Keel Lost $141 Million To Exit Bitcoin Mining

The Call That Closed The Book

On August 10 2026, Keel Infrastructure held its second-quarter earnings call under a name that did not exist a year ago. The company began life as Bitfarms, one of the larger publicly traded Bitcoin miners on the Nasdaq and Toronto Stock Exchange, and CEO Ben Gagnon used the call to confirm what the rebrand had signaled: every US Bitcoin mining operation is now decommissioned. Between April 1 and August 7, Keel sold 1,085 Bitcoin for roughly $75 million, leaving 1,861 Bitcoin on the balance sheet, and management told analysts it intends to sell the rest by the end of the year.

The numbers on the rest of the income statement matched the scale of the decision. Revenue came in at $30 million, down 50 percent from $61 million in the second quarter of 2025. The company reported a $141 million net loss for the quarter, against an $11 million net profit a year earlier - a swing of more than $150 million in twelve months. Shares fell more than 11 percent the same day, even though the loss was, in large part, the company choosing to close a chapter rather than being forced to.

From Miner To Landlord: A 2.2-Gigawatt Pipeline

Keel is not shutting down, it is repointing. The company's new business is leasing power and shell infrastructure to AI and high-performance-computing tenants, and Gagnon detailed a project list that totals close to 2.2 gigawatts of secured and expansion capacity. Moses Lake in Washington state carries 18 megawatts and is the first site scheduled to be ready for service in 2027. In Pennsylvania, inside the PJM grid, Sharon brings 110 megawatts and Panther Creek has 350 megawatts secured with room to expand past 500, while Scrubgrass is running a load study for a campus above 1 gigawatt. In Quebec, Sherbrooke's 96 megawatts came from consolidating three separate Bitcoin-mining power contracts into a single agreement built for AI hosting, pending provincial sign-off.

The company is paying for the transition with capital, not cash flow. Total liquidity reached $819 million, up from $533 million in May, after Keel upsized a convertible note offering from $350 million to $458 million specifically to fund power buildout at Panther Creek and Scrubgrass. It also hired Ganesh Aiyer, formerly chief business officer at Digital Realty, as president - a data-center executive brought in to run a data-center business, not a mining one.

Power Was Always The Constraint

Gagnon's most quoted line from the call was blunt: "Power is the constraint. Everything else is downstream of it." He also told analysts that "the defining constraint of the most important technology of our lifetime was not chips or capital, it was power" - a claim aimed at AI, but one that describes Bitfarms' entire history just as well. A Bitcoin miner's core asset was never really mining rigs or hashing algorithms; it was a portfolio of power purchase agreements and grid interconnection rights, the same assets an AI data-center operator needs.

That is the uncomfortable part of the story. If the scarce input - grid-connected power, in usable megawatts, at a price that clears - was the real asset the whole time, then the years spent optimizing mining efficiency, chip generations and Bitcoin price exposure were optimizing the wrong variable. The strategy changed. The constraint that mattered did not.

Repositioning Or Rationalizing: The Test Keel Passed

Every failing strategic bet gets defended, at some point, as "the cost of repositioning." The test for an owner-reader is not whether the new story sounds better than the old one - it almost always will, because someone is selling it to you on an earnings call. The test is whether the company is willing to book the loss on the old bet in full, in public, in one quarter, rather than smoothing it across several. Keel decommissioned every US mining site and committed to selling essentially its entire Bitcoin position by year-end. That is a company admitting the original bet's economics never worked at the scale it needed, not one quietly hedging while it waits to see which story wins.

The distinction matters because the alternative - keeping a shrinking legacy business alive "for optionality" while a new one is built beside it - usually means neither gets fully funded and neither gets a clean verdict. Ask of any pivot you are evaluating, in your own portfolio or a company you hold: is the old bet being closed, or merely deprioritized? A company that cannot answer that plainly on an earnings call is usually still defending the sunk cost, not the new plan.

What To Watch Next

Three things will tell an owner whether Keel's bet is paying off. First, whether the remaining 1,861 Bitcoin actually gets sold by the stated year-end target, since a slower exit would suggest less conviction than the call projected. Second, whether Panther Creek and Sharon convert their "active commercial discussions" into signed AI hosting contracts in 2027, since uncommitted capacity in PJM is only valuable once a tenant is under contract. Third, whether the 11 percent stock drop on the announcement day proves to be the market pricing in execution risk on an unproven business model, or simply the last exit for shareholders who bought Bitfarms for Bitcoin exposure and never wanted power infrastructure at all.

Keel's own framing is that scarce power, not mining scale, was the asset worth keeping - and that everything else, Bitcoin included, was downstream of it. Owners do not need to agree with that thesis to learn from the discipline of testing it this cleanly: name the constraint, close the position that was never really about the constraint, and let the market re-rate you on the thing you say actually matters.

Read next: 555 Was the Real Price. The Cap Said 325.  |  FERC Just Put the Rules for Connecting Big Loads on a 60-Day Clock