Murray Wood: Comrades in cowboy hats
Murray Wood says that person demanding Alberta separate from Canada in a social media post might be a farmer of a differ...
This story was first published on RealAgriculture.com on May 9, 2026.
In the rural municipality of Rosser, just outside Winnipeg, a 94,000 square-foot canola and pea processing facility is being converted into a Bell Canada AI data centre.
The building was completed in 2021. Merit Functional Foods, which built the facility with more than $100 million in federal financing, went into receivership in 2023, unable to compete with low-priced pea protein imports from China.
Read more:
- What is a data centre? Expert explains amid Regina debate
- B.C. says AI and data centre projects must compete for power in new selection process
- Landowners push for safeguards around proposed data centre near Regina
After sitting idle for several years, the building is being repurposed to house computing servers instead of grain crushers.
A province over, Bell also announced a much larger AI data centre in the rural municipality of Sherwood, where ground broke in early May. The $12 billion project is expected to be partially operational in 2027. Residents have raised concerns about drainage, water, noise, road impacts and the pace at which the project moved toward local approval.
These are not isolated cases. They are the first visible landings of a larger capital wave.
AI is digital, but the infrastructure behind it is not.
A data-centre is a warehouse-scale building, and what it requires to operate at scale extends well beyond its walls, from power substations, to cooling systems, to access roads, to transmission corridors that can collectively cover several hundred acres per major facility.
The Bell Regina facility alone will draw continuous electricity roughly equivalent to a mid-sized city at peak demand. That electricity requires transmission infrastructure, which in return requires land to cross and occupy. The digital economy thus has a physical footprint that competes for the same acres as agriculture.
An 86-year-old farmer in Pennsylvania, U.S., was offered more than $15 million for 261 acres wanted for data centre development. He said no. He said he loved the land and did not want to see it built over.
That story still feels exceptional because most farmers are not making land decisions against sentiment alone. They are making them inside a much harder set of conditions: tight margins, volatile markets, debt, labour stress, succession pressure, and a lot of uncertainty about what returns from farming will look like five or 10 years from now.
In Canada, the average farm operator was 56 in 2021. More than 60 per cent were 55 or older. Only 12 per cent of farms reported a written succession plan. At the same time, the value of land and buildings on Canadian farms rose 22.7 per cent between 2016 and 2021, and average farmland values rose another 9.3 per cent in 2025. That is the setting into which artificial intelligence (AI) infrastructure is arriving.
Canada’s federal government has committed $925.6 million over five years to establish sovereign, large-scale AI computing infrastructure, with the stated goal of positioning Canada to compete globally.
Innovation, Science and Economic Development Canada (ISED) has since opened a formal process for large-scale commercial AI data centres to be built in Canada.
Provinces including Ontario, Quebec, British Columbia, and Alberta are pursuing their own incentives, policies, and targeted investments to position themselves as attractive destinations for AI and data centre development.
Agriculture is used to trade-offs. AI infrastructure does not sit outside that logic. It lands directly inside it.
A data centre proposal could alter the farm decision environment in ways that extend beyond the parcel being rezoned. It could change what nearby land is worth to a buyer with a different use in mind. It could affect how long holding on still makes sense. It could also reshape succession decisions, especially when the next generation is uncertain or absent, and when the return from selling may look more predictable than the return from producing.
The pressure starts before any rezoning vote. Once a new alternative use can offer a price that agriculture cannot justify through production alone, the market calculus around adjacent land has already shifted. The acres may still be farmed. The market context has changed anyway.
It is easy to say farmland should be protected. It is harder to ask what exactly we are protecting when the return from selling may look more certain than the return from farming. Protect it for whom, and under what economic conditions? That question becomes sharper when profitability is uneven, weather risk is rising, and succession is unresolved.
There is a further complication. Some will argue that this could be an AI bubble, that some of the promised facilities will never materialize, and that current announcements are running ahead of durable demand. Recent analysis identified 309 data centres across Canada, but it also pointed to a lot of speculation behind the boom.
But that uncertainty may not necessarily reduce pressure on farmland. It may actually intensify it.
If producers think the window is real but short, some may feel stronger pressure to move while the money is there. If they think the outside offer may not last, waiting can start to feel like the riskier choice.
The Merit Functional Foods story is worth holding alongside the Pennsylvania farmer’s.
One illustrated a farmer who was able to say no to a very large offer. The other illustrates what happens when a public investment in agricultural processing falls short of its objectives.
Neither story is simple and both are relevant to where this is heading. They reflect a shift in conversation beyond technology alone to one of integration and long-term farm viability.
— Elisabeta Lika is an applied economist, policy strategist and a research associate with Canadian Agri-Food Policy Institute
- Listen to RealAg on the Weekend, Saturdays at 4 p.m. and Sundays at 1 p.m. on 650 CKOM and 980 CJME.
Read more:
