AI data centers: what they are, how they work and what they cost
An AI data center is a facility built to train and run AI models: racks of GPU servers drawing 100 kW or more each, cooled by liquid, on a site chosen for its power supply. This guide explains how one differs from a traditional data center, what is inside, where the power comes from, what it costs per megawatt and per gigawatt, which campuses are the largest, and why so many Bitcoin miners are now AI landlords.
Updated November 13, 2026. Every figure carries its source and date; see the list at the end.
What an AI data center is
An AI data center is a data center built to train and run artificial intelligence models. Where a conventional facility hosts websites, business software and storage on ordinary servers, an AI data center is filled with GPU servers that work together as one very large computer. Those servers use so much electricity, and give off so much heat, that the building around them has to be designed differently from the ground up.
NVIDIA calls these buildings AI factories: electricity and data go in, and the output of AI models, counted in tokens, comes out. The name fits the scale. The largest campuses now under construction are measured in gigawatts, the output of a large power station, where a big conventional data center is measured in tens of megawatts.
The numbers that define one, each sourced below:
- About 120 kW per rack for NVIDIA's GB200 NVL72, against 10 to 30 kW for a conventional rack.
- About $11.3 million per megawatt to build in 2026, before the chips (JLL).
- More than four years to get a grid connection in primary US markets (JLL).
- About 415 TWh of electricity used by the world's data centers in 2024, heading for about 945 TWh by 2030 (IEA).
AI data center vs. traditional data center
| Traditional data center | AI data center | |
|---|---|---|
| Main job | Websites, business software, storage | Training and running AI models |
| Chips | CPUs | GPUs and other accelerators |
| Power per rack | 10 to 30 kW | About 120 kW today; about 600 kW planned for 2027 |
| Cooling | Air | Liquid to the chip, or immersion |
| Network | Standard Ethernet | A high-speed fabric linking thousands of GPUs into one cluster |
| Typical size | Tens of megawatts | Hundreds of megawatts to gigawatts |
| Where it goes | Near users and fiber | Where power is available |
Density. A conventional enterprise rack has run at 10 to 30 kW for years, and the Uptime Institute's 2025 survey found few facilities above 30 kW. An NVIDIA GB200 NVL72 rack, the workhorse of 2025 and 2026 AI training clusters, draws about 120 kW. NVIDIA's roadmap goes further: Vera Rubin systems ship in the second half of 2026 at similar rack power, and the company has said its 2027 Rubin Ultra design will reach roughly 600 kW per rack, with power delivered over 800-volt DC.
Power and operating cost per system, from a single GPU to a full server, are listed on aPanel: for example the NVIDIA H200 and the NVIDIA DGX B200.
Everything else follows from that number. A building designed for 10 kW racks cannot simply be refilled with 120 kW racks: the power feeds, the switchgear, the floor loading and, above all, the cooling all have to change.
Training and inference also pull in different directions. Training clusters tolerate distance from users and go where power is cheap and available, which is why the largest campuses are rising in West Texas and other frontier markets. Inference, serving answers to users, is moving closer to population centers as it grows.
Inside an AI data center
Four systems make up the building. Each one is sized by the same number, the power the racks draw.
- The power train
- Electricity arrives from the grid at high voltage and is stepped down at a substation on the site. Switchgear and transformers carry it into the building, uninterruptible power supplies and batteries bridge any gap until backup generators start, and busways deliver it to each rack, where it is converted to the direct current the servers use.
- Compute
- Racks of GPU servers. One NVIDIA GB200 NVL72 rack holds 72 GPUs and is built, shipped and cooled as a single unit. A training cluster is thousands of these racks.
- The network
- A back-end network, InfiniBand or high-speed Ethernet over fiber, links every GPU so the cluster can train one model as a single machine. A separate front-end network connects the cluster to storage and to its users.
- The cooling loop
- Cold plates on the chips pass heat to a coolant. Coolant distribution units move that heat into the building's water loop, and dry coolers, chillers or cooling towers outside release it. See the liquid cooling guide for how each part is chosen.
Storage for training data, fire protection, security and the software that monitors every breaker and pump complete the picture. The order in which these arrive on site sets the schedule: the grid connection and the large transformers take longest, which is why a site that already has them is worth so much. For the power side in depth, read the data center power guide.
Power: how much, and how to get it
The International Energy Agency estimates that data centers used about 415 TWh in 2024, roughly 1.5% of the world's electricity, and projects about 945 TWh by 2030. The United States accounted for 45% of the 2024 total. In April 2026 the IEA reported that data center demand grew 17% in 2025, with AI-focused sites growing faster.
