicated power supplies** (including building their own power plants) for new AI data centers so that local ratepayers are less exposed to AI-driven load growth and associated cost increases. In parallel, industry investment in AI infrastructure continues to accelerate: **AMD and Meta** announced a reported **$100B** multi-year partnership for up to **6 gigawatts** of AI compute capacity using AMD’s *Instinct* GPU platforms (including MI450-based systems), *EPYC* CPUs, and *ROCm*, underscoring the scale of new deployments that intensify power and cooling requirements. Separately, commentary on AI’s societal impact highlighted how AI adoption is reshaping institutions (e.g., courts, academia, open source ecosystems) and amplifying “arms race” dynamics, but it did not address the specific US power-policy proposal or the AMD–Meta infrastructure deal.
Jun 29, 2026
Nvidia is preparing to resume shipments of its H200 data center GPUs to China following the U.S. government's decision to allow limited exports under strict licensing and a 25% tax. CEO Jensen Huang confirmed that demand for the H200 in China is extremely high, but actual deliveries are pending final regulatory approvals. The move comes amid ongoing U.S. export controls that have restricted Chinese access to advanced AI accelerators, prompting Nvidia to restart its supply chain and prepare for significant orders once licenses are finalized. As a result of these export restrictions, China's AI infrastructure market is rapidly consolidating around domestic champions such as Baidu and Huawei, who now control over 70% of the country's "GPU cloud" market using homegrown AI chips. This shift is supported by substantial investment in domestic silicon development and a strategic move by Chinese tech giants to vertically integrate hardware, software, and cloud services. Meanwhile, the global surge in AI data center construction is straining electricity grids, with companies like Meta and Microsoft exploring alternative power sources, including small modular nuclear reactors, to meet the growing energy demands of AI workloads.
Jun 29, 2026