AI Infrastructure
The compute, chips, data centers, and energy behind AI, and the constraints that shape what gets built.
47 stories · page 2 of 2
Nvidia Co-Signs OpenAI's $500B Ohio Data Center Lease
OpenAI is in advanced talks to lease a massive computing campus being built on former federal nuclear land in Ohio, with Nvidia acting as financial guarantor for both the lease and the developer's financing, signaling that AI chip makers are now behaving more like co-investors than suppliers.
June 11, 2026 · 2 min read
Meta Builds Its First India Data Center With Reliance
Meta and Reliance Industries announced a deal to build a 168-megawatt data center in Gujarat, India, a move that reflects how global companies are now treating India as a serious location for AI infrastructure, partly driven by new Indian data rules that require customer data to stay inside the country.
June 10, 2026 · 2 min read
China Opens First Wind-Powered Underwater Data Center
China has put the world's first wind-powered underwater data center into full commercial operation off Shanghai, using seawater for cooling instead of electricity-hungry air conditioning, at a time when AI's appetite for power is making energy the single biggest constraint on how fast any country can build out AI infrastructure.
June 9, 2026 · 3 min read
Google Pays SpaceX $920M Monthly for AI Computing
Google signed a $30 billion deal to rent computing capacity from SpaceX's xAI data centers, a move that signals just how badly the world's biggest AI companies are scrambling for computing power they cannot build fast enough themselves.
June 6, 2026 · 3 min read
US Voters Ban Data Centers as Opposition Spreads
Residents of Monterey Park, California just became the first US voters to permanently ban data centers via ballot, part of a fast-growing national movement that has already blocked over $156 billion in projects and is starting to affect where and how quickly AI infrastructure gets built.
June 5, 2026 · 2 min read
Taiwan Builds the World's AI Factories, Then Uses AI to Run Them
NVIDIA's latest showcase at GTC Taipei reveals that Taiwan's manufacturers, from Foxconn to TSMC, are not just making the world's AI hardware but applying AI to their own factory floors, cutting defect inspection time by two-thirds, speeding up production layouts by 70%, and deploying humanoid robots on assembly lines.
June 10, 2026 · 3 min read
AWS Rebuilds Search Infrastructure for AI Agents
Amazon has overhauled a core piece of its cloud infrastructure to handle the unpredictable, burst-heavy traffic patterns of AI agents, and the move signals a broader reshaping of internet plumbing that will affect costs and capabilities for any business deploying or buying AI-powered tools.
June 5, 2026 · 2 min read
AWS Rewires Its Data Centers, Cuts Hardware Use by 69%
Amazon has replaced the decades-old internal wiring design used inside its data centers with a new system that uses 69% fewer networking devices and 40% less energy, and has already made it the default for most new AWS facilities worldwide.
June 1, 2026 · 2 min read
Korean Chip Startup Raises $135M to Cut AI Server Costs
XCENA, a South Korean startup founded by Samsung and SK Hynix veterans, has raised $135 million on the premise that the real bottleneck in running AI at scale is not computing power but memory, and that solving it could let companies do the same work on far fewer servers.
June 1, 2026 · 3 min read
Microsoft in Talks to Supply Its Own Chips to Anthropic
Microsoft is negotiating to give Anthropic access to its Maia 200 chips, a move that would mark the first real-world test of Microsoft's own AI hardware and deepen a relationship that already involves a $5 billion investment and a $30 billion cloud spending commitment.
June 2, 2026 · 2 min read
xAI Buys $2.8B More Gas Turbines While Facing Pollution Lawsuit
Despite being sued for illegally running dozens of unpermitted gas turbines near Memphis, xAI's IPO filing reveals it is committing another $2.8 billion to buy more of the same equipment, exposing a real risk that courts could shut down its AI operations entirely.
June 2, 2026 · 3 min read
Google Launches Open-Source Runtime for Long-Running AI Agents
Google has released Agent Executor, a free open-source tool that keeps AI agents running reliably for hours or days without crashing, losing progress, or getting stuck, directly addressing the biggest reason enterprise AI agent pilots fail before reaching full production.
