AI data center energy use hit a record high in 2025, with Google reporting a 37% jump in electricity consumption in just one year. The figure, revealed in Google’s 11th annual Environmental Report published on June 30 2026, is a wake-up call for every country hoping to attract hyperscale data centre investment, including Pakistan.
What Google’s 2025 Environmental Report Actually Says
Google’s 11th annual Environmental Report shows electricity consumption rose 37% in 2025, its largest single-year increase ever, pushing total electricity use up more than 250% since 2019.
To put that in plain terms, Google alone is now consuming vastly more power than most medium-sized countries use in a year. Despite the surge, Google says it cut operational emissions 2% year-over-year by matching 100% of its electricity with renewable purchases for a ninth straight year and signing deals for over 12 gigawatts of new clean energy.
But the good news on operations is offset by a harder truth in the supply chain. Supply-chain emissions grew 25% year-over-year, with data center construction alone adding about 2.3 million tons of CO2-equivalent, which Google attributes largely to semiconductor suppliers on carbon-heavy grids in Taiwan, Japan, Vietnam, and India.
Most telling of all is this admission from Google itself: Google’s own report states its “AI infrastructure buildout is currently accelerating faster than the grid is decarbonizing.”
Why AI Data Center Energy Demand Keeps Growing
The reason for this explosion in power use is straightforward. Every time you type a query into an AI tool like Gemini or ChatGPT, dozens of powerful computer chips called GPUs or TPUs (special processors designed for AI tasks) spring into action inside a data centre. These chips run hot and need constant cooling, which adds even more electricity on top of the compute itself.
Cooling systems and environmental control can account for over 30% of electricity in less-efficient enterprise data centers. Even the most efficient hyperscale facilities still burn significant power on cooling alone.
The IEA estimated global data centres consumed around 415 TWh in 2024, about 1.5% of global electricity. That figure is projected to reach around 945 TWh by 2030, nearly doubling in six years. The AI boom is the main driver of that growth.
Microsoft tells a similar story. Microsoft’s 2025 sustainability report noted that, alongside a 168% increase in energy since 2020, the company’s total emissions have grown by 23.4%, citing AI and cloud expansion as key growth-related factors.
The Pakistan Angle: Big Ambition, Bigger Power Challenge
Pakistan has been actively positioning itself as a destination for data centre investment. The numbers look promising at first glance. The Pakistan data center infrastructure market grew from USD 280.53 million in 2020 to USD 644.49 million in 2025 and is anticipated to reach USD 2,178.81 million by 2035.
The Pakistan data center market stands at an installed IT load of 23.53 MW in 2025 and is forecast to reach 53.30 MW by 2030, reflecting a 17.77% CAGR. There is also real momentum on the ground. XDS DataCentre and Al Nahal IT Park partnered in October 2025 to launch Pakistan’s first AI liquid-immersion data center in Karachi, featuring containerised DR solutions to ensure fast access and regulatory-compliant storage.
On the investment side, MeetKai announced a landmark joint venture with the Government of Pakistan to build a national Sovereign AI stack, establishing Pakistan’s first comprehensive national AI platform built on local data, optimised for local languages, and governed under Pakistan’s own laws and institutions.
But here is the problem. The Pakistan data centre market faces several challenges, including inadequate infrastructure, unreliable power supply, limited internet connectivity, and security concerns. These are not small obstacles. A hyperscale AI data centre energy requirement is enormous, and it must be available 24 hours a day, 365 days a year, with zero tolerance for outages.
Imported liquefied natural gas generation costs have risen as high as PKR 28.4 per kWh, eroding the price advantage of on-net hosting. With legacy grid hardware dating back to pre-independence, average transmission and distribution losses remain at 14%.
Operators respond by installing gas turbines and lithium-ion battery arrays, but these investments increase build costs by around 20%, extending payback periods beyond seven years. That kind of financial pressure makes it harder for Pakistan to compete with countries that offer cheap, stable, green power to data centre investors.
Pakistan’s Renewable Card Could Change the Game
There is a genuine silver lining. Pakistan is not just battling power problems; it also sits on a large renewable energy resource that is still mostly untapped for data centres.
With over 60,000 megawatts of renewable energy potential, particularly in Sindh and Balochistan, Pakistan is well-placed to reduce costs and cut carbon emissions by utilising solar panels and engaging in long-term power purchase agreements with wind farms.
There is also the question of surplus capacity. Pakistan generates about 10 GW of surplus electricity, much of which is wasted due to underutilisation. Redirecting even a fraction of that surplus to power-hungry AI data centres could turn a costly liability into a revenue-generating asset.
Surplus electricity earmarked for AI workloads, the rollout of 4G and imminent 5G, and cross-border connectivity delivered by China-Pakistan Economic Corridor (CPEC) fibre routes position the Pakistan data center market for sustained double-digit growth.
Global hyperscalers are also acutely aware of the energy problem. Google signed agreements for more than 12 gigawatts of net-new clean energy in 2025 alone, including restarting Iowa’s 600 MW Duane Arnold nuclear plant with NextEra Energy and a 3 GW hydropower framework with Brookfield. Any country that can offer verifiably clean, cheap, and stable power will have a strong pitch to these companies. Pakistan’s renewable potential is real, but it needs serious investment and policy support to convert that potential into the kind of guaranteed uptime that hyperscalers demand.
What Needs to Happen Next
Industry experts are clear about what Pakistan must do to turn this global AI data center energy wave into local economic gain.
- Grid modernisation. Ageing transmission infrastructure and high line losses must be addressed. Hyperscalers need near-100% uptime; Pakistan’s current grid cannot guarantee that without major upgrades.
- Renewable energy deals. Long-term power purchase agreements (PPAs) tied to solar and wind projects in Sindh and Balochistan could offer data centre operators the green, low-cost energy they are actively seeking globally.
- Clear policy and incentives. Experts stress the need for policy support, infrastructure upgrades, and a push for green data centers to fully unlock Pakistan’s potential in the digital economy.
- Skills development. Pakistan produces 35,000 IT graduates against a stated requirement of 100,000 professionals, creating a talent shortfall evident in delayed commissioning schedules. This gap must narrow.
You can read Google’s full environmental commitments on sustainability.google. For global data on AI energy trends, the International Energy Agency (IEA) publishes detailed, regularly updated reports on data centre electricity consumption worldwide.
Frequently Asked Questions
How much energy did Google’s AI data centers use in 2025?
Google’s electricity consumption rose 37% in 2025, its largest single-year increase ever, pushing total electricity use up more than 250% since 2019. The company matched all of this with renewable energy purchases.
Why does AI use so much electricity?
AI models require thousands of specialised chips running at full speed to process requests. These chips generate enormous heat, so data centres also need powerful cooling systems running around the clock. Available power is not keeping up with demand, creating a major bottleneck for cloud, AI and data centre proliferation.
Does Pakistan have the power supply needed to host AI data centers?
Not yet at hyperscale. Challenges such as limited electricity supply and high costs of internet connectivity may hinder the market’s growth. However, Pakistan’s large renewable energy reserves and surplus generation capacity offer a realistic path forward if the right policies and grid investments are made.
What is the global outlook for AI data center energy by 2030?
Data centre electricity consumption is projected to reach around 945 TWh by 2030, under 3% of global electricity consumption in the IEA base case. That is more than double the 2024 figure and a sign that the energy challenge will only intensify as AI adoption keeps accelerating.
