Keel and Hive Rally as Bitcoin Miners Pivot to AI With Asset Sales and $115M Funding

Wednesday’s pop in Keel Infrastructure (KEEL) and Hive Digital Technologies (HIVE) isn’t just a sympathy move with Bitcoin. It’s a repricing of business models as miners increasingly convert hashing power economics into high-performance compute (HPC) capacity for AI.

The moves were concrete. Keel—formerly Bitfarms—closed the sale of its Paso Pe, Paraguay mining site, netting $13 million. Management had estimated up to $30 million at one point, so the final take reflects roughly 56% less than the high end. The company signaled it has no remaining non-core assets to manage or sell, effectively drawing a line under its Latin American footprint and redirecting capital to an AI-centric, North America–focused buildout. Keel framed the sale as pulling forward roughly two to three years of estimated free cash flow under current market conditions and redeploying it immediately into its HPC/AI pipeline.

Hive took the funding route, closing a $115 million private offering of convertible notes, with proceeds earmarked for GPUs and data center development, among other uses. That augments a six-month push into AI infrastructure, including a November agreement involving Dell via Hive’s Buzz subsidiary.

Price action has followed the narrative. Over the last month, KEEL has climbed more than 40% and traded around $3.06 on Wednesday, up roughly 9% intraday. HIVE has rebounded more than 31% in a month, recently at $2.66 and up over 7% on the day. Bitcoin itself is up about 4% in 24 hours to near $79,000, still about 37% below its October peak of $126,080.

What matters here isn’t the label shift from “mining” to “AI,” but the cash flow math behind it. Bitcoin mining returns compress quickly when network hash rate climbs or power contracts reprice, and the most recent halving has historically amplified that sensitivity. In that environment, monetizing a peripheral site today—even at a discount to initial hopes—can be rational if the proceeds accelerate GPU-ready capacity that captures stronger dollar-per-kilowatt returns. AI training and inference workloads, when paired with dense power, reliable cooling, and low-latency networking, often yield revenue profiles that are less binary than mining block rewards and more tied to multi-year customer contracts.

This pivot isn’t riskless. AI demand looks durable, but customers can be concentrated, procurement cycles are lumpy, and service-level penalties introduce operational rigor that mining didn’t require. Converting an ASIC-optimized footprint into GPU-grade infrastructure means rethinking design: liquid or immersion cooling, higher rack densities, robust fiber paths, and power usage effectiveness that enterprises scrutinize. Capital structure also shifts. Hive’s convertibles reduce immediate dilution but add future overhang and interest obligations; execution needs to outpace that drag.

Still, investors are often rewarding clarity. Keel’s clean exit from Latin America removes jurisdictional drift and FX noise, even if North American power markets can be pricier. It also aligns with hyperscaler and OEM ecosystems, where supply chain access to GPUs, firmware support, and SLA credibility become competitive edges. Hive’s prior Dell-linked initiative supports that same message: partner where it improves procurement and integration, rather than trying to brute-force a standalone supply chain in a constrained GPU market.

There’s a psychological layer too. Markets have treated “hash rate for hash rate’s sake” as a weaker story post-halving, while “compute as a service” feels tangible when accompanied by signed notes, announced capex plans, and site-level conversions. The ethical debate on energy use doesn’t disappear; it simply changes shape—from securing a permissionless ledger to enabling AI workloads. Operators that disclose PUE, grid interconnection benefits, and curtailment strategies could see an incremental credibility premium.

Net-net, both companies are exchanging volatile, difficulty-driven yield for contractable compute economics. Keel’s decision to realize two to three years of expected free cash flow upfront, even at $13 million versus a higher prior estimate, looks like a trade for speed. Hive is choosing scale via convertibles to secure GPUs and build capacity while the window is open. If they execute on power, cooling, and customer acquisition, the market’s recalibration of their multiples may continue. If not, this rerating will prove transient.