AI data center costs: +66% for gas power plants, what changes for SMEs

Updated October 14, 2026

Gas power plants cost 66% more than two years ago, and they take 23% longer to build. If your company uses cloud services or AI tools, this number directly affects your next bill.

Does this increase in energy costs really affect my company too?

Yes, even if you don't build data centers. Every cloud or AI tool you use depends on energy infrastructure that is becoming more expensive: sooner or later, that cost shows up on your bill.

According to reports by TechCrunch , construction costs for natural gas power plants have increased by 66% in two years. Completion times have extended by 23%. This is not a local problem: the phenomenon is systemic, with its epicenter in the United States but with effects spreading on a global scale.

The mechanism is simple. Big data center operators — Microsoft, Google, Amazon, Meta — consume huge amounts of electricity. When energy costs more, their margins get squeezed. At that point, they have two paths: absorb the cost or pass it on to customers. In recent quarters, the trend is clearly toward the second option.

For an Italian SME using an advanced CRM, a marketing automation platform, or an artificial intelligence API (the service that allows integrating AI functions into software), this translates into evolving price lists and more expensive contract renewals.

What really happened: the numbers behind the +66%?

Electricity demand from data centers has grown faster than the capacity to produce it. The result is a market under pressure on two fronts: cost and construction speed.

The 66% increase in construction costs for gas power plants is not an isolated figure. It must be read alongside the 23% extension in construction timelines: it means that new energy capacity arrives later and costs more. A double disadvantage for those who need to power rapidly growing data centers.

According to IEA (International Energy Agency) , over the next five years, data centers will account for about 50% of the growth in electricity demand in the United States, with a total addition of over 420 TWh — a huge amount, comparable to the annual consumption of entire European countries.

The situation is also moving in Europe. Still according to the IEA , electricity consumption by data centers in the EU in 2026 will be 30% higher than 2023 levels. New facilities are already under construction, but generation capacity is struggling to keep up.

How does the energy cost reach my cloud bill?

The path is direct: gas power plant → data center → hyperscaler → SaaS provider → your company. Each link in the chain passes the price hike on to the next.

Hyperscalers — the big cloud providers like AWS, Microsoft Azure, and Google Cloud — build or lease data centers that consume industrial amounts of electricity. When energy costs rise, their operating margins shrink. In recent quarters, all three have revised the rates for various services, particularly those related to AI processing and machine learning.

But the price hike doesn't stop with hyperscalers. Many second-tier platforms—advanced CRMs, marketing automation tools, predictive analytics software—rely on third-party cloud infrastructure. These providers might also adjust their prices in upcoming contract renewal cycles.

Are Italian SMEs already at a disadvantage regarding energy costs?

Yes. Italian businesses pay 30% more for electricity than the European average. Every cloud price hike adds to an already high cost base.

According to Confindustria (based on 2025 Eurostat data), in the first half of 2025, Italian companies paid an average of 278 euros per MWh (megawatt-hour, the standard unit of measurement for large-scale consumption) for electricity. The European average was 216 euros: a 30% gap that weighs on the competitiveness of every business.

Added to this is what was noted by Confcommercio : in 2025, the average electricity bill for Italian service sector businesses was 28.8% higher than in 2019. The gas bill was even 70.4% higher. Anyone running a shop, a studio, or a service company knows this burden well.

In this context, every increase in cloud costs is not an abstract problem: it adds to an already compressed cost structure.

Why does the Italian cloud market keep growing despite the costs?

Because the benefits of cloud adoption still outweigh the price hikes for most companies. But rising spending makes it even more important to choose where to invest wisely.

According to Cloud Transformation Observatory of the Politecnico di Milano , in 2025 the Italian cloud market reached 8.13 billion euros, with a 20% growth compared to the previous year. SME spending on public and hybrid cloud hit 690 million euros, an increase of 18%, with an adoption rate stable at 67%.

Growth is real. But the stakes are rising too: the more an SME spends on the cloud, the more exposed it is to future price hikes. Those who don't monitor their spending risk having costs spiral out of control just as price lists go up.

The topic of the energy infrastructure supporting these investments is also explored in the article on FERC and the acceleration of AI data centers , which analyzes how new power grid rules are changing the outlook for tech companies. Regarding global investments, it is also worth reading the analysis on SoftBank and the 75 billion in French data centers : a measure of how high the stakes are.

Three concrete scenarios: what changes for an e-commerce, a professional firm, and a manufacturing company?

