#DataCentersMayFaceOversupplyBy2028

Data Centers May Face Oversupply by 2028, But the Bigger Risk Is a Mismatch

What Just Changed?

The data center boom has entered a new phase. Massive amounts of new capacity are being planned and built to support artificial intelligence, cloud computing and growing digital services. Yet the market is still struggling with a basic problem: demand is arriving faster than usable capacity can be delivered.

That creates an unusual setup. The industry can experience shortages today while still creating the conditions for oversupply in some markets later. The question is no longer simply whether enough data centers will be built. It is whether the right facilities, with the right power, technology and customers, will come online at the right time.

What Happened First?

The first major signal was the explosion in AI demand. Powerful AI models require enormous computing resources, and companies have been racing to secure data center space, electricity and advanced processors.

By the first half of 2026, North American data center demand had reached record levels. Around 25 gigawatts of capacity was absorbed during the first six months, roughly double the level from a year earlier. Vacancy remained extremely low even as construction accelerated.

That tells us something important: the current market is not suffering from a simple lack of demand.

What Came Next?

The construction pipeline then became enormous. More than 66 gigawatts of data center capacity was under construction across North America, while a large majority of that capacity had already been committed to customers.

At the same time, construction itself became harder. Electricity connections, grid upgrades, equipment availability, land and local infrastructure became major bottlenecks.

In other words, developers could announce huge projects, but announcing capacity is not the same as delivering usable computing capacity.

Why Does 2028 Matter?

2028 is important because many projects being planned or contracted today are designed to become operational around that period.

This creates a timing risk.

If AI demand continues growing rapidly, new capacity could be absorbed almost immediately. But if demand growth slows while thousands of megawatts reach completion together, certain markets could suddenly have more capacity than customers need.

That would be a very different environment from today.

Where Could Oversupply Appear?

Oversupply does not necessarily mean the entire global data center industry will collapse into excess capacity.

The bigger possibility is a regional or technological mismatch.

One location may have plenty of buildings but not enough electricity. Another may have cheap power but poor network connections. A facility designed for one generation of AI hardware may become less attractive when newer, more efficient processors arrive.

A data center can therefore exist physically while still being economically unattractive.

What Is Happening With Investment?

Investment is moving at extraordinary speed. Global data center spending is projected to rise dramatically through 2028 as companies compete to build the infrastructure required for AI.

That spending creates opportunities for developers, power companies, equipment manufacturers, construction firms and technology providers.

But rapid capital spending also increases the importance of discipline.

When an industry becomes one of the biggest investment themes in the market, projects can sometimes be approved because everyone expects future demand to remain strong. That works when demand keeps accelerating. It becomes dangerous when expectations run ahead of reality.

What Does Current Demand Tell Us?

The strongest evidence today still points toward tight supply rather than broad oversupply.

North American vacancy has remained around historically low levels, while much of the capacity under construction is already preleased. Major cloud providers are also reporting strong commitments for future capacity.

This means the phrase “data center oversupply by 2028” should not be treated as a certainty.

It is better understood as a risk that could emerge if construction catches up with demand faster than expected.

What Could Change the Picture?

The most important variable is AI usage.

If businesses and consumers continue using AI at a rapidly increasing rate, demand for computing could remain strong enough to absorb much of the new capacity.

But if AI adoption slows, companies reduce infrastructure spending, model efficiency improves faster than expected, or customers delay large deployments, the supply-demand balance could change quickly.

Better AI models can actually create another twist. More efficient computing can reduce the amount of hardware needed for a given workload. That could slow the growth in physical capacity required even while total AI usage continues rising.

Why Is Power So Important?

Electricity may become the strongest filter separating valuable data centers from weaker projects.

A building without reliable power cannot serve large AI workloads. Grid connections can take years, and some planned projects have already faced delays because sufficient electricity was unavailable.

This means the most valuable capacity may not simply be the largest capacity. It may be capacity that has reliable power, strong network connections, efficient cooling and a committed customer.

That distinction could become much more important by 2028.

What Happens to Older Hardware?

Another issue is the speed of technology change.

AI processors can become outdated much faster than the buildings that contain them. A data center may have a useful life measured in decades, while the computing equipment inside it can be replaced in a much shorter cycle.

This creates a possible mismatch between long-term infrastructure investment and rapidly changing technology.

Facilities that can upgrade efficiently should have an advantage. Facilities built around inflexible designs may struggle to compete.

What Should Investors Watch?

The next few years should be about utilization, not just construction announcements.

Watch how quickly new capacity is leased. Watch vacancy rates. Watch whether major customers continue signing long-term contracts. Watch electricity availability and project delays.

Also watch AI infrastructure spending. If major technology companies suddenly begin cutting capital expenditure, that would be a much stronger warning sign than a single delayed project.

Hardware prices and rental rates matter too. Falling prices are not automatically bearish, but falling prices combined with declining utilization would suggest that supply is starting to outrun demand.

What Is the Real 2028 Risk?

The real risk is not simply “too many data centers.”

It is too much of the wrong capacity in the wrong place, arriving at the wrong time.

Some markets could remain extremely tight while others develop excess capacity. Premium facilities with secured power and strong customers may continue performing well even if weaker projects struggle.

That is why headline construction numbers can be misleading. A massive pipeline does not automatically mean massive usable supply.

What Comes Next?

The data center industry is entering a period where execution will matter more than announcements.

AI demand remains powerful, and current evidence still shows serious capacity shortages. At the same time, the scale of planned investment means the market cannot assume that every future megawatt will be equally valuable.

By 2028, the winners may be the operators that secured power early, built flexible facilities, controlled costs and matched capacity closely with real customer demand.

The biggest lesson is simple: today’s shortage does not guarantee tomorrow’s balance.

Stay careful, question the headlines, and think independently before deciding what the next phase of the data center boom will look like.

‍$AVGO