U.S. data centers consumed 183 terawatt-hours of electricity in 2024, a figure driven significantly by the surge in AI. This massive power consumption problem is forcing companies like Okta to expand renewable energy purchases.
AI adoption is skyrocketing due to its perceived benefits, but its underlying energy consumption rapidly escalates. This poses a significant environmental challenge.
As AI integration deepens across industries, companies will increasingly face pressure to offset their energy usage. Sustainable AI practices become a critical component of corporate responsibility and operational strategy.
The Escalating Energy Demands of Data Centers
- Electricity consumption from data centres is estimated to amount to around 415 terawatt hours (TWh), according to the IEA.
- Data centres account for approximately 1.5% of global electricity consumption, according to the IEA (data from before 2025).
These figures aren't just statistics; they reveal data centers, the very backbone of our AI future, are already guzzling a significant chunk of global power. Ignoring their sustainability is no longer an option.
Okta's Proactive Stance on AI Sustainability
Okta adopted a sustainable strategy in April to manage the artificial intelligence software used by its employees, according to Trellis Group. The company expanded its purchases of renewable energy certificates.
These certificates cover the additional energy needed to run the top two third-party AI tools used by the company. Okta's move isn't just good PR; it's a blueprint. By directly offsetting the energy demands of their AI tools, they're showing other companies how to put their money where their mouth is. The question is, who's next?
A Rapidly Growing Footprint: US Data Center Consumption
The U.S. isn't just participating in the AI boom; it's leading the charge in energy consumption. At 183 TWh in 2024, according to Pew Research, we're gobbling over 44% of the global data center electricity. That's not just a statistic; it's a glaring red flag.
This isn't a slow burn. Data center energy use nearly tripled from 76 TWh in 2018 to an estimated 176 TWh in 2023, say Belfer Center and Pew Research. This surge isn't coincidental; it's the direct consequence of our AI obsession. We need efficiency and sustainable power, yesterday.
The Future Landscape of AI Infrastructure
With over 4,000 data centers already running or planned in the U.S. Pew Research confirms one thing: AI's energy appetite is only going to swell. This isn't just growth; it's an impending energy crisis.
We need more than just good intentions. Without concrete offsetting strategies, the tech industry isn't just risking its reputation; it's threatening to overwhelm our grids and derail climate targets. This isn't a hypothetical; it's a looming reality.
The Unavoidable Truths of AI's Footprint
Forget a diffuse problem. A third of U.S. data centers are crammed into Virginia (643), Texas (395), and California (319), Pew Research tells us. This isn't just a geographic detail; it means these regions will bear the brunt of AI's energy and infrastructure headaches. Are they ready?
The real environmental villain of AI is its insatiable energy demand, churning out carbon from power plants. Consider this: training one major AI model can spew as much carbon as five cars over their entire lifespan, according to some estimates. That's not a minor footprint; it's a wake-up call for immediate renewable energy adoption.
Can we curb AI's power hunger? Absolutely. We're talking smarter algorithms, specialized energy-efficient chips, and radical cooling systems like immersion technology. This isn't just about 'nice-to-haves'; by 2026, these aren't options, they're mandates for any tech company serious about managing this escalating demand (data from before 2025).
If current trends persist, the unchecked expansion of AI's energy demands will likely force a reckoning with grid capacity and climate goals far sooner than many anticipate.