The impact varies based on how much each company depends on AI-intensive cloud services. Here are three cases to understand where your situation stands.

Illustrative examples: numbers are meant to show the reasoning, they are not real customer data.

The topic of AI infrastructure diversification is also central to the analysis on AWS and AI chips sold to third-party data centers : a sign that the market is fragmenting, with new options for those who want to reduce their reliance on a single provider.

Which energy alternatives could change the game?

Renewables, geothermal, and advanced batteries are in development, but none are yet able to replace gas as a baseload source for data centers in the short term.

Renewables—solar and wind—have a structural problem: intermittency. They do not produce energy consistently, and data centers need a continuous power supply, 24 hours a day. This is why natural gas will remain a structural component of the energy mix for the coming years.

Several hyperscalers are investing in long-term renewable energy purchase agreements (so-called PPAs, Power Purchase Agreements) to stabilize their costs. But these agreements take years to show their effects.

On the technology front, the efficiency of AI chips is improving rapidly. New processors consume less energy per unit of computation compared to previous generations. This could stabilize energy demand for equivalent workloads in the medium term, even though absolute demand will continue to grow.

Solutions like the GM sodium-ion batteries and the advanced geothermal from Fervo Energy are promising, but still in the scaling phase. In the short term, cost pressure remains.

Why are we talking about this now: do market data confirm the urgency?

Yes. Data center electricity demand is growing at a pace that generation infrastructure cannot keep up with, and Italy is already starting from a disadvantaged position.

According to IEA (Electricity 2026) , in the USA, electricity demand rose by 2.1% in 2025 and will grow by about 2% annually until 2030. About half of this increase is driven by data centers. It's not a temporary spike: it's a structural trend.

On the pricing front, according to IEA , in 2025, electricity prices for energy-intensive industries in the EU were on average about double those in the USA. A gap that makes European companies—and Italian ones in particular—structurally more vulnerable to global energy price hikes.

To this, we add a worrying figure from the Digital Innovation Observatory in SMEs of Politecnico di Milano : 76% of Italian SMEs have not invested and do not plan to invest in AI, and only 7% have started structured training programs. Those who don't prepare today risk having to play catch-up in a context of higher costs.

How to get started: concrete steps to protect your digital budget?

Five practical actions, in order of priority, to reduce exposure to cloud price hikes without stalling innovation.

What are the most common mistakes in this situation?

The three most common mistakes are waiting until renewal to notice a price hike, choosing AI tools solely for their features without looking at the pricing model, and failing to measure the return on digital investments.

How to tell if my cloud spending is under control?

Three indicators are enough to get a clear picture: monthly spending variation, cost per unit of AI output, and the percentage of services with fixed vs. variable pricing.

So, is it worth worrying about now or can it wait?

It depends on how much your company relies on AI-intensive cloud services. Those who wait risk having to manage price hikes in an emergency, with no room to maneuver.

It is worth it if your company uses AI APIs, predictive analytics, or intensive cloud workloads: price hikes are already underway and acting now gives you more negotiating leverage.

A middle ground If you use the cloud for standard applications (email, management software, basic CRM): monitor renewals and start building a spending baseline, but there is no urgent need to change everything.

Not yet if your company is still in the process of evaluating digital tools: you have time to choose solutions with predictable pricing from the start, without having to fix contracts you've already signed.

How we see it at SHM

For us, AI is a tool to give people back their time, not a cost item to be passively endured. We believe that integrating artificial intelligence into a company means, first and foremost, understanding the economic model behind it: how much it really costs, how it scales, and where it can become a problem.

Working with Italian companies across different sectors, we've often seen the same pattern: an AI tool is adopted for its features without analyzing the pay-per-use pricing. Then the renewal comes, or the end-of-month invoice, and the cost is double what was expected. Not because the tool doesn't work, but because no one did the math beforehand.

Our approach always starts with an analysis of expected costs and benefits before recommending any platform. If you are evaluating an Strategic AI integration in your company, or you want to understand how to optimize existing digital spending, we can help you build a clear picture. We also offer AI training for companies and support in the data processing and business intelligence , to turn numbers into concrete decisions. You can explore all topics related to digital infrastructure and AI in the cluster hub Project management and technology .

The energy market is volatile. Cloud costs will rise further. But those with a clear strategy — the right tools, well-negotiated contracts, defined KPIs — turn a risk into a competitive advantage.

Frequently Asked Questions

The issue isn't whether costs will rise, but whether your company finds out in advance or on the invoice.