Capacity and pipeline

Reading order

Read the metric type before the number. IT power, facility power, site power, utility power and announced capacity with an unclear definition use the same megawatt unit and are not the same fact. Then read the scope: building, phase, campus or region. Then read the three dates: when the source itself is dated, the as-of time the claim describes, and when a human last checked the row. Then read the evidence state. Only after those four labels should you ask whether two rows could ever be added.

Keep campus totals off their own phases. A parent campus figure already contains, or looks forward to, the buildings underneath it. Adding them produces a double count even when both sentences are true. Keep current commissioned wording off future “once operational” wording. Keep a marketing available-power envelope off a balance-sheet sentence. If those tests fail, the rows stay adjacent and unsummed.

After the labelled rows, open the project pages. The capacity table is an index, not a substitute for original wording. Falun, STACK and Conapto each have notes that the definition of IT versus facility power is incomplete. Those notes travel with the number. A spreadsheet that drops the note has already changed the claim. Quote the sentence, not only the integer.

• Same metric type.
• Same scope, and no phase plus its parent campus.
• Same as-of window, not a future sentence plus a balance-sheet date.
• Show how many rows were excluded as unknown.

Why Swedish capacity will not collapse to one number

Swedish connection capacity is not a published national leftover. Svenska kraftnät’s indicative transmission-network capacity map is a national layer. The authority’s own framing is that it does not replace the formal connection process and does not show underlying-network capacity. Regional and local network owners answer for the property. Those answers are case files, often unpublished. A market page therefore cannot honestly print “Sweden has X spare MW for data centers.”

Bidding areas SE1 Luleå, SE2 Sundsvall, SE3 Stockholm and SE4 Malmö show transmission constraints in the wholesale price. They do not allocate megawatts to Falun, Staffanstorp, Sollefteå or Kista. The May 2026 government assignment, with a Svenska kraftnät report due 29 January 2027, is a review of area design. It does not refresh campus rows and does not create a national load census. National energy mix — hydropower described on a Meta page, renewable-energy wording on an operator brochure — is still not a connection offer.

Operator megawatts themselves mix measurement bases. Designed, announced, contracted and measured are different claims. A press sentence that power is “secured” is not filed here as a reviewed grid agreement. EcoDataCenter’s 80 MW press item names Falu Energi & Vatten and Ellevio in operator wording; the register stores that as reported, not as a reviewed contract. STACK’s 18 MW and Conapto’s 24 MW are likewise operator-stated campus figures without the underlying agreement documents.

What named primary sources show

EcoDC Holding’s 2025 annual report, in the English extract used here, says data centers C and D were fully completed and commissioned and that total capacity at the Falun site increased from 29 MW to 41 MW as of the balance-sheet date. The same extract says construction of E and F began in 2025, with company targets for later 2026 operation, after which Falun would amount to 65 MW. Those E and F dates are expectations, not commissioning certificates. The 41 MW and 65 MW sentences stay as two rows.

The Falun marketing page still states 80 MW total available power. A related press item describes securing 80 MW future power supply with Falu Energi & Vatten and Ellevio. That wording is operator-reported and is not filed here as a reviewed grid agreement. The 80 MW envelope is a third row. The three figures — 41, 65 and 80 — are not added together.

STACK Infrastructure’s Stockholm market page states 18 MW and one campus. The page does not define whether 18 MW is IT load or facility load. The row is stored as announced capacity with an unclear definition. Agreement documents are not published here. That 18 MW stays a sourced, non-additive Stockholm row beside Conapto, not a brick in a national wall.

Conapto’s 5 September 2024 release said Stockholm 4 South had opened and that the campus offered 5,200 m² of computer rooms and 24 MW of power. A data sheet for Stockholm 2 South states five computer rooms with 4 MW total capacity. The 4 MW phase figure is not added on top of the 24 MW campus figure. IT versus facility definition for the 24 MW is not fully specified.

atNorth’s SWE01 page states 12 MW site power capacity for the Kista metro hall. Occupied IT load is not published. SWE02 Akalla is a 30 MW metro site announced on 29 January 2026; SWE04 Sollefteå is a 300 MW plan announced on 23 February 2026 on a 50-hectare plot. Those figures are announced or planned site power, not live IT load, and they are not added to SWE01. GleSYS Falkenberg has two dated rows: 3 MW utility power in the 1 October 2020 heat-delivery release to Falkenberg Energi, and 5 MW total capacity on the later operator page. Whether 5 MW superseded 3 MW through expansion is not explained on those pages. They are not added to 8 MW.

Most hyperscale and several colocation rows disclose no megawatts. Meta’s cited page has no current operator MW. A 2011 Associated Press sentence about a 120 MW energy need is a historical announced claim, not current operating capacity. Microsoft’s 2021 region materials name Gävle, Sandviken and Staffanstorp without facility MW. AWS Europe (Stockholm) has no MW on the cited launch pages. Digital Realty states 22.7k m² and no campus MW on the cited page. Equinix SK1 and SK2, Bahnhof Pionen and Thule, GleSYS Stockholm, evroc Arlandastad and EcoDC Borlänge have unknown MW in this file. evroc’s 16,000 GPU design is stored as GPUs, not converted.

Common misreads

The arithmetic misread is to sum Falun 41 + 65 + 80, or to add STACK 18 to Conapto 24 to Digital Realty’s floor area, and call the result Sweden. Those rows do not share a definition, a scope and an as-of date. Falun 65 is a later campus total if E and F enter operation; it already looks forward through the 41 MW sentence. Falun 80 is an available-power envelope and a related future-supply press claim. STACK 18 and Conapto 24 are operator campus figures in Stockholm. They are kept visible as sourced rows. They are not bricks in a national wall.

The pipeline misread is to treat development as hidden live load. SWE02 and SWE04 have announced site-power figures; evroc and Borlänge still have unknown MW. Hectares in Sollefteå, building rights in Arlandastad and a GPU design are not load. Company aim dates — evroc’s H1 2025 ground-break and H2 2026 operation, atNorth’s Q4 2027 live wording for SWE02, EcoDC’s E and F quarters — remain aims until a dated commissioning source exists.

The map misread is to colour SE1–SE4 or the indicative transmission map as leftover IT power. Bidding-area price is not a connection offer. The transmission map is not underlying-network capacity. A hydropower-rich national mix is not a yes to a hall outside Falun or Staffanstorp. Those mechanisms belong in the power guides. They do not generate a missing national total on this page.

What to ask next

For every disclosed row, ask which original sentence produced it and whether IT versus facility power is defined. For Falun, ask which buildings were commissioned as of the balance-sheet date, which buildings remain in construction, and which paper — if any — sits behind the 80 MW available-power wording. For STACK 18 and Conapto 24, ask for the campus sheet that defines the unit and for the network-owner agreement if a data room requires it. Those documents are not in this launch set.

For unknown rows, ask whether the operator has simply not published megawatts, whether a facility page is missing, or whether two dated claims disagree. Meta, Microsoft and AWS are not-disclosed on the cited pages. Hydro66 is not-yet-checked against a current official facility page. GleSYS 3 MW and 5 MW remain two reported rows until a source explains the difference. Do not convert any of those gaps into zero, and do not borrow a directory figure to close them.

For pipeline rows, ask what would change the label from development to construction or operational. A land purchase, a hectare count, a GPU design or an acquisition notice is not that change. Then ask the network owner which written position exists for the plot — inquiry, reservation, agreement or energized load — and which reinforcements, if any, sit on the critical path. Write the answers with dates. If the owner will not put a number on paper, record that as a gap.

• Which MW type, scope and as-of date does each sourced row carry?
• Which unknown rows would be excluded from any attempted sum?
• Which network-owner paper, if any, sits behind a “secured power” sentence?

What this does and does not prove

This page proves that the reviewed register contains a small set of disclosed megawatt rows and a larger set of explicit unknowns. It proves that Falun 41, 65 and 80 are stored as three non-additive observations, and that STACK 18 and Conapto 24 are stored as separate operator-wording rows. It proves that development projects are counted as projects, not as hidden load. It does not prove a national Swedish megawatt total, a leftover connection figure for any bidding area, or an energization date for E, F, SWE02, SWE04, evroc or Borlänge.

It also does not prove occupied IT load, contracted utility capacity, or heat actually delivered. Those facts need different papers — customer meters, network agreements, district-heating settlements — that the cited operator pages generally do not publish. DataCenterSweden can keep original wording and evidence states visible. It cannot invent the missing national sum. If a later primary source discloses a defined megawatt with a date, that row can be added to the register. Until then, the gap stays a gap.