Volta presents its Norwegian project with more than 130 megawatts of gross power on its commercial website . Bitdeer described 121 megawatts of IT load leased to a Volta subsidiary in Tydal in its update published August 26 . The figures measure different things; they do not calculate a performance difference.
For a compute buyer, that distinction opens the relevant conversation: which dedicated resources will be delivered, under which commitment and with which service responsibilities. Project power explains physical scale. The commercial decision needs to reach the work the customer will be able to execute.
Two measures answer different questions
| Published figure | What it describes | Source |
|---|---|---|
| More than 130 gross MW | Norwegian project power; it does not by itself measure available compute | Volta, website consulted October 13, 2026 |
| 121 MW IT load | Capacity in the lease described for a Volta subsidiary | Bitdeer, July report published August 26, 2026 |
Collapse notes
These figures describe announcements and an infrastructure transaction; they do not verify how much compute capacity is available to a buyer today. Nor are they comparable measures of AI model productivity.
A gross figure includes the energy context supporting the facility. An IT-load figure brings the explanation closer to equipment performing computation. Offer evaluation still needs relevant conditions: cluster configuration, access and the contracted availability period.
The electricity scenario more than doubles data-center demand
Estimated global data-center electricity consumption. The 2030 bar represents the IEA Base Case published in April 2025.
TWh
Tap or hover for details. All values are also available in the data table.
Collapse notes
Historical estimates and a Base Case; 2030 is not an observed result. Includes conventional and AI data centers; excludes cryptocurrency and data transmission networks. TWh measures annual energy and MW measures power, so these figures cannot be directly compared with Volta’s announced 130 gross MW or 121 MW of IT load.
Explore the data table
| Category | Electricity consumption |
|---|---|
| 2020 | 269 TWh |
| 2023 | 361 TWh |
| 2024 | 416 TWh |
| 2030 (Base Case) | 946 TWh |
The IEA series compiled by Statista puts global consumption at 416 TWh in 2024 and projects 946 TWh for 2030 in its April 2025 Base Case. This is sector context: contracted delivery needs to specify availability and electricity connections alongside its power figure.
Duration changes the commercial conversation
Volta’s AI Factories offer targets AI labs and companies through dedicated hardware and long-duration commitments. That purchase differs from acquiring shared resources for a temporary need. The customer must connect the commitment with the work it expects to sustain.
A company preparing extended training needs to coordinate capacity with data, people and project timing. Contracting too early commits spending before capacity can be used. Arriving late can delay the team’s work. Delivery dates and the start of the economic obligation belong in the sales argument.
Inference gains weight in operators’ expected electricity mix
Reported and expected electricity-consumption distribution among executives at organizations owning or operating data centers. Capgemini survey, January 2026.
Tap or hover for details. Select a series above to show or hide it.
Collapse notes
Worldwide online survey. Current mix: 77 executives at operators already hosting AI; expectations: 140 executives at operators hosting AI or planning to within 12 months. These are different cohorts, not a longitudinal panel. Expectations do not measure future consumption or describe Volta’s workload mix.
Explore the data table
| Category | Reported current mix | Expected in 3–5 years |
|---|---|---|
| AI training | 15% | 25% |
| AI inference | 10% | 35% |
| Other IT workloads | 75% | 40% |
Capgemini separates training from inference: the latter moves from 10% of reported consumption to an expected 35% in different survey cohorts. That distinction directs a commercial question for Volta: which work the buyer needs to sustain, and on which timetable, rather than treating every AI purchase alike.
The proposition also needs to explain responsibility across infrastructure layers. Power, cooling, connectivity and operations affect the service the buyer experiences. Combining them in one offer has commercial value when it simplifies responsibility without hiding delivery conditions.
A suitable account can complete that decision
Acquisition needs to distinguish buyers requiring dedicated capacity from those exploring options. Announced project size can attract both. A useful conversation identifies planned work, its duration and the preparation required to use the service.
That information lets sales present the configuration and commitment relevant to the account. It also permits an incompatible purchase to be rejected before both sides spend time negotiating. Commercial progress is measured by agreements a customer can use, alongside interest in project scale.
The promise of compute as infrastructure becomes persuasive when it translates into precise delivery. The megawatt opens the explanation; the contract needs to complete it with identifiable capacity, a timetable and responsibility for the service the buyer expects.
In this article
Marian Vasile
Marian Vasile is a MarTech and growth professional. His work spans product marketing, SEO, paid advertising and deep online tracking, including behavioral analytics and attribution.
