Don't Mistake Keel's $400,000 Insider Buying for a Signed AI Lease

Don't Mistake Keel's $400,000 Insider Buying for a Signed AI Lease

The retail bull case for Keel Infrastructure rests on a stack of SEC filings. When the company's insiders disclosed a second open-market purchase within days, the stock edged higher overnight and the message boards did what they always do: management is buying, so the AI data-center pivot must be real. Insiders rarely buy their own stock just for fun, and that instinct deserves a fair hearing — but a couple of Form 4s are not an investment thesis. They are a starting point. Start with the arithmetic.

$400,000 Against a $2 Billion Market Cap

Within the past week, Keel disclosed three open-market purchases. CEO Benjamin Gagnon bought 58,888 shares for roughly $196,000 at about $3.33 in mid-August, and shares jumped 8.4% in the session after the filing. Chief Operating Officer Liam Wilson then bought 26,472 shares at $3.78 in an open-market purchase, and added a second $100,000 tranche at $3.25 a few days later, near the recent lows.

Combined, that is three insider purchases totaling $396,200 over the past 12 months, with zero insider sales. It reads as conviction until measured against the stock's roughly $2.08 billion market capitalization: about two one-hundredths of a percent of the company, and less than a third of one percent of a single session's dollar turnover. Even the CEO's entire direct stake after buying — around $4.5 million worth of stock — is a rounding error next to the enterprise.

Insider trades are disclosed on SEC Form 4s, and open-market cash purchases are the strongest kind of insider signal because the executive is spending his own money rather than banking compensation. But they are a confirmation signal, never the thesis itself, and when the supposed smart-money stamp is measured in hundredths of a percent of market value, the response is to check the numbers the filings don't contain — not to conclude that insiders voted with a position large enough to matter.

The Numbers the Chatter Skips

Here is what the insider-buying headlines leave out: Keel just reported revenue of $30.4 million in the second quarter, down 50% year over year, along with a net loss of $65 million that came in wider than the consensus forecast of roughly $0.08 per share. The revenue collapse is not mysterious. The company — formerly crypto miner Bitfarms — shuttered its U.S. Bitcoin mining operations this spring to redeploy the sites toward AI and high-performance-computing hosting, and the quarter paid the price. The operating loss came to $140.8 million, though $84 million of that was non-cash depreciation, while adjusted EBITDA — earnings before interest, taxes, depreciation, and amortization, a rough proxy for cash earnings — swung from a positive $6.6 million a year earlier to negative $23.7 million.

The trajectory looks worse before it gets better. Quarterly revenue stepped down from roughly $52 million in Q4 2025 to $37 million in Q1 2026 to $30 million in Q2 2026. The old engine has been switched off and the new one has not started.

The Moat Is Real. The Lease Is Not.

Put a price on that. The market values Keel at about $2.08 billion against roughly $190 million of trailing revenue — approximately 11 times trailing sales — for a company with a negative gross margin and trailing free cash flow burn of roughly $400 million over the past year. With no earnings to speak of, a price-to-sales multiple is the market's way of prepaying for profit that does not exist yet. And 11 times sales is no contrarian discount inside its own universe: the AI-hosting complex trades at similar or considerably higher multiples, with Core Scientific near 13 times and IREN near 20 times trailing sales. The whole group is priced on the same AI-megawatt story.

That is where the peer comparison exposes the gap that matters. In this business, value is created at the lease signing: a data center with a hyperscale tenant is a contracted stream of cash flow, while one still in negotiation is an expensive hole in the ground being filled with somebody's capital. Keel has disclosed no signed tenant lease; management describes active negotiations at three priority sites. The market has arguably already paid for the AI story before a tenant has signed anything.

The assets behind the story are genuinely scarce, which is why the name belongs on a watchlist rather than in a punchline. A 2.2-gigawatt development pipeline with established grid interconnections in Pennsylvania, Washington, and Quebec is exactly the power-and-interconnection package that AI buildouts fight over. But a moat survives stress only when someone pays to use it. Keel holds roughly $819 million of liquidity — about $698 million in cash plus $121 million of unencumbered Bitcoin — and management has said that buffer is designed to carry the near-term sites through lease execution. A meaningful slice of that runway arrived via a $458 million convertible note financing, notes that can convert into equity and put a dilution overhang behind the multiple. Fuel buys time. It is not revenue.

Don't Chase a $100,000 Filing

The market regime makes the timing call simple. This is the reverse of the classic contrarian setup: the AI-infrastructure complex is loved, not hated. AInvest's aggregate signal still labels the stock a Buy even as its fundamental rating trails its liquidity score, and sell-side targets sit well above the current price. The crowd wants to believe, and the insider filings are now the excuse.

The price action says the same thing. The stock is down about 30% over the past month after the Q2 drubbing — it fell by over 12% on the day of the report alone — and the recent bounce arrived on elevated retail turnover within a 52-week range stretching from $1.21 to $7.37. A bounce triggered by a $100,000 Form 4 is a news squawk, not a bear trap closing on dried-up volume. A $100,000 filing is not a reason to grab a falling knife before it has stopped falling; that is a momentum lottery ticket, not a low-risk entry.

None of this makes Keel a sell. Insiders sometimes do know things, the pipeline is real, and the balance sheet can wait out negotiations. The judgment here is about the retail reaction specifically: a $400,000 insider purchase is not a signed lease, and the market is being asked to value a negotiation-stage developer as though the AI contracts were already banked.

The risk-managed call is to wait rather than chase. If a tenant lease is signed at economics that justify the buildout, or if the financials inflect, the story changes and deserves fresh work. Until then, investors buying this stock on the strength of a couple of insider Form 4s are paying for the crowd's excitement, not for evidence. For money that wants the AI-megawatt trade, the better risk/reward sits with the operators who have already converted megawatts into signed contracts. Don't let a $400,000 filing talk you into a $2 billion bet.