US forecasts vary with their assumptions. Lawrence Berkeley National Laboratory, in its report for the Department of Energy, put US data centers at 176 TWh in 2023, or 4.4% of the country's electricity, rising to between 6.7% and 12% by 2028. EPRI's 2026 study projects between 9% and 17% by 2030.
The queue is the constraint
Money and land are rarely what holds a project back now. The grid is. Berkeley Lab counted about 2,061 GW of generation and storage waiting in US interconnection queues at the end of 2025, with a median of about five years from request to operation. On the demand side, JLL reports average waits of more than four years for a grid connection in primary data center markets.
Texas shows the scale. ERCOT's large-load queue reached 474.7 GW in June 2026, 90% of it data centers, more than five times the grid's record peak of 91 GW set that July. Texas Senate Bill 6, signed in 2025, now requires loads of 75 MW or more to disclose backup generation and allows new large loads to be curtailed in emergencies, and in August 2026 the governor ordered an audit of the data center queue.
Bringing your own power
Operators are answering with power of their own. The Stargate campus in Abilene, Texas, connected to ERCOT but also built about 360 MW of on-site gas turbines, with plans for more. Hyperscalers have turned to nuclear: Microsoft's 20-year agreement to restart Three Mile Island Unit 1 (about 835 MW), Meta's 20-year contract for the full 1.1 GW output of the Clinton plant in Illinois, and Amazon's 1.92 GW arrangement with Talen Energy at Susquehanna. The IEA counts 45 GW of conditional agreements between data centers and small modular reactor developers.
Cooling past the limit of air
Air cooling struggles somewhere around 100 to 120 kW per rack. Liquid carries far more heat per unit volume, so AI racks are cooled by liquid that reaches the chips directly. Cooling has historically consumed a large share of a data center's electricity, which makes the choice of system a cost decision as much as an engineering one.
- Rear-door heat exchangers
- A liquid-cooled coil on the back of the rack. Industry estimates put the practical range at roughly 20 to 50 kW per rack, so they suit mixed halls rather than dense AI clusters.
- Direct-to-chip liquid cooling
- Cold plates on the GPUs and CPUs carry heat to a coolant loop. This is the standard for NVIDIA's GB200 and later systems; industry estimates run from about 50 kW to well over 200 kW per rack.
- Single-phase immersion
- Servers sit in tanks of non-conductive fluid that is pumped through a heat exchanger. Widely used in Bitcoin mining, and increasingly tested for AI.
- Two-phase immersion
- The fluid boils on the chip and condenses on a coil above the tank. It removes heat very efficiently but needs specialised fluids and sealed tanks.
Efficiency is usually measured as power usage effectiveness (PUE): total facility energy divided by IT energy. The Uptime Institute reports an industry average near 1.56 that has barely moved in six years, while liquid-cooled AI designs target figures much closer to 1.1. Water is the other side of the ledger: Berkeley Lab estimated that US data centers used about 17 billion gallons directly in 2023, a figure that could double or quadruple by 2028. Closed-loop liquid systems use far less water than evaporative air cooling. For an operator's view of the problem, read designing for the AI cooling crisis and Bitmain's liquid cooling for AI data centers.
Sites, costs and markets
JLL puts the 2026 global average construction cost at about $11.3 million per megawatt, up from $7.7 million in 2020, with premium markets above $15 million. Cushman & Wakefield reports costs per megawatt up 21% since late 2024. Long-lead equipment sets the pace: large transformers can take about three years to arrive. Liquid cooling adds roughly $1 million to $2 million per megawatt over an air-cooled build.
Demand is outrunning supply. CBRE's mid-2026 count of eight primary North American markets showed 7.5 GW under construction, more than 80% of it preleased, and vacancy of 1.4%. JLL, counting every North American market, put more than 66 GW under construction and vacancy near 1% for a third year, with 77% of new construction in frontier markets led by West Texas. Many tenants are now signing for capacity delivered in 2028.
- Northern Virginia
- About 4,500 MW of inventory, 0.2% vacancy and 2,420 MW under construction (CBRE, H1 2026).
- Atlanta
- About 1,800 MW of inventory and roughly 2,900 MW under construction, the most in North America (CBRE).
- Dallas
- About 1,430 MW of inventory and more than 765 MW under construction (CBRE).
- Phoenix
- About 1,070 MW of inventory (CBRE).
- Chicago
- About 910 MW of inventory (CBRE).
