AI server power demand to surpass conventional systems by 2027
Gartner predicts global data center electricity consumption will reach 565 TWh by 2026, driven significantly by AI workloads. By 2027, AI-optimized servers are expected to consume more power than conventional servers, with implications for data center infrastructure and operations planning. This growth will strain grid supply, necessitating efficiency upgrades and high-efficiency cooling.
Key Takeaways
- AI-optimized servers will account for 31% of data center electricity consumption in 2026, totaling 175TWh.
- Electricity used for cooling and supporting infrastructure is forecast to reach 195TWh in 2026, up from 159TWh in 2025.
- Power consumption by AI-optimized servers is projected to reach 258TWh in 2027, exceeding the 200TWh expected for conventional servers.
- Global data center electricity consumption is forecast to exceed 1,200TWh by 2030, potentially outstripping available grid supply.
Why It Matters
The shift toward AI-optimized hardware is fundamentally altering the data center cost structure, as infrastructure and cooling needs now scale at nearly the same rate as compute. For the streaming ecosystem, this indicates that the long-term cost of AI-driven personalization and encoding will be dictated by power security rather than silicon availability. Supply constraints in primary hubs may force a migration of non-latency-sensitive workloads to secondary markets with better grid access. Watch the adoption rate of liquid cooling and onsite power generation, as Gartner suggests these will become essential to protecting margins by 2027.
Additional Context
The projection of 565TWh for 2026 aligns with broader findings that power availability is the primary bottleneck for AI scaling. In May 2026, Goldman Sachs Research reported that U.S. data center demand alone is forecast to double to 66GW by 2027, up from 31GW in 2025. This surge is creating acute regional pressure; the PJM Interconnection, which serves 65 million people in the U.S., saw capacity market prices spike tenfold for the 2026/2027 period, largely attributed to data center load growth. Per Sightline Climate in April 2026, between 30% and 50% of large-scale data centers scheduled for 2026 opening face delays or cancellations due to these grid constraints and long lead times for electrical equipment like transformers. Institutional analysts are also highlighting the infrastructure investment required to meet this demand. The International Energy Agency (IEA) projected in June 2026 that global data center energy use could reach 945TWh by 2030, with AI tasks being up to 1,000 times more energy-intensive than traditional web queries. To address this, high-density facilities are shifting toward "onsite power" strategies; a June 2026 Bloom Energy report noted that one-third of data centers are expected to use 100% onsite power by 2030 to bypass utility interconnection delays. Simultaneously, tech giants including Amazon and Google increased their combined planned capital expenditure for 2026 by an estimated 75% to secure the energy-ready sites necessary for next-generation AI training and inference.
Read full article at newindianexpress.com
Get this in your inbox → Subscribe
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source