Private project, public pressure

Costs: a headline investment number is not a public benefit plan

Large data centres can bring investment. They can also consume power capacity, hardware supply, land, labour, public attention, and municipal infrastructure.

Whose money, whose hardware, whose resources?

“This is a private investment” is not the end of the question. Large private projects can still create public costs through power systems, roads, municipal services, water infrastructure, emergency response, land-use pressure, and cleanup risk.

AI demand also changes hardware markets. GPUs, high-bandwidth memory (HBM), servers, power equipment, cooling gear, switchgear, transformers, and skilled labour all get pulled into the same global buildout.

Plain-English HBM: high-bandwidth memory is specialized ultra-fast memory used beside high-end AI chips and GPUs. When hyperscalers buy mountains of it, everybody else feels the shortage eventually.

Do

Publish local procurement targets.
Identify public infrastructure costs.
Disclose tax treatment and incentives.
Show permanent jobs separately from construction jobs.
Explain hardware, fuel, water, and grid pressure.
Require public reporting after opening.

Don’t

Use a headline investment number as proof of public benefit.
Ignore imported equipment and offshore value.
Pretend private capital cannot create public risk.
Say “jobs” without numbers, duration, skills, and wages.
Make local communities absorb impacts for a vague future economy.

The resources checklist

Resource	Question	Condition
Money	What public costs or incentives exist?	No hidden subsidies. Publish the deal.
Hardware	Does the project worsen GPU, memory, server, or equipment supply pressure?	Be honest about global scarcity and local benefit.
Labour	Who gets the work?	Alberta apprenticeships and local procurement.
Infrastructure	Who pays for roads, substations, gas, water, sewer, and emergency capacity?	Private benefit should not quietly become public cost.

Sources and notes

TrendForce on AI server demand, memory pricing, and supply pressure.
TrendForce on AI and data-centre demand worsening memory supply-demand imbalance.
Meta Sturgeon County announcement for 1 GW scale, CA$13B investment, closed-loop dry cooling language, partners, and public reporting statements.
Reuters reporting on Meta scale, ability to scale, Greenlight timing, 250 MW interim Capital Power supply, and natural gas requirements.