All articles tagged with #ai data centers
US trade deficit widened in December as imports rose 3.6%, led by computer chips and other tech goods and boosted by capital goods for AI data centers. The 2025 goods deficit hit a record, while exports fell slightly; trade with China declined but deficits with Taiwan and Vietnam grew, and tariffs stopped delivering a manufacturing rebound.
X-Energy, backed by Amazon, just won the first U.S. nuclear fuel license in more than 50 years, allowing enriched-uranium fuel production at two Tennessee facilities for the next 40 years with production slated to begin in 2028; the license supports Amazon’s plan to power AI data centers with carbon-free nuclear energy and comes alongside a $500 million Amazon investment in X-Energy for SMR development.
AI firms are pouring billions into building sprawling data centers that consume vast water and electricity, contributing to higher energy bills and grid strain. In response, Meta, Amazon, and others are funding advertising campaigns and lobbying efforts to sway public opinion and policymakers as projects face opposition and delays across towns and states.
Micron Technology is riding a surge in AI‑driven demand for memory chips, boosting revenue, cash flow and buybacks. With a forward P/E around 11.5 and a projected $100 billion high‑bandwidth memory TAM by 2028, the stock trades at a material discount to growth peers, implying potential upside as AI data‑center spend remains robust. Yet memory is cyclical, so a long‑term horizon is prudent as buybacks reduce shares and earnings grow, even if near‑term growth moderates.
Advanced Micro Devices extended its rally to seven straight sessions as investors grow upbeat about demand for its AI-focused server CPUs, buoyed by expectations of a Q4 revenue beat and higher guidance driven by Turin data-center CPUs. Analysts project strong server-CPU momentum, with potential 10%–15% upsell pricing and at least 50%/server growth this year, supported by hyperscalers locking in capacity. Bernstein also lifted estimates on server momentum, while OpenAI’s Helios collaboration reinforces the AI narrative, though GPU competition with Nvidia remains a consideration.
AI data centers are expanding rapidly, with Virginia hosting the world’s largest cluster and major investments from Meta, Microsoft, and Amazon, but the U.S. electrical grid is aging and may struggle to meet rising energy demand. Officials have floated an emergency electricity auction through PJM to cover growing costs from these centers, while residential rates have climbed and utilities are exploring higher charges for large customers. Projections suggest data centers could consume about 6.7%–12% of US electricity by 2028, and cooling needs could push water use up by about 170% by 2030, highlighting the challenge of coexisting data centers and households.
The Trump administration and a bipartisan group of governors pressed PJM to reform the Mid-Atlantic/Midwest grid to power a wave of AI data centers, urging technology companies to finance new power plants for the region; the agreement covers about $15 billion in generation projects, calls for an emergency capacity auction to keep prices affordable, and directs PJM to act quickly to strengthen reliability while protecting residential customers.
A memory-chip shortage driven by surging AI data-center demand is lifting storage and memory costs, triggering pricing pressure across devices in 2026. Nothing’s Carl Pei warns of 30%+ price hikes on some models, and Samsung hints at repricing; Dell and HP signaled 15–20% increases in the near term. IDC expects smartphone prices to rise 3–8% and PC prices 4–8%, with global device shipments potentially down about 9%. TrendForce also projects reduced smartphone production, while larger brands may gain share as smaller players struggle.}
Trump intervenes in the 13-state PJM market, urging an emergency 15-year power auction and price caps to curb rising Northeast electricity costs tied to AI data centers, while governors press for faster new generation and an all‑of‑the‑above energy approach to sustain growth.
PJM Interconnection trimmed its 2027 summer peak-demand forecast to about 160 GW from roughly 164 GW, citing that some planned data-center projects lack firm electric-service commitments. Despite still-strong long-run growth (about 17% by 2030 driven by data centers) and a 2028 peak seen above the 2006 record of about 165 GW, the revision injects skepticism into AI-driven consumption projections and could influence capacity auctions and consumer electricity costs. Shares of major power producers fell on the news.
Microsoft says it will fund energy and grid upgrades near its AI data centers to prevent higher local electricity bills, while investing in water replenishment, schools, libraries, and AI-skills training in host communities, and it will not seek property tax abatements.
Applied Digital plans to power AI data centers with natural-gas-fired steam turbines from Babcock & Wilcox to bypass a looming gas-turbine delivery bottleneck, enabling earlier capacity. The company is targeting 1 GW online now and 5 GW in five years, with 2028 as the planned online window for the steam-turbine plants and a 15-year lease for 200 MW at Polaris Forge 2. Siemens Energy will supply the turbine-gen sets, and the move could give Applied Digital a multi-year lead over competitors by delivering capacity years earlier than traditional turbines—though growth hinges on turbine availability and power generation expansion.
Meta has signed nuclear power agreements with TerraPower, Oklo, and Vistra to support its AI data centers, including the upcoming Prometheus project in Ohio, aiming to add up to 6.6 gigawatts of clean energy by 2035, reinforcing its commitment to sustainable and reliable energy sources.
Meta has secured multiple nuclear power agreements with TerraPower, Oklo, and Vistra to power its AI data centers in Ohio, supporting up to 6.6 gigawatts of clean energy by 2035, including funding for new nuclear units and existing plant expansions, to ensure reliable energy supply for its data infrastructure.
OpenAI and SoftBank are investing $1 billion in SB Energy to expand AI data center infrastructure for the Stargate initiative, including building a large data center in Texas and deploying AI tools for SB Energy employees, as part of a broader push to support AI growth through significant energy infrastructure investments.