Telus Data Center Exit in Vancouver Shows Why Siting Risk Shapes AI Capacity Plans
Telus has stepped back from two planned AI data centers in Vancouver, and the developer says both will go ahead with a new partner. For operators of sovereign AI capacity, the story is about how partners, permits and power allocations move capacity between sites.

What happened
Telus has decided not to co-develop the two Vancouver AI data centers it announced with developer Westbank in May 2026, DCD reported on 16 November 2026. Telus gave no reason. Westbank says both projects will go ahead with a new partner that will be announced in the coming months, and Telus says it will keep expanding its Kamloops facility while it evaluates new partners and locations in British Columbia for long-term sovereign compute.
The two sites are very different projects. At 150 West Georgia Street, a new ten-storey building was planned for a 2029 launch. Daily Hive Urbanized, republished by MobileSyrup, reported that the site needs no rezoning because a 2023 text amendment already permits bulk data storage there, and that the building would sit above Creative Energy's steam plant, with heat recovered from the data center helping to produce steam that heats more than 200 downtown buildings. The second project converts Westbank's M3 office building at 111 East 5th Avenue in Mount Pleasant, which Westbank's project website lists at 26 MW. DCD said M3 was meant to come online at the end of 2026 and scale through 2029, but the previous city council sent its rezoning back to staff in July, and a new public consultation runs from 2 to 15 December 2026.
According to DCD, the two Vancouver developments were planned at more than 150 MW combined by 2032, with closed-loop liquid cooling and heat recovery, and Telus had already secured an initial 85 MW from BC Hydro under the province's competitive bid process. Telus is expanding Kamloops to 20 MW by the end of 2028, according to Daily Hive Urbanized.
The local setting has changed too. DCD reported large protests against the Vancouver projects over water and power use, environmental impact and a lack of consultation, and said the new city council elected on 17 October meets on 18 November to consider a motion on tightening data center rules. The Goldstream News Gazette reported that BC Hydro limits power for all new data center projects to 200 MW a year for two years from 2026, with projects already underway when the rules were set, such as these, free to connect.
Why it matters for an AI data center operator
Sovereign AI capacity is planned across several sites and several years of utility work. When a co-development partner leaves, a rezoning goes back for consultation, or power depends on a bid process, the capacity on the plan moves between sites and dates slip. In our reading, the risk in the Telus case sits in partners, permits and power.
In our view, an operator running a multi-site AI buildout needs four numbers per site that stay true whatever the headlines say. The first is power actually allocated by the utility against the megawatts on the plan. The second is rack power and cooling readiness: whether racks are energized and the cooling loop is online before NVIDIA, AMD and other accelerators ship. Servers that arrive at a site still waiting for a consultation sit idle.
The third is GPU pool capacity by site, so work can be placed where capacity already exists when one site slips. The fourth is heat recovery and efficiency reporting, PUE, WUE and EED included. DCD noted that projects using heat recovery and reducing water use generally have an easier time getting a connection under the BC process, which makes those numbers part of the siting case as well as a compliance task.
What the operator sees in Sensaka
Sensaka for AI Data Centers keeps power, cooling and accelerators in one inventory across sites. The AI Cockpit takes questions in plain language and returns a custom board and a 30-second data video, on premises and without an external AI model. Preset questions cover the checks above: Which racks or zones are near their rated power? Is the liquid-cooling loop healthy: temperature difference, flow and leaks? Which GPUs are idle and could be reclaimed?
GPU pool operations shows allocated, busy and idle capacity for each pool and standard spec, and GPU-hours reserved against GPU-hours actually busy for each tenant, project and site. When a new site is late, that is where an operator sees which existing clusters have room.
Liquid cooling monitoring reads CDU status, supply and return temperature, pressure differential, flow and leak events, next to rack power density. Before more GPU servers are installed, teams review rack space, power density, temperature and cooling capacity together, and the platform highlights racks where power or cooling may limit deployment even when rack units remain free. This is monitoring: any control action stays with the site's own procedures. Follow-up work becomes a ticket synced with ServiceNow or Jira.
Takeaway
A partner exit changes when and where capacity arrives. Operators cannot control council agendas or partner decisions. They can keep allocated power, cooling readiness, GPU pool headroom and efficiency reporting current for every site, so the next change in the plan is a scheduling question rather than a surprise.
Frequently asked questions
Why did Telus exit the Vancouver AI data center projects?
Telus has not given a reason. A Telus spokesperson told DCD on 16 November 2026 that the company remains committed to building high-performance compute infrastructure in Canada, and Telus said it will keep expanding its Kamloops facility.
Are the two Vancouver AI data centers still going ahead?
Westbank says yes. A Westbank spokesperson said the 150 West Georgia Street and M3 projects will proceed with a new partnership that will be announced in the coming months.
How much power was planned for the Telus and Westbank sites?
DCD reported that the two Vancouver developments were planned at more than 150 MW combined by 2032, and that Telus had secured an initial 85 MW from BC Hydro under the province's competitive bid process for AI and data centers.
See rack power, cooling readiness and GPU pool headroom for every site in one record. Start a free 30-day pilot on your own hardware, inside your own network.
Start a Pilot →Sensaka for AI Data Centers
GPU pools, workload queues, GPU-hour and token metering, and an operations cockpit with an AI Cockpit, on one on-premises platform.
Related articles & analysis
