#DataCentersMayFaceOversupplyBy2028

Data Centers May Face Oversupply by 2028: Is the AI Boom Building Too Much, Too Fast?

The Empty Building

Imagine a company spending billions of dollars to build a huge data center. It buys land, installs powerful servers, prepares cooling systems, and signs contracts for electricity. The plan looks excellent because demand for artificial intelligence is growing rapidly. But a few years later, the building is ready while customers are spending less than expected, computing equipment has become outdated, or electricity costs have climbed too high.

What happens then? The company still has bills to pay, even if its expensive infrastructure is not being used fully.

This is the risk behind the growing debate about whether data centers could face oversupply by 2028. The concern is not that artificial intelligence will suddenly disappear. It is whether the enormous amount of infrastructure being built today will generate enough business to justify its cost tomorrow.

The Big Question

Could there really be too many data centers by 2028?

Yes, some markets could experience oversupply, but a global surplus is far from certain. Data centers are expanding rapidly to support AI models, cloud computing, online services, and business software. At the same time, companies are announcing major construction projects based on expectations of future demand.

The important difference is between announced capacity and useful, operating capacity. A project on a planning document does not automatically become a functioning data center. It still needs financing, permits, construction, electrical connections, equipment, and customers.

Even when a building is completed, its capacity may not be profitable if businesses do not rent enough computing power. That is why the number of projects alone cannot tell us whether the market is becoming overcrowded.

The Current Reality

Is supply already catching up?

Not everywhere. Recent market data shows that demand is still exceeding available space in many major locations. In the first quarter of 2026, the 16 largest data center markets tracked globally had approximately 16 gigawatts of capacity, while the average vacancy rate fell to 6.7%, down from 8.3% a year earlier.

In simple terms, more buildings were available than before, but the share of empty space became smaller. This suggests that new supply was not keeping pace with demand in those markets.

So why are people worried about oversupply by 2028? Because building a data center takes time. Companies must make investment decisions years before knowing exactly how much computing power customers will need. If many projects arrive at once, the balance could change quickly.

The Regional Difference

Will every country face the same risk?

No. Data center markets do not all grow at the same speed. One region may have more buildings than customers, while another struggles to find enough electricity and suitable space.

In February 2026, an industry forecast estimated a 16% probability of a data center supply glut across Asia-Pacific by 2030. China was identified as the region's most exposed market, while Singapore, Malaysia, and South Korea appeared better positioned under the forecast's assumptions.

This is a regional forecast, not proof that oversupply will happen by 2028. It does, however, show why investors must examine individual countries rather than treating the entire data center industry as one market.

Local conditions matter. A location with reliable electricity, strong internet connections, clear regulations, and nearby customers may attract businesses even when another location has too much capacity.

The Power Problem

Can a new data center operate without enough electricity?

No. A building can be complete and still be unable to run at full capacity if its power connection is delayed. AI servers consume large amounts of electricity, and cooling systems require additional energy to prevent equipment from overheating.

This creates a strange situation: the industry can invest heavily in new buildings while still struggling to activate them. In the United States, estimates published in 2026 pointed to a substantial gap between projected data center electricity demand and available power capacity through 2028.

That is not the same as having too many operating data centers. It means the industry could face a shortage of usable capacity even while construction continues.

For investors, the distinction is important. A company with a completed building but no reliable power may have less value than a competitor with fewer buildings that are already connected, operational, and earning revenue.

The AI Factor

What if AI demand keeps growing?

That is a real possibility. AI is being used for customer support, software development, research, image generation, business automation, and many other tasks. These applications need computing resources, and more frequent use can create additional demand.

However, growing AI usage does not guarantee that every infrastructure project will succeed. Businesses must still earn enough money from their services to cover computing expenses.

Consider two AI companies. The first has many paying customers and uses its computing resources efficiently. The second spends heavily on computing capacity but struggles to turn user activity into revenue. Both may contribute to demand today, but their ability to support long-term infrastructure costs can be very different.

This is why the number of AI users alone is not enough. Investors must also examine how much customers pay, how often they use the services, and whether the business can operate profitably.

The Hidden Risk

Could expensive equipment lose value before the buildings do?

Yes. Data center buildings can remain useful for decades, but the technology inside them changes much faster. New generations of AI chips can offer better performance, improved energy efficiency, or more computing power in the same space.

A facility designed around older equipment may need costly upgrades to remain competitive. If a company spends too much on hardware just before a major technology change, the expected returns can weaken.

There is also a financing risk. Developers often borrow money or depend on long-term customer contracts to support construction. If rental income falls, costs rise, or customers cancel plans, debt payments can become harder to manage.

The result may not be a completely empty building. It could instead be lower profits, delayed projects, falling valuations, or consolidation as stronger companies acquire weaker competitors.

The Warning Signs

How can we tell if oversupply is approaching?

There are several useful indicators to watch between now and 2028.

First, monitor vacancy rates. If available space rises consistently across several markets, it could indicate that supply is growing faster than demand.

Second, examine rental prices and contract terms. Falling rents, larger discounts, and shorter customer commitments may suggest that operators are competing more aggressively for business.

Third, look at utilization. This means how much of a facility's available computing capacity is actually being used. Low utilization can hurt returns even when a building appears busy from the outside.

Fourth, watch investment plans. If major customers reduce spending, delay projects, or cut their expected computing needs, developers may need to reconsider their expansion plans.

Finally, examine financing costs and electricity access. A project that depends on expensive borrowing or uncertain power supplies may struggle even if demand remains strong.

No single indicator proves an oversupply cycle is beginning. The strongest warning would be several of these signals appearing together across the same market.

The Investment Lesson

Does this mean the data center boom is ending?

Not necessarily. A growing industry can still produce failed projects, overvalued companies, and regional imbalances. The internet expanded enormously after the dot-com boom, but that did not prevent investors from losing money on businesses that had poor economics or unrealistic expectations.

The same lesson applies to AI infrastructure. Long-term demand can be real while individual investments are still unattractive.

Rather than assuming that every data center company will benefit, investors should look at debt levels, operating costs, customer commitments, available electricity, and the expected return on construction spending. They should also consider whether a business can adapt when technology changes.

For ordinary readers, the main point is simple: a large construction budget does not automatically mean a strong investment opportunity.

The Final Verdict

What should we expect by 2028?

Data centers may face oversupply in selected regions if construction grows faster than paying demand. But the available evidence does not establish that a global oversupply is inevitable by 2028. Several major markets are still experiencing tight supply, and electricity constraints may prevent planned capacity from becoming operational.

The outcome will depend on how quickly new facilities are completed, how much AI computing businesses actually use, how technology evolves, and whether customers generate enough revenue to support the investment.

For Dragonfly Official readers, the most useful approach is to separate the AI growth story from the financial performance of the companies building its infrastructure. Both matter, but they are not the same thing.

The real question is not simply how many data centers the world will build. It is how many will have reliable power, paying customers, efficient technology, and sustainable profits.

By 2028, which will matter more: the amount of data center capacity being built, or the money that capacity can actually earn?

‍$AVGO