What an AI data center costs
There are two prices, and they differ by a factor of four or more. The first is the building: land, shell, power and cooling. At JLL's 2026 average of about $11.3 million per megawatt, a one-gigawatt campus works out to roughly $11 billion before a single server is installed.
The second is the building with its computers. NVIDIA's chief executive has put a one-gigawatt AI factory at roughly $50 billion to build, as reported in September 2026. The same report gives NVIDIA's own share per gigawatt as about $18 billion for its Hopper generation, about $25 billion for Grace Blackwell and about $40 billion for Vera Rubin. The chips, in other words, now cost several times the building that holds them, and their share is rising with each generation.
That changes how projects are financed and how fast they must fill. A building can be paid off over decades. A new GPU generation arrives about every year, so the computers age far faster than the walls around them, and an owner needs paying tenants the moment the power is on. It is one reason most new capacity is leased before it is finished, as the market figures above show.
The largest AI data center campuses
The biggest sites are announced in gigawatts and built in phases. Planned capacity and what is running today are very different numbers, so both are shown where a source gives them.
| Campus | Location | Planned | Status |
|---|---|---|---|
| Meta Hyperion | Richland Parish, Louisiana | 2 GW by 2030, scaling to 5 GW | Announced July 2025 |
| OpenAI Stargate (Project Jupiter) | Doña Ana County, New Mexico | 2.2 GW | Foundation work in April 2026; target late 2028 |
| OpenAI Stargate | Shackelford County, Texas | 2.0 GW | Under construction; first building due late 2026 |
| OpenAI Stargate | Saline Township, Michigan | 1.4 GW | Foundation work in April 2026; target late 2028 |
| OpenAI Stargate | Port Washington, Wisconsin | 1.3 GW | Ground broken; target late 2028 |
| OpenAI Stargate | Abilene, Texas | 1.2 GW | About 0.3 GW running in April 2026, four of eight buildings |
| OpenAI Stargate | Milam County, Texas | 1.2 GW | First building rising in April 2026; target late 2028 |
| Meta Prometheus | New Albany, Ohio | 1 GW | Due online in 2026 |
| Amazon Project Rainier | New Carlisle, Indiana | About 1 GW of IT power | Operating and ramping in mid-2026 |
Stargate figures are Epoch AI's estimates of total facility power, published April 17, 2026; together the seven US Stargate sites come to more than 9 GW planned. Meta's figures are from the company's July 2025 announcement. The Rainier figure is from a June 2026 compilation of Epoch AI's tracking.
Timing varies as much as size. Abilene, built by Crusoe with natural gas generation on site as well as a grid connection, had four of its eight buildings running in April 2026 and is due to be complete at the end of 2026. The other large Stargate sites are not expected to be complete before the end of 2028. Three of the seven are in Texas, where the AI build-out around Dallas is the subject of the last section.
Why Bitcoin miners are becoming AI landlords
A Bitcoin mine is, physically, most of a data center: an energized grid connection, substations, land, cooling and crews used to running thousands of machines around the clock. What it lacks is the redundancy and fit-out that AI tenants require, and that can be added far faster than a new grid connection can be won. JPMorgan analysts estimated in 2024 that getting power three to four years sooner is worth $105 million to $140 million per megawatt to a hyperscaler. The deals that followed show how much of the market agreed:
- Core Scientific contracted about 590 MW across six sites to CoreWeave, worth about $10.2 billion in projected revenue over 12-year terms. CoreWeave's bid to buy the company outright was rejected by shareholders and ended in October 2025. Read how Core Scientific's operating system began.
- IREN signed a $9.7 billion, five-year contract with Microsoft for NVIDIA GB300 capacity at Childress, Texas, and delivered the first 50 MW phase in August 2026.
- Hut 8 leased 245 MW at River Bend, Louisiana, to Fluidstack for Anthropic, with Google backstopping the lease: about $7 billion over 15 years, and up to $17.7 billion if every option is taken.
- Cipher Mining signed about $5.5 billion over 15 years with AWS and about $3 billion over 10 years with Fluidstack, backed by Google.
- TeraWulf signed Fluidstack deals at Lake Mariner, New York, and Abernathy, Texas, with Google backstops and a Google equity stake.
- Galaxy contracted 526 MW of critical IT load at its Helios campus in West Texas to CoreWeave and delivered the first 133 MW in July 2026.
- HIVE runs GPUs alongside Bitcoin mining and is building a large AI site near Toronto. Read HIVE's dual-engine strategy.
