[ExI] Where should we build new multi-gigawatt AI Data centers?

Fri Nov 6 14:06:51 UTC 2026

*The local opposition to new AI data centers in the USA has started to
worry me, so I asked GPT another question. *

*ME: What are the prospects of building multi-gigawatt AI Data centers in
Canada?*

*GPT: " *Canada is potentially one of the world's best locations for
enormous AI data centers. In some respects, I think it has a better
long-term opportunity than Iceland because it combines abundant freshwater,
a cold climate, enormous hydroelectric resources, and a much larger land
area and electricity system. But there is a paradox: *Canada has enormous
energy resources, yet it currently has surprisingly little spare
electricity available in the places where data-center developers want to
build. *The distinction between energy resources and electricity that can
actually be delivered to a particular site is crucial.

I would look particularly at four provinces.

1. Quebec

Enormous hydroelectric resources, relatively inexpensive electricity and a
cold climate. Probably one of the most attractive long-term locations,
provided Hydro-Québec can expand generation and transmission

2. Alberta

The most interesting near-term development story. It has abundant natural
gas and is actively attracting data-center investment. Dedicated power
plants could bypass some grid constraints.

3. Manitoba

Extensive hydroelectric resources and a relatively small population. Worth
investigating for long-term developments, although new transmission and
generation would be necessary for gigawatt-scale projects.

4. British Columbia

Excellent hydroelectric resources and climate. However, the provincial
government is rationing new electricity allocations for AI and data centers.

*The surprising leader in actual investment: Alberta*

Alberta is particularly interesting because Meta has announced an
investment exceeding US $9.1 billion in a major AI data center in Sturgeon
County. The associated electricity project is a 932-megawatt
natural-gas-fired power plant. Operations are expected to begin in the
second half of 2030.

This illustrates an important strategy: instead of waiting for the
provincial electricity grid to supply an enormous new load, developers can
build dedicated power generation alongside the data center. Alberta has
substantial natural-gas resources, so this approach is feasible. However,
Alberta has received requests for approximately 21 GW of data-center
connections, while its electricity system operator has indicated that only
1.2 GW of additional data-center load could be accommodated by 2028 without
compromising reliability. That is a remarkable difference between
developers' ambitions and the electricity actually available.

*Why Quebec might be the biggest long-term opportunity*

Quebec has something Alberta doesn't: an enormous established hydroelectric
system. Hydro-Québec generates large quantities of relatively inexpensive,
low-carbon electricity. Its reservoirs also provide flexibility to balance
electricity demand. But Quebec has already encountered a problem:
technology companies have expressed interest in more electricity than the
province can readily provide. Hydro-Québec has consequently restricted new
electricity allocations for large data-center developments while it
evaluates its future generating and transmission requirements. For a
company thinking 10–15 years ahead, however, Quebec deserves serious
consideration.

*British Columbia is another interesting case*

British Columbia has abundant hydroelectricity, but its government has
decided that electricity allocation must be competitive.

Its current two-year allocation framework provides:

   -

   300 MW for AI data centers.
   -

   100 MW for conventional data centers.
   -

   200 MW for hydrogen-for-export projects.

Its 2026 application round received 15 applications requesting nearly 800
MW in total.This illustrates that abundant hydroelectric resources do not
automatically translate into unrestricted access for data-center developers.

*How much data-center construction is Canada actually anticipating?*

A useful October 2026 industry assessment from Colliers found:

   -

   Approximately 1.415 GW of operating data-center capacity in Canada.
   -

   At least 5.5 GW of named projects either planned or under construction.
   -

   Alberta accounts for most of the visible development pipeline.

Those figures are encouraging, but planned projects are not necessarily
financed, approved, connected to electricity or ultimately constructed.

The Canadian government's own energy outlook is more cautious. Its baseline
scenario anticipates an additional 1.5 GW of data-center electricity demand
by 2030. Its high-growth scenario anticipates 2.7 GW.

Canada has also introduced national principles requiring data centers to
protect electricity ratepayers, minimize water consumption and provide
lasting benefits to Canadian communities.

*My biggest question: Could Canada accommodate a 3 GW AI campus?*

A hypothetical 3 GW AI data center ( annual electricity consumption at
continuous full power would be 26.28  terawatt-hours) that would require:

   -

   Several new large power-generation projects.
   -

   Major transmission infrastructure.
   -

   Substantial backup generation and grid support.
   -

   A dedicated cooling and water-management system.
   -

   Long-term agreements protecting the electricity supply.

I think this is technically achievable in Canada, particularly through a
combination of hydroelectricity, natural gas, nuclear power and new
renewable generation. But I would not expect a 3 GW facility to materialize
quickly using existing spare electricity.
My particular interest would be Quebec and Manitoba for
hydroelectric-powered AI, and Alberta for dedicated-power developments. One
final thought: I suspect Canada could eventually become an important
alternative to the United States for AI computing, particularly for
training large models where a few milliseconds of additional network
latency are relatively unimportant.

The question isn't whether Canada has enough natural resources. It does.
The question is whether it can build generating capacity, transmission
lines and data centers quickly enough to capitalize on those resources.
That is an infrastructure and political challenge, rather than a
technological one."

*John K Clark*

On Thu, Nov 5, 2026 at 8:42 AM John Clark <johnkclark at gmail.com> wrote:

> *Many newly elected Democratic governors have advocated a pause an AI Data
> Center construction, so I asked GPT the following question:*
>
> *ME: What state looks to you like the most attractive for building large
> data centers in the next few years?*
>
>
> *GPT*: "I would pay particular attention to these five states:
>
>    1.
>
>    Georgia: Strong site-readiness measurements, but local opposition is
>    significant.
>    2.
>
>    Tennessee: Attractive combination of infrastructure and development
>    opportunities.
>    3.
>
>    South Carolina: Another promising southeastern location.
>    4.
>
>    Texas: Enormous potential, but substantial uncertainty over
>    electricity and permitting.
>    5.
>
>    Pennsylvania: Particularly interesting because of its natural-gas
>    resources and existing electrical infrastructure.
>
> Florida is also competitive, but faces its own land-use, water,
> electricity and hurricane-related challenges.
>
> I think we may be approaching a situation in which the availability of
> electricity, rather than the availability of money or computer chips,
> becomes the principal constraint on AI expansion. That could produce some
> surprising winners among states that currently receive relatively little
> attention in discussions about the AI industry."
> *ME: Iceland has lots of freshwater and cheap energy due to hydroelectric
> and geothermal power, so would that be a good place to build multi gigawatt
> AI data centers?*
>
> GPT: "Iceland is potentially an excellent location for a specialized AI
> computing facility in the 300 Megawatt range, but Iceland does not
> currently have the electrical infrastructure to accommodate the enormous
> multi-gigawatt campuses being contemplated by the largest American
> technology companies.  But two other countries probably could accommodate
> them,  Norway and Finland. Norway has enormous hydroelectric resources,
> Finland has substantial nuclear and other generation capacity, and both
> have much larger electricity systems than Iceland. I would investigate
> those two countries very seriously as potential European destinations for
> gigantic AI computing facilities."
>
> *John K Clark*
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20261106/c84c01ed/attachment-0001.htm>