GPU ColocationHut 8 Canada · October 2026

Dense GPU racks.
Canadian soil.
Capacity now.

Place your own GPU servers in Tier III designed Canadian data centres. Hut 8 provides the power, liquid-cooling-capable infrastructure, connectivity and security around them: up to 135 kW per rack, with 5 MW of high-density capacity available now in Kelowna and Mississauga.

Capacity as at Q3 2026. Confirm current availability with Hut 8.

▶ Hear the 90-second summary

Short on time? Get the summary in 90 seconds.

InfraHub's independent take: what Hut 8 is offering, who it suits, and what to ask before you book.

About 1:32 · captions on
Press play to start

Prefer to skim? Switch to Read.

Audio isn't available in this browser. Use the Read tab for the full summary.

Written by InfraHub, not Hut 8, and narrated with an AI voice. Figures are the provider's own claims.

135 kW+Per rack, so a full GPU node can sit in one cabinet
5 MWHigh-density capacity available now, as at Q3 2026
99.99%Uptime SLA at both sites: 53 minutes of downtime a year at most
2Tier III designed sites: Kelowna, BC and Mississauga, ON
The density problem

A GPU cluster doesn't fit a data centre built for general IT.

A general-purpose facility spreads power across a floor. A GPU cluster concentrates it in a single cabinet. Three things decide whether a site can host yours, and a fourth decides whether you can start on time.

Power per rack

Ask for the kilowatts delivered to one cabinet, not a facility total. The right figure depends on your hardware.

135 kW and up

Cooling

As rack power rises, air alone does less of the work. Liquid-cooling capability matters as densities climb.

Liquid-cooling capable

Resilience

Redundant power and cooling mean one failure doesn't stop training or inference. Check the design standard, power path and generator runtime.

Tier III designed

Time to deploy

Capacity is a finite pool and your timeline matters. High-density deployments here typically go live in a few weeks.

Typically weeks

What is GPU colocation? You own the compute: GPU servers you buy and control. The facility provides the power, cooling, network and security around them.

The sites

Two Tier III designed sites. Pick the one that fits your cluster.

Both are liquid-cooling capable, carrier-neutral and connected to Hut 8's 100G national backbone. They differ on cabinet size, energy source and location.

Kelowna, BC

KEL1
Tier III designedUp to 135 kW / rack50U cabinets

Best for liquid-cooled GPU clusters on 95% renewable hydroelectricity, larger 50U cabinets, or a second site alongside Vancouver.

Mississauga, ON

GTA1
Tier III designed135+ kW / rack42U cabinets

Best for dense GPU deployments close to the Greater Toronto Area, with 42U cabinets and five carriers on site.

Kelowna (KEL1)Mississauga (GTA1)
Available now2.5 MW of 3 MW secured2.5 MW of 3.5 MW designed
Power per rackUp to 135 kW135+ kW
Cabinets50U42U
CoolingChilled-water closed loop plus air, N+1, two 200-tonne chillers, cold aisle containment, winter free cooling. Liquid-cooling capable.Chilled water. Liquid-cooling capable.
Power path2N (A+B)2N power and 2N UPS
Generators2 × 1,500 kW, 29+ hours at full load10 × 700 kW (N+1), 24+ hours
Energy95% renewable hydroelectricityUtility grid
CarriersTelus, Rogers, Zayo, BellRogers, Bell, Zayo, Cogent, Telus
LocationDescribed by Hut 8 as one of the lowest geographic risk profiles in North AmericaAbout 25 km from Toronto
Connected, secured, in Canada

The network, security and residency around the racks.

Carrier-neutral connectivity

100G ports, dark fibre, cross-connects and diverse fibre entrances. A 100G national backbone links all five Hut 8 Canada data centres.

100G backbone

Security and uptime

Every site is SOC 2 Type II certified, with card and biometric access, 24/7 NOC-monitored CCTV and a 99.99% uptime SLA.

SOC 2 Type II

Canadian data residency

Equipment at every Hut 8 Canada site is hosted in Canadian facilities, which supports PIPEDA data residency requirements.

PIPEDA-supporting
InfraHub's takeWritten by InfraHub, not by Hut 8. Figures are the provider's own.

Best fit

  • Teams that own, or are buying, their own GPU servers
  • AI workloads that need to stay in Canada
  • Deployments where a full GPU node should sit in one cabinet

Probably not

  • You want to rent GPU capacity rather than host your own hardware
  • You want the AI stack built and run for you: see Hut 8's Managed AI campaign
  • Your racks run at standard densities: ask Hut 8 about its other three sites

Ask before you commit

  • The per-rack kW for your hardware, not the facility total
  • Whether your nodes need liquid cooling, and when
  • How much of the 5 MW remains for your timeline
  • Which site fits: renewable hydro in Kelowna or proximity to Toronto
Deployment

From shortlist to live in a few weeks.

