374 articles
Small modular reactors emerge as a potential solution for AI data centers' massive power needs, with over $15 billion in global SMR investments planned.
Indian manufacturers supplying AI data center components have surged up to 500% in value as hyperscalers invest billions in infrastructure worldwide.
AI data centers now consume more power than entire cities, making grid infrastructure stocks like Sterling, Infineon, and Vertiv crucial investment plays.
AI data center success now depends on securing megawatts of power, not just GPUs, as startups like Crusoe and Nscale prove with massive deployments.
Google considers Samsung for AI chip production as TSMC bottleneck intensifies, potentially breaking Taiwan's 73% foundry monopoly stranglehold.
Arm's neural rendering brings ray-traced graphics and AI-powered frame generation to smartphones, potentially reshaping mobile gaming with desktop-quality.
Marvell's new 102.4 Tbps switch chip tackles AI data centers' hidden bottleneck: network infrastructure that can stall entire training operations.
Amazon claims its data centers use 7x less water than industry average, but experts warn global metrics mask serious local water stress issues.
Tesla's planned Terafab chip facility could produce more silicon than the entire U.S. consumes annually, reshaping semiconductor independence.
Microsoft is quietly opening Windows 11's AI features to millions of older PCs with RTX 30-series GPUs, undermining the Copilot+ PC advantage.
TensorWave raises $350M from AMD to build an all-AMD AI cloud with 8,192 GPUs, betting against Nvidia's 90% market dominance on cost and memory advantages.
AI data centers create invisible heat zones warming surrounding areas by up to 9°C, potentially affecting 340 million people within 6 miles.
AI companies are ditching flexible GPUs for hard-coded inference chips that cut token costs by 80% but lock in specific models permanently.
General Motors enters the data center power business with sodium-ion batteries, targeting 165% growth in AI infrastructure demand by 2030.
Tech giants are forming joint ventures for 70% of AI data center deals as $5.2 trillion in infrastructure needs exceed any single company's capacity.
iOS 27's advanced AI features require 12GB unified memory on M3, M4, or M5 chips, plus latest iPhones, creating a tiered access system.
AI data centers bring towns $70M annually but spread $582B in grid costs across 67 million residents who can't vote on these projects.
Engineers design coral-like copper cooling fins that could cut data center water use by a third, offering a breakthrough for AI infrastructure.
NVIDIA's Blackwell chips demand 140 kilowatts per rack, forcing data centers to abandon traditional air cooling for liquid systems or immersion tanks.
MIT startup Ferveret adapts nuclear reactor cooling to cut data center energy use by 35%, potentially solving AI's massive power problem.
D-Matrix ships AI chips with built-in memory that run inference 10x faster than Nvidia GPUs, sidestepping the DRAM shortage choking the industry.
Cerebras' 68% IPO surge to $311 per share triggers a financing frenzy among AI chip startups, with Groq, SambaNova and Tenstorrent racing to go public.
Taiwan attracts $5+ billion in AI data center investments from AWS, Google, Microsoft, NVIDIA, and Foxconn as Asia's computing hub emerges.
The U.S. treats Canada's AI infrastructure as vital to national security while using trade rules to block Canadian digital governance of those same data.
Cerebras challenges Nvidia's AI dominance with inference chips, drawing Wall Street coverage projecting 70% upside and 40-50% market share potential.
Xiaomi achieves 1,000 tokens per second AI inference on standard GPUs, matching what Cerebras and Groq built with millions in custom chips.
FERC's approval will restart Three Mile Island by 2027 to power AI data centers, cutting the timeline by four years through creative grid solutions.
AI data centers are adopting liquid cooling, heat recovery, and renewable energy investments to slash carbon emissions while meeting reliability demands.
NVIDIA partners with Samsung on custom AI chips, signaling a shift from GPUs to specialized ASICs that could save companies billions annually.
Google orders 3 million AI chips from Intel while Nvidia evaluates the foundry, breaking TSMC's monopoly and reshaping the chip supply chain.
Nigeria's 209 trillion cubic feet of gas reserves could power AI data centers as tech giants seek energy, but execution challenges remain.
Apple's macOS 27 Golden Gate will only run on M-series Macs, officially ending Intel support and unlocking AI features impossible on older hardware.
AI data centers will consume up to 73 billion gallons of water annually by 2028, with each ChatGPT query using 519 milliliters when including power.
Amazon, Google, and Microsoft are building custom AI chips to challenge Nvidia's dominance, threatening the company's near-monopoly position.
The 2028 Russian uranium ban could reshape AI data center power as tech giants race to secure nuclear energy before the hard deadline.
NVIDIA's RTX Spark brings Apple-style unified memory plus CUDA AI support to Windows, forcing Apple to rethink M5 architecture for the first time since.
Pennsylvania's $90 billion AI data center boom transforms the Rust Belt state into America's unexpected infrastructure powerhouse with abundant energy.
AMD launches Stardance, a summer challenge where teens build AI apps and robotics projects to earn coins redeemable for professional hardware prizes.
Nvidia's Groq acquisition creates hybrid GPU-LPU systems that could dominate the AI inference market by combining training and deployment chips.
Intel's Gaudi2 chips offer researchers free access to 336 AI accelerators, providing a cost-effective alternative to NVIDIA GPUs for training models.
AI data centers could consume up to 12% of U.S. electricity by 2028, potentially driving up power bills as the grid struggles to keep pace.
Nashville Zoo's 180,000-signature petition opposes a data center project 50 yards from endangered clouded leopards, citing noise and lighting threats.
Cerebras' stock fell 20% after its IPO despite a $20 billion OpenAI deal and 15x faster inference chips, exposing AI hardware investment risks.
Apple's M5 Ultra chip delivers 72 trillion AI operations per second on your desktop, keeping data private while eliminating cloud delays.
AI data center demand is driving $250M in climate tech funding as energy startups pivot from policy bets to infrastructure economics.
Texas counties lack legal power to stop 248 planned data centers targeting rural areas, leaving communities defenseless against AI infrastructure boom.
Meta plans to sell stock to raise billions for AI infrastructure, betting that owning data centers and power grids matters more than algorithms.
New York becomes the first state to pause large data center permits, forcing AI companies to navigate political risk alongside power shortages.
NVIDIA launches Nemotron Coalition with startups to distribute open AI models, expanding beyond GPU sales into software partnerships and ecosystems.
Constellation Energy's nuclear bet on AI data centers faces transmission delays and missing contracts, sending stock down 26.3% in six months.
