Understanding Data Center Demand Growth

Happy to finally share BloombergNEF US Data Center Outlook. This report combines our AI data center primer and US forecast into one incredible deep dive. We left no server rack unchecked – from AI training model demands to project construction timelines – this outlook covers it all. Key Findings: ⚡ BNEF projects US data-center power demand will more than double by 2035, rising from 34.7GW today to 78.2GW. Meanwhile, energy consumption could nearly triple, with average hourly electricity demand jumping from 16.2GWh to 49.1GWh. PJM is expected to remain the biggest by 2035 –followed by Ercot and then the Southeast. ⚡ BNEF’s relatively conservative forecast isn’t downplaying AI – it simply factors in real-world constraints like interconnection delays and build timelines. In the US, a data center takes seven years to reach full operation. For interconnections alone, developers face waits of 2–3 years in Chicago or 7–11 years in parts of Virginia and Texas. ⚡ Four companies – Amazon Web Services (AWS), Google, Meta and Microsoft – currently control 43% of US data-center capacity in 2024, wielding substantial influence over energy infrastructure planning and investment. ⚡ Data-center location decisions hinge many things like power cost, clean power, workforce availability, and tax incentives. But in the age of AI, speed-to-market and scalability top the list. Some developers co-locate near power plants or stranded renewables; others use remote campuses with bridging technologies to accommodate massive AI workloads. Read more here: https://lnkd.in/gAcgP9it Special thanks to Nathalie Limandibhratha (our lead author), along with Tom Rowlands-Rees, Jennifer W., Ben Vickers, and Ashish Sethia, for the many hours and dedication that made this note possible. And to our global counterparts – Jinghong Lyu, Ian Berryman, and David Hostert – it has been a pleasure to hack this data center topic together. What's in the report? ▪️ Section 1: Key findings on growth, AI’s role and hyperscaler influence. ▪️ Section 2: Basics of data-center types, components and efficiency metrics. ▪️ Section 3: How AI training drives massive power needs, cost and design shifts. ▪️ Section 4: Factors influencing where data centers are built. ▪️ Section 5: Regional analysis of major and emerging US data-center markets. ▪️ Section 6: BNEF’s demand and capacity outlook through 2035. Looking to dive deeper into the data?
 The downloadable Excel (included with this report) features: ▪️ All charts & underlying data from the study ▪️ US-wide, project-level data covering every operating data center (April 2025) ▪️ County-level data on pipeline capacity for data centers (April 2025)

Your electricity bill just became the canary in the coal mine for America's biggest infrastructure worry The numbers from this month's energy report aren't just statistics—they're market signals calling for attention. Electricity prices surged 4.5% in May alone. That's nearly double the overall inflation rate. Behind this spike? Data centers have tripled their consumption to 176 terawatt hours in the past decade. Industry projections suggest they could double or triple again within three years. Think about that timeline. We're not talking about gradual shifts over decades. This is explosive demand growth hitting aging infrastructure that was designed for a completely different world. Here's what caught my attention: private companies are now moving into private power generation because the grid simply can't keep up. When Fortune 500s start building their own power plants, that's not innovation—that's admission of system failure. Strategic Reality Check For senior energy leaders: This demand surge represents the biggest grid modernization opportunity since rural electrification. The question isn't whether we'll invest in infrastructure—it's whether clean energy gets the lion's share of that investment or we default back to fossil fuel buildout. For project developers and engineers: Data centers represent concentrated load that's perfect for on-site renewable development. These facilities need 24/7 power, have capital to invest, and increasingly have net-zero commitments. That's your ideal customer profile. For emerging professionals: Understanding the intersection of digital infrastructure and energy systems is becoming table stakes. The companies solving this puzzle will define the next decade of energy markets. What Nobody's Talking About The IEA projects that by 2030, the U.S. will use more electricity processing data than manufacturing aluminum, steel, cement, and chemicals combined. Yet most of our grid planning still assumes demand growth patterns from the 1990s. Smart money is already moving. Utilities that figure out how to partner with hyperscalers on integrated renewable + storage solutions will dominate the next investment cycle. Those that fight distributed generation will lose customers to private power altogether. The grid wasn't designed for this moment. But the infrastructure we build to handle it will define American competitiveness for the next fifty years. Are we treating this AI demand surge as a problem to manage or as the biggest infrastructure investment opportunity of our careers? Because right now, it feels like most of the energy sector is still figuring out that the game has changed. #GridModernization #CleanEnergy #DataCenters #EnergyTransition #Infrastructure

