Fortune reported on July 26, 2026 that data centers have not always pushed electricity prices in the direction many people assume. Citing new Electric Power Research Institute research, the article said data centers historically caused average retail electricity rates to fall modestly from 2015 through 2024. The catch is the part business owners should pay attention to: that past pattern may not hold if the current AI infrastructure buildout creates large fixed costs before the expected demand appears.
That matters because AI is no longer just a software subscription sitting neatly inside the technology budget. The tools a business uses may depend on cloud providers, model vendors, data centers, power contracts, grid capacity, network routes, and local permitting decisions. When those assumptions move, pricing and reliability can move with them.
The Business Risk Is In The Assumptions
The practical issue is not whether AI data centers are good or bad. It is whether a business has approved AI projects, cloud migrations, automation tools, or data-heavy workflows without understanding the infrastructure assumptions underneath them.
Fortune reported that EPRI's work found economies of scale helped explain why data center load could lower average prices in the past. But the same article warned that the trend could reverse if utilities and grid operators build capacity for AI demand that does not arrive as expected. In that case, the fixed costs may be spread across fewer customers than planned.
For New Jersey and Pennsylvania businesses, the PJM angle makes this feel less abstract. PJM serves a large regional power market that includes New Jersey and Pennsylvania, and recent reporting has tied rising capacity-cost concerns to data center demand. Even if your company never hosts a server rack, your vendors and cloud providers may be exposed to the same power and capacity pressures.
What Owners Should Ask Before Approving The Spend
Before signing off on a large AI, cloud, or SaaS expansion, owners should ask for more than a feature demo and a monthly license price. A better review asks how the service would hold up if infrastructure costs change.
- What pricing assumptions are locked in? Ask whether the quote includes usage-based compute, storage, bandwidth, model calls, support tiers, or renewal escalators.
- Where could power-sensitive costs show up? Ask whether the vendor can pass through cloud, hosting, or capacity increases during the contract term.
- What happens if the AI workload grows faster than expected? Ask for a usage forecast and a spending guardrail before a pilot becomes a production dependency.
- What happens if the workload disappoints? Ask whether the business can scale down, pause, export data, or leave without paying for unused capacity.
- Who owns continuity? Ask how the vendor handles regional outages, cloud constraints, degraded AI features, and fallback processes.
The Budget Conversation Comes Before The Tool Choice
Many AI proposals start with the exciting part: faster reports, automated service tickets, easier customer support, or better document handling. Those benefits can be real. The problem is that a useful pilot can become expensive when usage, data volume, retention, integration work, and vendor infrastructure costs are not visible early.
A practical owner review should separate three questions. What business outcome are we buying? What infrastructure or vendor dependency makes it possible? What would make the cost or reliability story change after approval?
That framing helps avoid a common mistake: treating AI like a simple app purchase when the real commitment may be closer to a cloud, data, and operations decision.
A Sensible Next Step
If your organization is planning a meaningful AI or cloud expansion this year, ask your IT provider or internal team for a short dependency review before approving the budget. It should identify the vendors involved, the expected usage model, the contract renewal risks, the data-retention costs, the fallback process, and any regional infrastructure concerns that could affect pricing or uptime.
The utility bill may not arrive with the AI vendor's logo on it. But the cost assumptions behind that bill can still shape the technology plan.
Sources and further reading
- Data centers were actually making electricity costs cheaper, but the $7 trillion buildout with no guaranteed AI demand is threatening the trend
- Have Data Centers Raised Your Electric Bill? Causal Evidence from the United States
- Powering Intelligence 2026: Updated Scenarios of U.S. Data Center Electricity Use and Power Strategies
- Trump expands a voluntary pledge to protect consumers from high utility bills from AI data centers
