AI data center costs: +66% for gas power plants

The context: when AI demand meets the physical limits of energy

In just a few years, generative artificial intelligence has completely changed the nature of global energy consumption. Large language models, real-time inference systems, and training pipelines require unprecedented amounts of electricity. However, the power grid and generation infrastructure are not upgrading at the speed of software.

According to reports by TechCrunch , the cost of building natural gas power plants jumped by 66% over two years. Plus, construction times dragged on by 23%. This isn't just happening in one spot: it is a widespread trend, starting mainly in the US but sending ripples across the globe.

So, we're facing a physical bottleneck. Electricity generation capacity can't keep up with the growing demand from data centers. This structural imbalance has direct consequences on the operating costs of the entire digital supply chain.

The numbers that matter: +66%, 23%, and the weight of kilowatts on AI

The 66% increase in construction costs is significant in itself. However, it becomes even more relevant when read together with the 23% lengthening of construction times. Together, these two indicators describe a market under pressure on both critical dimensions: cost and speed.

According to the analyses of IEA — International Energy Agency , the electricity demand of data centers could double by 2026 compared to 2022 levels. Therefore, the problem is not going to ease anytime soon. On the contrary, the pressure on installed capacity is set to increase further with the spread of next-generation AI models.

Furthermore, the financial component must be considered. The capital needed to build new generation capacity has grown disproportionately. As a result, major data center operators—Microsoft, Google, Amazon, Meta—are internalizing much higher energy costs than three years ago. These costs, sooner or later, are passed down the value chain.

In particular, McKinsey it is estimated that energy infrastructure investments for data centers could exceed 500 billion dollars by 2030. Therefore, the scale of the phenomenon is systemic, not episodic.

From the gas power plant to the cloud invoice: the cost transmission chain

To understand the practical implications, it is useful to trace the path that connects a gas power plant to an Italian company using cloud services or AI tools. The mechanism is straightforward, even if it is not always visible to those who buy software licenses.

First of all, major hyperscalers build or rent data centers. These data centers consume massive amounts of electricity. So, when energy costs go up—due to power plants that are more expensive to build and run—the operating margins of hyperscalers get squeezed.

After that, these providers have two choices: absorb the costs or pass them on to customers. In recent quarters, we have seen a growing trend toward the second option. In fact, Microsoft Azure, AWS, and Google Cloud have updated the pricing for several services, especially those related to AI processing and machine learning. Therefore, small businesses using AI APIs or cloud environments for data analysis are already dealing with changing price lists.

On top of this, we need to think about the impact on tier-two SaaS vendors. Lots of marketing automation platforms, advanced CRMs, and predictive analytics tools rely on third-party cloud infrastructure. Because of this, these vendors might hike their prices in upcoming contract renewals too.

Strategic reading: what it means for Italian SMEs in B2B and retail

Italian SMEs are in a unique position. On one hand, they are speeding up the adoption of advanced digital tools, driven by the need to compete in increasingly automated markets. On the other hand, they operate with limited budgets and a high sensitivity to cost changes.

Therefore, the energy dynamic described in this article is not an abstract issue. It is a variable that concretely enters into digital investment planning. In particular, anyone evaluating the adoption of AI solutions for customer service, content generation, or predictive sales analysis must incorporate this variable into their cost models.

However, this doesn't mean giving up on innovation. On the contrary, it means adopting a more selective and conscious approach. We at SHM Studio we work daily with companies that need to balance digital ambition and economic sustainability. In fact, choosing the right tools — in terms of computational efficiency and pricing model — becomes a concrete competitive advantage.

Similarly, retail companies pouring money into AI customer experience personalization need to keep in mind that inference costs could go up. So, it is smarter to lean toward solutions with predictable pricing and efficient usage models instead of platforms that charge strictly on a variable pay-as-you-go basis.

The work in progress: the unknowns that no one can quantify

There are some variables that make an accurate prediction about cost trends difficult. First of all, the speed of renewable energy development. If the construction of new solar and wind capacity were to accelerate significantly, the pressure on natural gas could ease.

However, renewables have intermittency issues that currently make them insufficient to cover the base load of data centers. Therefore, natural gas remains a structural part of the energy mix for the coming years. Despite this, several hyperscalers are investing in long-term renewable energy purchase agreements, trying to stabilize their energy costs.

Plus, AI chip efficiency is improving fast. The new processors from NVIDIA, AMD, and internal teams at Google and Amazon use less power per unit of compute than older generations. Because of this, the energy demand for the same workload might level off in the medium term, even though total demand keeps going up.

In summary, the picture is characterized by structural uncertainty. SMEs would do well not to assume stability of cloud costs in their multi-year plans.

Operational implications: how to navigate digital investments

Faced with this scenario, some operational guidelines can help SMEs navigate the context with greater awareness. These are not universal recipes, but rather common-sense principles applied to a transforming market.

• Cloud contract review: it is advisable to check the price adjustment clauses in contracts with cloud service providers. Furthermore, it is useful to negotiate multi-year commitments when possible, to lock in current rates.
• Computational efficiency: prioritize AI solutions that optimize computational resource consumption. In fact, a smaller, specialized model can offer equivalent performance at a lower cost compared to a large general-purpose model.
• Supplier diversification: avoiding dependence on a single hyperscaler. Therefore, evaluate multi-cloud architectures or on-premise solutions for more intensive workloads.
• Real-time cost monitoring: implement FinOps tools to keep cloud spending under control. Consequently, it is possible to intervene quickly in case of anomalies or unexpected increases.
• AI ROI assessment: every investment in AI tools should be accompanied by rigorous return measurement. Thus, it is essential to define clear KPIs before adoption, not after.

For SMEs seeking structured support in this phase, the services of Digital marketing and AI by SHM Studio include a preliminary analysis phase of expected costs and benefits. Furthermore, the team at SHM Studio can support the choice of the most suitable platforms for the company's risk profile and budget.

Outlook: a market that is recalibrating, not stopping

It is important not to see this data as doom and gloom. Rising energy costs won't stop AI adoption. Still, they will weed out the weaker solutions on the market. That's why the platforms that survive and thrive will be the ones offering real value at a price people can afford.

For Italian SMEs, this is actually a favorable time to adopt a more mature approach to AI. Instead of chasing every new technological trend, it is better to focus on specific use cases with measurable impact. For example, customer service automation, campaign personalization Linkedin or Google Ads , or improving content quality through tools for assisted copywriting .

Finally, the SEO and the web presence remain investments with a stable cost-benefit ratio, less exposed to fluctuations in global energy costs. Therefore, diversifying the digital investment mix is a prudent strategic choice now more than ever. To delve deeper into these topics, the SHM Studio blog provides up-to-date analyses tailored to the needs of Italian businesses. For personalized advice, you can contact the team through the page contacts .