U.S. Power demand enters its fastest growth cycle in 25 Years, and no surprise it is driven by AI Data Centers... The EIA now projects the largest four-year increase in U.S. power consumption since 2000. Demand growth accelerates meaningfully in the second half of the decade, shifting from modest increases to a step change that materially affects generation, transmission, and grid planning requirements. Key Points • Four consecutive years of rising power demand (2024–2027) are forecast, the first such stretch since 2007. • Demand growth is expected to run at 1% year over year in 2026, accelerating to 3% in 2027. • Data centers are the primary driver of incremental load, outweighing residential and traditional commercial demand growth. • AI, cloud, and high-performance computing workloads are creating persistent, high-density baseload demand, not intermittent or seasonal load. All this means that Utilities are already seeing multi-GW load requests tied directly to data center and AI infrastructure pipelines, forcing real-time changes to capital planning, interconnection strategy, and long-term resource development.

The problem with estimating data centre demand is that a lot of guessing goes on. Take the oft quoted estimate that Australia will need around 175 new data centres by 2030 to meet the growth in cloud services and artificial intelligence. This number pops up in developer pipelines, analyst reports, and market commentary. The 175 estimate mostly counts big new campuses, ignoring the hundreds of expansions and retrofits required as AI workloads surge. It misses the distributed edge nodes and regional AI hubs that will bring real-time AI closer to users for low-latency inference. As AI moves from pilot projects to production-scale workloads embedded into every process, a whole ecosystem is emerging: hyperscale training clusters, older sites upgraded with high-density racks, and edge sites linked to local grids — all adding megawatts that never appear in neat “new build” figures. Second, the average energy load per site is fairly outdated. Many forecasts still assume a “medium” data centre might draw 5–10 MW, based on older cloud and enterprise storage workloads. But AI workloads are entirely different. A single AI rack can draw 60–100 kW, not 5–10 kW — and these racks run at near peak, 24/7, for training and constant inference. So-called “medium” sites could easily need 20–50 MW each, while hyperscale AI campuses may exceed 100 MW each. Real-world grid connection requests are already showing that many new sites are targeting these much higher loads. As those doing enterprise strategies know, Enterprise AI take-up is only just beginning. McKinsey and Gartner both highlight that while about 70% of companies now experiment with AI, only about 5% have deployed it at full scale. By 2026, Gartner expects 80% of enterprises will be running GenAI models in production, not just testing them. This means demand will shift from small pilot jobs — easily handled in shared cloud clusters — to always-on, high-density GPU clusters powering daily operations. Every sector from finance and retail to logistics and mining will be adding persistent AI workloads that look more like a new heavy industry than a back-office server room. The math is not hard to do: if you multiply a realistic 20–50 MW per site by a footprint closer to 250–400 AI-ready sites (including expansions, retrofits, and edge), you get a total demand closer to 5–10 GW by 2030 — two or three times what many base-case charts suggest today. Put simply: the AI era means more sites than anyone is counting, more power per site than we’re planning for, and enterprise demand that will hit much faster than traditional forecasts assume. Who wins? Nobody if we don’t start to build. If we don’t adjust our energy strategies and policies now, supply constraints will emerge just as AI becomes central to every globally competitive sector. Lynda Osborne Sin Yin Long Simone Rennie Phillip Vrettakis Nick Turton https://lnkd.in/g7m3bKYu

