When bitcoin miners pivot to AI infrastructure, they bring more than real estate—they bring power expertise that Big Tech desperately needs and can't build fast enough.
The Summary
- BCE Inc. signed a major AI infrastructure deal with a former Bitcoin mining company, highlighting how crypto mining infrastructure is becoming the foundation for AI compute
- Keel (formerly Hive Digital) hired data center veteran Ganesh Aiyer from Digital Realty to lead commercial growth, sending its stock up 10%
- Record AI infrastructure spending is driving capital raises across listed companies, with former mining operations sitting on exactly what AI labs need: power and cooling
- The deal positions Canada as a data sovereignty play for enterprises that want AI compute without sending everything to U.S. hyperscalers
The Signal
BCE Inc., Canada's largest telecom, just bet that the future of AI infrastructure runs through a company that used to mine Bitcoin. The deal with Keel marks a turning point: bitcoin mining operations aren't just pivoting to AI, they're becoming strategic partners for Fortune 500s that need compute capacity yesterday.
Keel rebranded from Hive Digital earlier this year, signaling its shift from crypto mining to AI infrastructure. The company's stock jumped 10% after announcing it hired Ganesh Aiyer, a former Digital Realty executive who spent years building hyperscale data centers. That's not a vanity hire. Digital Realty manages over 300 data centers globally. Aiyer knows how to talk to enterprise buyers, navigate power contracts, and scale infrastructure that enterprises actually trust.
"Former mining operations are sitting on exactly what AI labs need: power and cooling at scale."
Here's why this matters more than another AI infrastructure deal. Bitcoin miners already solved the hardest problems AI companies are facing right now:
- Securing massive power contracts in regions with cheap, reliable electricity
- Building cooling systems that handle extreme heat loads 24/7
- Operating facilities in jurisdictions with clear regulatory frameworks
- Managing uptime requirements where downtime costs millions per hour
AI infrastructure spending is hitting record levels, and the constraint isn't chips anymore. It's power. Nvidia can ship H100s. But getting 50 megawatts of reliable power with redundant cooling in under 18 months? That's the bottleneck. Former miners already have it.
The BCE deal also plays into something enterprises care about more than Silicon Valley admits: data sovereignty. The partnership enhances Canada's AI capabilities while keeping data inside Canadian borders, which matters for regulated industries like finance, healthcare, and government contractors. Enterprises want AI without exposing proprietary data to U.S. hyperscalers or overseas jurisdictions with murky privacy laws.
The Implication
Watch for more telcos and enterprises to bypass traditional cloud providers and cut deals directly with infrastructure players that control power and real estate. The AI infrastructure stack is fragmenting. Hyperscalers still dominate model training, but inference and private AI workloads are moving to distributed providers that can offer lower latency, better economics, and jurisdictional control.
If you're tracking the agent economy, this is the plumbing that makes it possible. Autonomous agents running 24/7 need inference capacity that's cheap, fast, and reliable. Former mining operations turned AI infrastructure providers are building exactly that. The companies that figured out how to run ASICs profitably in 2022 are now the ones powering the agents everyone will be using in 2027.