Beyond the hype. Where is the real value?

The research behind the headline

Every few years a theme arrives that makes investors forget to look at the numbers. In 2026 it is artificial intelligence and, riding alongside it, the computing infrastructure that bitcoin miners are repurposing for AI workloads. Toronto has a cluster of both. The question this brief asks is not whether the themes are real — they are — but whether the economics of two specific Canadian companies can justify what the market is asking you to pay for them.

We picked these two deliberately because they are not obviously “value” stocks. A copper producer trading below the replacement cost of its mines is the easy case. A loss-making software company and a bitcoin miner with convertible debt are the hard case: if the discipline in this publication cannot be applied to them, it is not much of a discipline.

Coveo: a cash-backed subscription business

Coveo sells AI-powered search, recommendation and generative-answering software to large enterprises, priced as a subscription. It is headquartered in Québec City and listed only on the TSX.

Read those four numbers in order and the shape of the business appears. Almost all of the revenue is recurring. The company has more than two and a half years of its current quarterly revenue sitting in cash. It is losing money, but not much relative to its scale, and the loss is an accounting loss that includes non-cash items such as stock-based compensation — the cash picture needs to be checked separately in the statement of cash flows.

The number that deserves the most attention is the last one. A net expansion rate of 99% means that the customers Coveo already had a year ago are, in aggregate, spending slightly less with it now than they were then. For a subscription software company that is the equivalent of a mine whose ore grade is drifting lower: the business can still grow, but every dollar of growth has to be won from new customers rather than harvested from existing ones.

What would move the needle, in either direction, is observable: the next two quarters of net expansion, the gap between IFRS loss and operating cash flow, and whether sales and marketing spend is producing new customers at an improving cost. None of those requires a view on the future of AI. They require reading three tables in a quarterly filing.

The U.S. investor’s wrinkle

Coveo has no U.S. exchange listing that we could confirm. U.S. brokers may offer access over the counter, typically with thinner liquidity than the TSX line and a currency conversion into Canadian dollars on the way in and out. For a position of any size, ask your broker what the OTC symbol is, what the spread looks like, and whether they charge a foreign-settlement fee.

HIVE: a tangible business with a complicated balance sheet

HIVE Digital Technologies operates bitcoin-mining data centres in Canada, Sweden and Paraguay and is building out high-performance computing capacity for AI workloads. It is listed on the TSX Venture Exchange and on Nasdaq, which makes it one of the easiest companies in this publication for a U.S. investor to hold.

This is a very different kind of business from Coveo. Revenue is real and relatively large, but it is produced by machines consuming electricity, and its size in any quarter depends on the bitcoin price and the network’s mining difficulty, neither of which the company controls. The quarter’s loss was dominated by a single item — the Swedish VAT provision — which tells you two things at once: the headline loss overstates the operating problem, and the operating business is exposed to tax and regulatory disputes in multiple jurisdictions.

The balance sheet is where a value reader should spend most of the time. US$208 million of cash sounds like a cushion until it is set against US$234.9 million of convertible loans. Convertibles sit ahead of common shareholders in a liquidation and, if converted, add to the share count. Either outcome has a cost to existing holders. The right way to think about HIVE is therefore not “a miner with a lot of cash” but “an operating business whose equity value is what is left after the convertible holders, the lessors and the other lenders are accounted for.”

Two companies, one discipline

USD. Quarter ended June 30, 2026. Revenue is not profit; both companies reported net losses.

The common thread is that neither company’s story is wrong. Enterprises do want better search. Computing capacity is scarce. The error investors make with themes like these is to let the story settle the price. The numbers above do not settle anything; they tell you precisely which questions to ask next quarter, and that is what a research brief is for.

What would change our view