Not every miner is converting. Most still mine, and many split sites between Bitcoin and AI. But the direction is clear enough that this conversion is the founding idea of Intelligence Disrupt: the people who own the power and the people who need it, on one floor.
Two of these stories were told on our own stage. Read how Galaxy rebuilt Helios and weighed AI against Bitcoin and what an insurer sees when AI moves in next to a Bitcoin mine.
Who is renting the capacity
Beyond the hyperscalers, a group of GPU clouds, often called neoclouds, rents capacity and resells compute. The largest, CoreWeave, reported second-quarter 2026 revenue of about $2.6 billion, a backlog near $104 billion, 1.5 GW of active power and about 3.7 GW contracted. Fluidstack appears as the tenant in several of the miner deals above, with Google standing behind the lease payments, a structure that lets a younger company sign long leases on a giant's credit.
Colocation rents reflect the shortage. JLL's figures put premium-market rents at roughly $140 to $160 per kW per month. For operators deciding between owning, leasing and renting GPU capacity, the trade-off is speed against control, and in 2026 speed is usually winning.
Dallas
The Dallas area is one of the largest and fastest-growing data center markets in the country. CBRE counted more than 1,400 MW of inventory in mid-2026, more than 765 MW under construction with 95% preleased, vacancy of 2.8% and another 3.7 GW planned. Oncor, the area's utility, reported 737 active large-load requests at the end of June 2026, about 282 GW of them from data centers, and plans more than $7 billion of transmission work along the southern DFW and I-35 corridor through 2034.
The draw is cost and speed: land that sells for a fraction of Northern Virginia prices, no state income tax, sales-tax exemptions on data center equipment and local property-tax abatements. The constraint, here as everywhere, is power. That is the subject of Intelligence Disrupt 2027, March 22–24 at the Irving Convention Center at Las Colinas, in the middle of that market. See passes or exhibit.
Questions
- What is an AI data center?
- A data center built to train and run AI models. It holds racks of GPU servers that draw 100 kW or more each, is fed with far more power per square foot than a conventional facility, and is cooled by liquid instead of air.
- How is an AI data center different from a regular data center?
- Density. A conventional rack runs at 10 to 30 kW and is cooled by air; an NVIDIA GB200 NVL72 rack draws about 120 kW and is cooled by liquid. AI sites are also far larger, hundreds of megawatts to gigawatts, and are placed where power is available instead of where users are.
- What is the largest AI data center?
- Among announced campuses, Meta's Hyperion in Louisiana is planned to scale to 5 GW. Epoch AI estimates OpenAI's Stargate sites in Doña Ana County, New Mexico, at 2.2 GW and Shackelford County, Texas, at 2.0 GW. Stargate's first campus, in Abilene, Texas, is planned at 1.2 GW and had about 0.3 GW running in April 2026.
- How much does a 1 GW AI data center cost?
- Roughly $50 billion with the chips, by the figure NVIDIA's chief executive has used, as reported in September 2026. The building, power and cooling come to about $11 billion of that at JLL's 2026 average of $11.3 million per megawatt.
- What is an AI factory?
- NVIDIA's name for an AI data center: a facility that takes in electricity and data and turns out AI model output, measured in tokens, the way a factory turns out goods.
- Do AI data centers use a lot of water?
- It depends on the cooling. Berkeley Lab estimated that US data centers used about 17 billion gallons directly in 2023 and that the figure could double or quadruple by 2028. Closed-loop liquid cooling uses far less water than evaporative air cooling.
- How much power does an AI data center use?
- The IEA puts a typical AI-focused data center at the electricity use of about 100,000 households, and the largest ones under construction at roughly 20 times that. A single NVIDIA GB200 NVL72 rack draws about 120 kW.
- How much electricity do US data centers use?
- About 176 TWh in 2023, or 4.4% of US electricity, according to Lawrence Berkeley National Laboratory, which projects 6.7% to 12% by 2028. EPRI projects 9% to 17% by 2030.
- What is liquid cooling in a data center?
- Coolant is brought to cold plates on the chips (direct-to-chip) or servers are submerged in a non-conductive fluid (immersion). Liquid carries far more heat than air, which is why racks above roughly 100 kW are liquid cooled.
- How much does it cost to build an AI data center?
- JLL puts the 2026 global average at about $11.3 million per megawatt, up from $7.7 million in 2020, with premium markets above $15 million. Liquid cooling adds roughly $1 million to $2 million per megawatt.
- How long does it take to get grid power for a data center?
