CoreWeave + Cerebras build huge Canadian site
What happened
CoreWeave and Cerebras are partnering with BCE to build what’s being billed as Canada’s largest purpose‑built AI data center, locking in alternative capacity and enterprise customers reported. That kind of regional capacity play can sway buyers who care about data residency and multi‑vendor procurement.
Why it matters
Bell (BCE) said it will underwrite a C$1.7 billion buildout for a 300‑megawatt AI campus outside Regina in the Rural Municipality of Sherwood. (bce.ca) The capacity split locks Cerebras in for 160 megawatts and CoreWeave for 140 megawatts, allocations disclosed in early coverage of the announcement. (ca.investing.com) Bell’s timeline calls for construction to start this spring, with the first data halls expected online in the first half of 2027 and full build‑out by the end of 2027. (datacenterdynamics.com) Bell projects the project will generate up to C$12 billion in economic value, create roughly 800 construction jobs and about 80 permanent roles once operational. (newswire.ca) CoreWeave plans to populate its 140 MW allotment with NVIDIA‑based GPU clusters, building on its earlier rollout of GB200 NVL72 instances and deep partnership with NVIDIA. (prnewswire.com) Cerebras is taking the 160 MW slot as part of its global inference and sovereign‑AI expansion after announcing a multi‑datacenter rollout using its wafer‑scale engines. (businesswire.com) CoreWeave’s existing commercial ties with NVIDIA include capacity agreements that let NVIDIA buy residual compute if unsold, a deal noted in SEC‑filings and reporting through April 2032. (datacenterdynamics.com)
Key numbers
- Bell (BCE) said it will underwrite a C$1.7 billion buildout for a 300‑megawatt AI campus outside Regina in the Rural Municipality of Sherwood.
- (bce.ca) The capacity split locks Cerebras in for 160 megawatts and CoreWeave for 140 megawatts, allocations disclosed in early coverage of the announcement.
- (ca.investing.com) Bell’s timeline calls for construction to start this spring, with the first data halls expected online in the first half of 2027 and full build‑out by the end of 2027.
- (datacenterdynamics.com) Bell projects the project will generate up to C$12 billion in economic value, create roughly 800 construction jobs and about 80 permanent roles once operational.
What happens next
- Bell (BCE) said it will underwrite a C$1.7 billion buildout for a 300‑megawatt AI campus outside Regina in the Rural Municipality of Sherwood.
- (ca.investing.com) Bell’s timeline calls for construction to start this spring, with the first data halls expected online in the first half of 2027 and full build‑out by the end of 2027.
- (datacenterdynamics.com) Bell projects the project will generate up to C$12 billion in economic value, create roughly 800 construction jobs and about 80 permanent roles once operational.
Quick answers
What happened in CoreWeave + Cerebras build huge Canadian site?
CoreWeave and Cerebras are partnering with BCE to build what’s being billed as Canada’s largest purpose‑built AI data center, locking in alternative capacity and enterprise customers reported. That kind of regional capacity play can sway buyers who care about data residency and multi‑vendor procurement.
Why does CoreWeave + Cerebras build huge Canadian site matter?
Bell (BCE) said it will underwrite a C$1.7 billion buildout for a 300‑megawatt AI campus outside Regina in the Rural Municipality of Sherwood. (bce.ca) The capacity split locks Cerebras in for 160 megawatts and CoreWeave for 140 megawatts, allocations disclosed in early coverage of the announcement. (ca.investing.com) Bell’s timeline calls for construction to start this spring, with the first data halls expected online in the first half of 2027 and full build‑out by the end of 2027. (datacenterdynamics.com) Bell projects the project will generate up to C$12 billion in economic value, create roughly 800 construction jobs and about 80 permanent roles once operational. (newswire.ca) CoreWeave plans to populate its 140 MW allotment with NVIDIA‑based GPU clusters, building on its earlier rollout of GB200 NVL72 instances and deep partnership with NVIDIA. (prnewswire.com) Cerebras is taking the 160 MW slot as part of its global inference and sovereign‑AI expansion after announcing a multi‑datacenter rollout using its wafer‑scale engines. (businesswire.com) CoreWeave’s existing commercial ties with NVIDIA include capacity agreements that let NVIDIA buy residual compute if unsold, a deal noted in SEC‑filings and reporting through April 2032. (datacenterdynamics.com)
