Half of New US Data Centers Stalled or Canceled

Plans for new data centers in the United States are hitting a major wall.

Even though giant tech firms want to spend billions of dollars, about half of the new projects for this year are on hold.

Construction is stopping because there is not enough power and parts to go around.

A Power Supply Crisis

Building a place for AI to live takes more than just fast chips.

It needs a massive amount of electricity and special gear like transformers and large batteries.

Right now, the wait time for a single high-power transformer has jumped from two years to five years.

These delays are making it very hard for companies to finish their work on time.

The China Connection

The United States is still very dependent on other countries for building materials.

Even with efforts to make more things at home, about 30 percent of the parts used to manage power still come from China.

Companies are also buying more batteries and equipment from places like Mexico and South Korea to fill the gaps.

If these supply chains stay stuck, all the money in the world will not be able to buy more computer space.

Future Growth Stalled

The goal was to turn on a huge amount of new power for AI by next year.

However, only about one-third of that power is actually being built right now.

With the grid already busy supporting electric cars and home heating, there is very little room left for new data centers.

Technology leaders must find a way to fix the power grid before the big AI dreams can become a reality.

A major player in the AI world just went dark, and the silence was loud.

DeepSeek, the Chinese AI firm that became a global star last year, suffered a service failure that lasted over seven hours.

This was the longest time the platform has been down since it first went viral.

Millions of people who use the tool for work and coding were left with nothing but error screens.

The Cost of Being Popular

Success brings its own set of problems for fast-growing tech companies.

DeepSeek has more than 350 million users and keeping the lights on for that many people is a massive job.

The company did not say why the system failed, but it shows that even the best AI models are only as good as the wires and servers they run on.

When a tool becomes this big, every minute it is offline causes a huge ripple through the business world.

Reliability Is the New Race

We are moving past the time when just having a smart AI was enough to win.

Now, the focus is shifting to which company can stay online without breaking.

For people who build their own apps using these tools, a long outage like this is a big warning sign.

If you cannot trust the tool to work every day, you might look for a safer choice.

This event proves that building a great brain is hard, but building a system that never sleeps is even harder.

Stable and steady service will be the most important goal for AI leaders this year.

Many firms find the cloud costs much more than expected, often triple their first estimates. This has led to a state of quiet panic as leaders deal with expensive workloads and major outages. Some are now moving tasks back to their own servers or using sovereign clouds to regain control.

There is a big difference between reliability and resilience. While a vendor handles specific parts, the user must design the whole system to survive a disaster. AI adds more risk because it is complex and spread out. If one small link breaks, the whole system can fail.

Using AI to quickly build apps, known as vibe coding, can lead to massive, surprise bills. AI often makes poor choices that use too much power or memory. While these tools feel fast, they often lack the human touch needed for efficiency, creating a major financial mess that is hard to fix later.

Microsoft and a company called Armada are joining forces to change how we use computers in far-off places.

They are building a special system that lets groups like the military or energy workers run powerful AI even when they have no internet.

This is a big deal for groups that must keep their data very safe and close to home.

Small Data Centers in a Box

The core of this plan uses something called Galleon modular data centers.

Think of these as small, tough computer rooms that can be moved anywhere, from a desert to a ship.

Inside these boxes, Microsoft’s software allows teams to process huge amounts of info right where it is found.

By doing the work locally, these groups do not have to worry about slow connections or sending private files across the world.

Total Control Over Data

Many governments and big businesses are worried about who can see their information in the public cloud.

This new partnership gives them a way to keep everything inside their own "trusted boundary."

It supports high-level security and follows strict rules that many industries must obey.

Even if the box is totally cut off from the rest of the world, the AI keeps working.

Intelligence Without the Internet

In the past, most AI needed a giant, central data center to think, but that is changing.

Now, real-time decisions can be made on the spot, which is vital for safety and quick action.

Microsoft and Armada are making it possible to have smart systems in the most difficult spots on Earth.

This move ensures that critical work stays resilient and secure no matter where the mission leads.

Michigan is about to become a major hub for the tech world.

A company called Related Digital is close to getting $16 billion to build a massive data center.

This project is being built specifically for Oracle to use.

It marks one of the largest single investments in computer infrastructure we have seen recently.

A Massive Scale

The project is so large that it needs its own giant power source.

Current reports say the site will use about 800 megawatts of electricity.

That is enough power to run hundreds of thousands of homes at the same time.

Oracle plans to use this space to run its cloud and AI services for customers everywhere.

Economic Impact

Building a center this big will create many local jobs in construction and technology.

It shows that companies are looking for new places to build away from the usual tech hubs.

Michigan offers the space and resources that these giant AI systems need to grow.

The state is quickly turning into a key player for the future of the cloud.

Oracle is making a massive bet that huge physical buildings are the only way to win the AI race.

Microsoft is spending a lot of money to grow its cloud business in Southeast Asia.

The tech giant plans to put $5.5 billion into Singapore over the next few years.

This money will build more data centers and help people learn how to use AI tools.

Singapore is already a top spot for tech, and this move makes it even more important for the region.

Big Plans for AI

Most of this money will go toward building more space for computers and data.

Microsoft wants to make sure its cloud services are fast and can handle the newest AI apps.

As more companies use these tools, they need more power and storage to keep things running smoothly.

This investment will help local businesses grow by giving them the best technology available.

Training and Support

It is not just about buildings and wires.

Part of the plan is to teach 100,000 people in Singapore how to work with AI.

Microsoft is also working with the local government to make sure the tech is used in a safe way.

By helping people learn new skills, the company ensures there are enough workers to run these systems.

Microsoft is betting big that Singapore will stay a leader in the global tech market for a long time.