Struggling marijuana farmers pivot to becoming AI data center sites to escape falling weed revenue

AI & Compute

Cannabis Growers Pivot to AI Data Centers as Weed Prices Collapse

Cannabis sales peaked in 2024 and prices are falling. Growers with megawatt-wired facilities are pitching them as 'mini data center nodes' for AI operators facing hyperscale opposition.

By
Rebecca Stone
Filed
Channel
AI & Compute
Read
3 min read

A 32,850-square-foot cannabis facility in Kalamazoo, Michigan, already wired for 1.5 megawatts and expandable to 2.5 megawatts in under a year — and 15 megawatts long term — is exactly the kind of property now drawing interest from AI data center operators, including, reportedly, OpenAI and Anthropic.

The facility, profiled by Forbes, also has fiber internet provisioning and a state-of-the-art HVAC climate control system. That combination of power, cooling and connectivity is what makes former grow houses tempting targets for conversion as the cannabis industry contracts.

Arizona Public Media reports that cannabis sales peaked in 2024 and oversupply has driven prices down. Faced with rising competition, costs and taxes alongside falling demand, several growers are considering selling their cultivation facilities or repurposing them into "mini data center nodes."

Moe Asnani, former owner of two Tucson D2 Dispensaries, is one of them. "For some reason, there's a lot more cannabis grown in Arizona than Arizonans use. And so, a lot of folks are getting out of that business, like myself, at least on the cultivation side. But these facilities can potentially be repurposed to be mini data center nodes," he told the broadcaster.

Asnani frames the smaller sites as a decentralization play. "The whole idea is if you create these smaller nodes where people control their own data, where they're not at the mercy of other people's updates and things like that — that's kind of where the idea came in, because I believe in decentralization," he said.

Why grow houses fit the bill

Cannabis farming is power- and water-intensive by design. Grow operations need dense lighting, aggressive climate control and reliable water — the same raw ingredients a data center demands, and increasingly hard to secure as hyperscale AI projects compete for grid capacity.

The concept has a name: "adaptive reuse." Tucson City Council member Paul Cunningham calls it a more palatable alternative to massive purpose-built sites on the outskirts of town.

"Data centers are dirty words, and everybody knows that. And no one wants to build a brand-new data center anywhere," Cunningham said. "But if we have buildings that were using that amount of water and that amount of energy already in their previous function and [it] can be done without being invasive, I'm OK with exploring that option."

His point lands hardest where hyperscale projects have stalled: massive sites from operators like OpenAI and Anthropic face mounting community opposition, and smaller distributed nodes could dodge some of that friction.

What could kill the deal

Massive demand for compute doesn't guarantee a payday for cash-strapped growers trying to unload unprofitable properties. Niki Krear, vice president of acquisitions at real estate investment trust New Lake Capital Partners, says operators hunt for a specific sweet spot.

"It's a combination of finding affordable power, dense fiber networks, and favorable tax policies," Krear told MJBizDaily.

Even though many cultivation facilities carry higher-than-average power capacity for lighting and climate control, data centers often demand two to three times more — and those upgrades can take up to five years.

The obstacles stack up quickly:

  • NIMBY opposition from neighbors of smaller sites
  • Zoning restrictions, moratoriums and regulatory hurdles
  • Missing economies of scale that raise per-unit costs

That last one matters more as AI chips gain power and efficiency over the next five years. Data centers will need to sell existing equipment and buy new GPUs to stay competitive, and suppliers typically offer favorable pricing on larger orders — meaning upgrades at these smaller sites will likely cost more per unit.

Still, with opposition to large new projects only growing, former cannabis farms could suit operators desperate to bring compute online quickly — provided a site checks most of the boxes on power, fiber and taxes.

Source: Tom's Hardware

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering technology and product news in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.

892 articles

Related articles

« Previous articleNext article »