Hut 8 says high-density deployments typically go live in a few weeks. The usual path has four steps.

01

Shortlist a site

Compare Kelowna and Mississauga on kW per rack, cabinet size, cooling and carriers.

02

Review the spec sheet

Check the site specification against your hardware: power, cooling, resilience and connectivity.

03

Choose your footprint

Half rack, full rack or private cage, all with A+B power.

04

Deploy and go live

Install your hardware with remote hands or 24/7 smart hands on call.

Footprint options

Pricing on request. Tell us your rack count and kW per rack and Hut 8 will scope it.

Start small

Half rack

A half-cabinet footprint for smaller deployments.

Dense node

Full rack

A full cabinet, enough at 135 kW and up for a complete GPU node.

Dedicated space

Private cage

A dedicated cage for larger or more isolated deployments.

Support: remote hands (billed hourly) or 24/7 smart hands. On-site equipment storage is available.

Frequently asked questions

Common questions about GPU colocation in Canada.

How much high-density capacity does Hut 8 have available in Canada?

As at Q3 2026, Hut 8 reports 5 MW of high-density capacity available now across its two flagship sites: 2.5 MW in Kelowna, BC (of 3 MW secured) and 2.5 MW in Mississauga, ON (of 3.5 MW designed). Capacity changes, so confirm current availability with Hut 8 before planning around it.

How many kW per rack can Hut 8 support for GPU deployments?

Both high-density sites support 135 kW and up per rack. Kelowna supports up to 135 kW per rack in 50U cabinets. Mississauga supports 135+ kW per rack in 42U cabinets. At that density, a full GPU node can sit in a single cabinet.

Can the Kelowna and Mississauga sites support liquid cooling?

Yes. Both sites are liquid-cooling capable. Kelowna uses a chilled-water closed loop plus air with N+1 cooling, two 200-tonne chillers, cold aisle containment and a winter free-cooling advantage. Mississauga uses chilled water.

How quickly can a GPU deployment go live?

Hut 8 says high-density deployments typically go live in a few weeks. The usual path is to shortlist a site, review its spec sheet and deploy. Timing depends on your hardware and the capacity available when you ask.

Does my data stay in Canada?

Yes. Equipment at every Hut 8 Canada site is hosted in Canadian facilities, which supports PIPEDA data residency. Every site is SOC 2 Type II certified, uses card and biometric access, and is watched by 24/7 NOC-monitored CCTV.

What happens if the power grid goes down?

Both sites are designed to keep running through a grid outage. Kelowna has 2N (A+B) power, N+1 cooling and two 1,500 kW generators with more than 29 hours of runtime at full load. Mississauga has 2N power, 2N UPS and ten 700 kW generators (N+1) with more than 24 hours of runtime. Both carry a 99.99% uptime SLA.

Should I choose Kelowna or Mississauga?

Choose Kelowna for liquid-cooled GPU clusters on 95% renewable hydroelectricity, larger 50U cabinets, or a second site alongside Vancouver. Choose Mississauga for dense deployments close to the Greater Toronto Area, with 42U cabinets and five carriers. Both connect to Hut 8's 100G national backbone.

What colocation options and support are available?

Hut 8 offers half rack, full rack and private cage colocation, all with A+B power. Support is available as remote hands or 24/7 smart hands, and Remote Hands and Eyes is billed hourly. On-site equipment storage is available, and every site is monitored by a 24/7 network operations centre.

Now accepting capacity requests

Ready to place your GPU cluster in Canada?

Tell us your rack count, kW per rack and timeline. Your request goes to Hut 8's team and to InfraHub.

  • Capacity available now at both sites, as at Q3 2026
  • Both sites Tier III designed and liquid-cooling capable
  • Canadian data residency: equipment hosted in Canadian facilities
  • Typical go-live in a few weeks for high-density deployments

Request capacity and pricing

Include your rack count, kW per rack and target date in the message.

    Canadian team · Your details are shared only with Hut 8 and InfraHub
    This campaign was built and managed by InfraHub.ca on behalf of Hut 8 Canada · Campaign Tier · October 2026. Specifications and capacity figures are provided by Hut 8 Canada and reflect its GPU colocation guide published 30 September 2026, as at Q3 2026. View Hut 8 Canada's InfraHub profile