- In primary US markets, more than four years on average, according to JLL. Newer markets can deliver power 12 to 24 months sooner, which is a large part of why West Texas and other frontier markets are growing.
- Why are Bitcoin miners becoming AI data centers?
- They already hold energized, interconnected sites. Hyperscalers and GPU clouds are paying for that head start: IREN, Hut 8, Cipher, TeraWulf, Galaxy and others have signed multi-billion-dollar AI hosting and lease contracts since 2024.
- What is PUE?
- Power usage effectiveness: total facility energy divided by the energy used by IT equipment. The industry average has held near 1.56 for several years, per the Uptime Institute. Liquid-cooled designs can run well below that.
- Why is Dallas a top data center market?
- CBRE counts more than 1,400 MW of Dallas-area inventory with 765 MW under construction, 95% preleased, at 2.8% vacancy in mid-2026. Land costs a fraction of Northern Virginia's, Texas has no state income tax, and ERCOT offers a large, if crowded, grid.
More guides
- GPU cloud: providers, prices and the power behind them
- Liquid cooling for data centers
- Data center power: how much it takes and how to get it
- Bitcoin miners turning into AI and HPC data centers
- Texas data centers: Dallas-Fort Worth and the AI campuses
- Stargate: every OpenAI data center site and where it stands
- NVIDIA GPU power and cooling requirements
- The largest data centers in the world
- How much does it cost to build a data center?
- PUE and how much power data centers use
Sources
- IEA, Energy and AI: executive summary · 2025
- IEA, Data centre electricity use surged in 2025 · 2026-04-16
- US DOE on the LBNL data center energy report · 2024-12-20
- EPRI, Powering Intelligence 2026 · 2026-02-26
- Berkeley Lab, Queued Up 2026 · 2026-06
- NVIDIA, Blackwell platform and liquid cooling · 2025-04-22
- Uptime Institute, Global Data Center Survey 2025 · 2025-07-29
- POWER Magazine, Texas data center interconnection queue audit · 2026-08-04
- ERCOT, Senate testimony on the grid · 2026-07-29
- CBRE, North American data center trends H1 2026 · 2026-08-27
- JLL, North America data center report, midyear 2026 · 2026-08-11
- Datacentres.com on JLL construction costs · 2026-05-12
- Dallas Innovates on CBRE DFW data · 2026-09-03
- Oncor, second-quarter 2026 results · 2026-08-06
- Core Scientific and CoreWeave partnership · 2025-02-26
- Core Scientific ends CoreWeave merger agreement · 2025-10-30
- The Block, IREN delivers first Microsoft AI cloud capacity · 2026-08-17
- DCD, Hut 8 and Fluidstack 245 MW deal · 2025-12
- The Energy Mag, TeraWulf and Fluidstack in Texas · 2025-10-28
- The Block, Galaxy delivers Helios phase I to CoreWeave · 2026-07-07
- CoreWeave, second-quarter 2026 results · 2026-08-11
- DCD, Meta and Constellation Clinton PPA · 2025-06-03
- Talen Energy and Amazon, Form 8-K · 2025-06-11
- Epoch AI, OpenAI Stargate: where the US sites stand · 2026-04-17
- TechCrunch, Meta's Hyperion and Prometheus data centers · 2025-07-14
- 24/7 Wall St., NVIDIA's chief executive on the cost of a 1-gigawatt AI factory · 2026-09-24
- NextBigFuture, major AI data center build projects (compiled from Epoch AI tracking) · 2026-06-25
The power is on this floor. Be in the room.
March 22–24, 2027 · Irving Convention Center · Dallas
Intelligence Disrupt and Mining Disrupt are NOT affiliated with any other brands, groups, events, meetups, individuals, sponsors, exhibitors, agents, or referrals, except the Industry Ambassadors listed on this site. We do not sell any lists. Beware of scams claiming otherwise.
There are imitators and scammers using our name to solicit sponsors and attendees. Always verify directly through our official Telegram group. Stay vigilant!
Intelligence Disrupt Policy: No refunds. No outside food or drinks. No third-party event solicitations.
Unauthorized solicitation, sales, advertising, product demonstrations, or distribution of promotional materials, including magazines, brochures, flyers, catalogs, business literature, and promotional giveaways, is strictly prohibited anywhere within the event venue unless expressly authorized by Intelligence Disrupt.
Marketing activities are reserved exclusively for approved exhibitors, sponsors, and authorized event partners. Attendees found engaging in unauthorized promotional activities may be required to cease immediately, surrender promotional materials, and may have their event credentials revoked without refund at the sole discretion of event management.