"Data centers accounted for at least 60% of last year’s increase in U.S. electricity demand, straining an aged grid that struggles to keep up despite investments made in upgrades, latest data and analysis from research firm BloombergNEF and Business Council for Sustainable Energy (BCSE) show. After remaining relatively flat through much of the 2010s, U.S. electricity demand rose 2% compared with the prior year and was up 8% over the past decade. Surging demand for artificial intelligence (AI) services more than any other use is driving the rapid development of power-hungry data centers, a trend the report said is “showing no signs of slowing down.” Electricity demand has ticked up in the last five years from electric vehicles and electrification of some industrial processes, “but the real story is data centers and AI boom,” Trina White, BNEF senior associate for North America’s energy transition team, told reporters Tuesday. Last year, electricity demand from data centers alone grew 21% compared with 2024, 150% in the past five years and 400% over the past decade, White pointed out." AI is reshaping U.S. electricity demand faster than the grid can keep up Daily Energy Insider Amena H. Saiyid https://lnkd.in/gQ7WsJFE

The power market is starting to separate credible demand from speculative backlog.   The cancellation of a 600 MW data center expansion at Stargate in Abilene reflects mounting friction in large load development. While the broader Stargate program remains intact, with plans for up to 4.5 GW of capacity, the decision by OpenAI and ORCL to halt this expansion highlights the growing gap between announced demand and what can realistically be delivered.    Interconnection queues across U.S. power markets have grown well beyond what can be built, and as operators and utilities tighten requirements, projects without clear paths to execution are being forced out.   Enverus Intelligence® Research continues to track this at a granular level. Our latest projections show U.S. data center capacity approaching 50 GW by 2030, driven by hyperscaler capital deployment and accelerating AI adoption. However, the path to that growth is not unconstrained.   Chip supply remains a bottleneck in the near term, and a meaningful share of expected demand, up to 20%, may rely on behind-the-meter solutions with uncertain timing.   What we are seeing is not demand destruction, it's demand refinement. The projects that move forward will be those aligned with power availability, supply chains, and realistic timelines.

The surge in data center demand is now a power infrastructure story as much as a technology one. Behind every new facility is a simple constraint: reliable, scalable power. What’s becoming clearer is that not all suppliers will benefit equally. A few dynamics stand out: • Speed to power is becoming the gating factor Land and capital are available. Grid capacity, interconnection timelines, and equipment lead times are not. • Reliability is non-negotiable Hyperscalers and operators are prioritizing proven solutions and partners that can deliver uptime, not just capacity. • System integration matters more than individual components Transformers, backup generation, switchgear, cooling, and controls all need to work together seamlessly under load. • Execution capability is a differentiator Manufacturing scale, supply chain resilience, field service, and commissioning capacity are becoming as important as product specs. The result: this isn’t just a demand surge. It’s a filtering mechanism. The companies that will benefit are those that can deliver at scale, on time, and with reliability built into the system. Everyone else will find it harder to translate demand into durable growth. #DataCenters #PowerInfrastructure #EnergyTransition #Grid #IndustrialStrategy

Goldman Sachs research projects data center power demand could rise 165% by 2030. That number isn't just an energy problem. It's a competitive problem. Every company investing in AI is also investing in electricity consumption — whether they realize it or not. Compute needs go up. Facility loads go up. And no utility is adding capacity on your timeline. We already know the grid can't build fast enough. New generation sometimes takes a decade. AI demand is compounding in months. But here's the part most people are missing: data centers aren't the only ones feeling this. Every large enterprise is next. The companies that haven't addressed the 20-40% of electricity they're already wasting are about to find out that inefficiency isn't just expensive — it's a constraint on what they can do next. Some companies are building their own supply behind the meter. Smart. But the faster move is treating demand reduction like infrastructure — not a facilities line item. The companies that get efficient now will have enough power to run their AI strategies. Everyone else will pay more or be waiting for a grid that isn't coming on time.

