HIVE's $84.9M Quarter Came With a $31.2M Impairment

HIVE Digital booked $84.9 million in revenue for the fiscal second quarter ended September 30, 2026. Its BUZZ GPU division chipped in roughly $8.8 million of that, a 70.7% jump from the same period a year earlier. And the company posted a net loss of about $78 million.

Three numbers, one quarter, and they tell you almost everything about what's happening to the bitcoin mining business right now.

Because this is the trade miners keep making. they've electricity contracts, land, and buildings full of transformers. Bitcoin mining monetized that in one way. Renting out GPUs monetizes it in another. HIVE is betting it can run both engines at once, and the latest filing is the first honest look at how that bet is going.

The Quarter in Numbers

Here's what the filing actually says: $84.9 million in revenue, $8.8 million of it from BUZZ, and $139 million in cash sitting on the balance sheet. That cash figure matters more than it usually would, because it's what buys time for a GPU build-out that hasn't paid for itself yet.

The key detail is the ratio. BUZZ is growing fast, and $8.8 million is still a little over 10% of company revenue. So the AI arm is real. It's just not yet the thing that carries the quarter.

That's not a criticism, it's arithmetic. Hashing sells at a price set by a global network. GPU time sells at a price set by whoever signed the contract. One is liquid and brutal. The other is sticky, but it demands customer relationships, network engineering, and staff who know how to keep a cluster cool and a client happy. Those aren't skills a mining farm picks up for free.

Then there's the loss. Roughly $78 million on the bottom line, and $31.2 million of it's a mining-equipment impairment.

Impairments aren't Noise

Investors have gotten very good at waving off impairments as noncash. Technically, that's correct. Nobody wrote a $31.2 million check this quarter. The charge reflects a revised view of what those machines are worth in future earnings.

But noncash isn't the same as meaningless. The write-down is management telling you, in the dead language accounting requires, that expected economics on part of the fleet have fallen. Hardware value depends on electricity costs, network difficulty, machine efficiency, and the price of bitcoin. Move any one of those and a rig bought two years ago might not justify the value it carried on the books.

So what does it signal? That older machines are under pressure. That's the message, and it's a message every large miner will eventually have to deliver or explain away.

The precedent here's important. Under disclosure rules, companies get real latitude on impairment timing, and auditors sign off on the assumptions behind it. Which means a write-down is never just a write-down. It's a statement about how management sees the next few years of difficulty and power costs, delivered in the least conversational format available.

HIVE did the conservative thing, taking the charge now rather than stretching carrying values across future quarters. Cautious readers should note that and move on.

Reading between the lines, though, the dual-engine pitch has a cost side that never shows up in the revenue line. GPU capacity doesn't arrive free. It comes with capex, with power agreements that need renegotiating, and with tenants who can walk if uptime slips.

What Gets Measured Next

So what should you actually track when the next quarter lands? Three things, and none of them is total revenue.

First, BUZZ margin. Growing a division at 70% is far less impressive if every dollar of that revenue costs a dollar ten to serve. Second, capex and contractual obligations. The $139 million in cash is a snapshot, not a cushion. If construction commitments run past it, the number shrinks quickly. Third, mining cost per coin against a difficulty curve that only bends one direction.

And here's the question that should bug anyone holding miner equity in 2026. If AI compute is genuinely the better business, why does the industry still measure itself against hash rate? Because the transition is expensive and slow, and because the AI side rewards operators who can sell a contract, not just run a machine.

The structural read is that miners are quietly turning into power companies with compute attached. TeraWulf has pushed AI leasing into the large majority of its revenue. Riot has sold bitcoin to fund data center contracts. HIVE is running a version of the same play with a smaller AI book and a bigger impairment charge.

Who wins? Operators with cheap, contracted power and existing tenant relationships. Who loses? Anyone who buys GPU capacity before signing a customer, and any shareholder who reads a 70% growth rate without reading the loss underneath it.

The takeaway is narrow and it's uncomfortable. HIVE's AI revenue is growing for real. But a $78 million loss, a $31.2 million impairment, and an AI division at roughly a tenth of revenue isn't a finished turnaround. It's a company mid-transition, and the accounting is being honest about that before the income statement is.

Watch the margins next quarter. Growth is the easy part to print.