June 2, 2026 · 2 min read
Alibaba Launches AI Chip and Model Built for Autonomous Tasks
Alibaba has released a new processor and AI model both designed to run complex, long-running automated tasks with minimal human input, signalling that the company is building its own end-to-end AI supply chain rather than simply replacing blocked US chips.
June 2, 2026 · 3 min read
NVIDIA Ships Its First Custom CPU to OpenAI, Anthropic, Oracle
NVIDIA has delivered the first units of its Vera CPU, a custom chip built specifically for AI agents, to Anthropic, OpenAI, SpaceXAI, and Oracle, marking the moment that a quiet but significant shift in AI infrastructure, from GPU-dominated to CPU-heavy, moves from theory to production hardware.
June 5, 2026 · 2 min read
NVIDIA and Dell Push Enterprise AI Out of the Cloud
At Dell Technologies World, Jensen Huang and Michael Dell announced a full hardware and software platform for running the latest AI systems inside a company's own data center, backed by real customer deployments and a cost-per-token improvement that makes on-premises AI decisively cheaper than the cloud for high-volume workloads.
June 5, 2026 · 2 min read
Kevin O'Leary's Utah Stratos Data Center: What Is Actually at Stake
A $100 billion AI data center backed by Kevin O'Leary was approved in rural Utah using a little-known military zoning authority, and the backlash from residents reveals a much broader pattern playing out across the US and Canada where AI infrastructure is quietly consuming public resources at a scale most communities are completely unprepared for.
June 2, 2026 · 3 min read
Databricks Kills the Data Pipeline. Here Is What Replaces It.
Databricks has launched a feature that automatically moves live business data from its Lakebase database into its analytics platform without any pipeline, a shift that removes a costly and fragile layer of data infrastructure that most mid-sized and large companies currently pay heavily to maintain.
June 2, 2026 · 2 min read
AI's Real Power Problem Is Not the Grid, It's the Milliseconds
As AI data centers push local power grids toward their limits, a deeper and less-discussed problem is emerging: the violent, split-second power surges produced by large groups of computing chips working in sync, which ordinary backup systems cannot handle and which now threaten to stall the entire AI buildout.
June 2, 2026 · 3 min read
Google and SpaceX's Space Data Center Talks, Explained
Google is in talks with SpaceX to launch data centers into orbit, a move driven less by proven economics and more by a desperate scramble for power on the ground and a timely need to boost SpaceX's upcoming IPO story.
June 2, 2026 · 3 min read
How Faster Image Delivery Is Becoming a Revenue Tool
Imgix, which processes 8 billion images daily for brands like Spotify and Porsche, has cut its image processing time in half by moving to Google Cloud's newest GPU-powered servers, a shift that reveals a broader pattern: the infrastructure behind visual media is quietly becoming one of the most direct levers businesses have over their online revenue.
June 2, 2026 · 2 min read
Nvidia's Bet on Micro Data Centers Near Power Substations
Nvidia and partners are building small data centers plugged directly into spare grid capacity at electrical substations across the U.S., a move that signals where AI infrastructure is heading next and has real consequences for businesses that rely on AI services, energy costs, and local grid stability.
June 2, 2026 · 3 min read
The Real AI Cost Problem Is Storage, Not Chips
Google has launched a high-speed cloud storage product aimed at AI workloads, and while it sounds technical, it points to a much bigger issue that every business running or evaluating AI needs to understand: the expensive computing hardware powering AI sits idle far more than anyone admits, and storage is usually the reason.
June 2, 2026 · 2 min read
The Hidden Cost of Running Your Own AI Engine
Superhuman's decision to hand off its AI serving infrastructure to Databricks after years of building it in-house reveals a pattern that every organisation deploying AI at scale will eventually face: the real cost of running your own AI is not hardware, it is the engineering time consumed just keeping the lights on.
June 2, 2026 · 2 min read