𝐏𝐨𝐰𝐞𝐫𝐢𝐧𝐠 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞: 𝐖𝐡𝐚𝐭 18 𝐦𝐨𝐧𝐭𝐡𝐬 𝐨𝐟 𝐫𝐞𝐜𝐨𝐫𝐝 𝐝𝐚𝐭𝐚 𝐜𝐞𝐧𝐭𝐞𝐫 𝐝𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭 𝐦𝐞𝐚𝐧𝐬 𝐟𝐨𝐫 𝐭𝐡𝐞 𝐔.𝐒. 𝐩𝐨𝐰𝐞𝐫 𝐬𝐲𝐬𝐭𝐞𝐦 Today EPRI released our updated analysis of U.S. data center electricity demand, the follow-up to what was our most-downloaded deliverable. The headline finding: Data centers could consume 9-17% of U.S. electricity by 2030, up from 4-5% today. These projections are roughly 60% higher than our 2024 estimates, driven primarily by record levels of development activity over the past 18 months. A few things I found particularly notable in this update: - The gap between announcements and reality. The report maps the full chain from nominal IT capacity to realized peak demand (accounting for cooling loads, ramp-up lags, and utilization patterns). Announced MW should be treated as a pipeline indicator, not a near-term peak forecast. Our facility-level data shows peak utilization in the range of 62-80% of nameplate capacity. - The state-level variation is extraordinary. Virginia is currently the only state where data centers exceed 20% of electricity demand. By 2030, seven additional states could join that group. Meanwhile, states like Indiana, Louisiana, and Mississippi with very little existing capacity are emerging as new development hotspots as developers prioritize power access and land availability for large AI training facilities. - Supply responses depend entirely on the policy environment. Under current policies, natural gas dominates incremental supply, with build rates in the high scenario more than double the recent five-year average. Under 24/7 carbon-free energy targets, portfolios shift to renewables, storage, and nuclear. Same demand, very different supply story, and a significant departure from our 2024 analysis due to changes in federal tax credits. This work was a genuine team effort (co-authored with Geoffrey Blanford, Tom Wilson, and Nils Johnson) with support from many colleagues across EPRI. It also connects directly to EPRI's DCFlex initiative, which is working with 60+ companies to address grid reliability, flexibility, and affordability as data centers scale. Check out the full report here: https://lnkd.in/gYHm9rn5

🌐 Global Data Center Landscape & India’s Shift Toward a Data-Centric Architecture 🇮🇳 The world operates ~11,800 data centers today, with the US holding the largest installed base. However, the next phase of data center growth is being driven by latency-sensitive workloads, data localization norms, and hyperscale expansion — positioning India as a critical regional data center hub. Why India is becoming data-center centric 🔹 Data localization & compliance (DPDP Act, sectoral regulations) 🔹 Hyperscaler & colocation demand driven by AI/ML, fintech, OTT, and enterprise cloud 🔹 Edge data centers to reduce latency across Tier-2/Tier-3 cities 🔹 Availability of renewable power & grid-scale redundancy Technical shift in design & infrastructure ✔ Higher power density racks (15–30 kW+, AI-ready facilities) ✔ N+1 / 2N electrical architecture with redundant MV/LV distribution ✔ Advanced cooling: liquid cooling, indirect evaporative cooling, hot/cold aisle containment ✔ Uptime-driven design aligned to Tier III / Tier IV philosophies ✔ Integrated BMS / DCIM for energy optimization and predictive maintenance ✔ Fire & life safety engineering compliant with NBC, NFPA 75/76, and TAC guidelines Key focus areas going forward ⚙️ Energy efficiency (low PUE targets) ⚙️ Water stewardship (WUE optimization) ⚙️ Modular & scalable construction ⚙️ Resilience against climate and grid instability 📌 India’s data center evolution is moving beyond capacity addition to engineering-led, resilient, and compliant infrastructure, making it a long-term global data backbone. #DataCenterDesign #DigitalInfrastructure #Hyperscale #IndiaDataCenters #CloudEngineering #AICapacity #CriticalInfrastructure #SustainableDesign