Tech leaders curb AI and cloud costs amid energy crunch
Executives repatriate workloads, adopt FinOps and schedule jobs to lower AI and cloud bills as rising fuel prices and heavy data‑center demand strain power supplies.
Technology firms are changing cloud strategies to manage rising AI and cloud bills. Companies are repatriating workloads to private data centers, adopting FinOps financial controls and shifting when and where compute runs as fuel prices rise and grid capacity tightens in some countries.
IT and cloud teams report sudden spikes in AI spending and growing uncertainty about long‑term costs. Greg Holmes, EMEA field CTO at Apptio, noted many organizations lack clarity on the cost of individual AI operations and said some have “blown through a whole year’s budget of AI in a couple of months.” That trend has driven investment in tools that break costs down to hourly or per‑operation levels.
Energy availability and price swings are changing how firms schedule workloads. Holmes pointed out there are “cheaper periods of the day to use energy” and some suppliers offer lower or even negative prices when supply is high. Companies that can delay nonessential processing are testing systems that shift jobs across time and location to reduce bills.
FinOps practices are being used to increase billing visibility. Tools that parse cloud bills hourly, trigger alerts for usage spikes and expose cost data through APIs allow developers and operations teams to see costs during development and deployment. Holmes said organizations are using those programmatic feeds to make cost information actionable for teams that generate consumption.
Regulators and grid operators have begun to limit where capacity can expand. Denmark paused approvals for new generative AI data centers after requests totaling about 60 gigawatts in a country with roughly 7 gigawatts of peak demand; AI projects accounted for about 14 gigawatts of the requests. In the UK, policymakers have signaled plans to prioritize grid connections for critical data centers and reduce speculative applications.
Enterprises weighing repatriation face tradeoffs. Moving workloads off public clouds can reduce exposure to provider price changes and license shifts, but it shifts costs to capital expenditure, energy bills and facility management. On‑premise AI clusters consume large amounts of power, and rising international oil and gas prices have increased operating expenses for firms running workloads in private facilities.
Cloud storage and other cloud fees are also rising. Organizations report higher storage bills that strain IT budgets and are tightening controls and monitoring consumption. Some are adopting hybrid approaches that place latency‑sensitive or high‑volume tasks on private infrastructure while keeping bursty or unpredictable workloads in public cloud environments.
Finance, engineering and procurement teams are creating shared accountability for cloud spending through formal FinOps programs. Those programs aim to make cost data visible and actionable across teams by using alerts and APIs. Holmes said such measures help teams respond quickly to cost spikes and make choices aligned with budget limits.
Shifts in energy markets and applications for data‑center capacity mean firms are adjusting placement and scheduling strategies for compute. Regulators and grid operators are changing approval and connection policies to reflect limits on power availability, affecting where and how both cloud and on‑premise capacity are planned.
