A friend asked me the other week how you would even value a company whose entire reason to exist is holding Bitcoin on its balance sheet. Not the Bitcoin itself, the equity. And the honest starting point is that you need a discount rate, and the discount rate needs a cost of equity, and the cost of equity is where most people quietly wave their hands and move on. So let me actually do the arithmetic and show where it gets uncomfortable.
The tool almost everyone reaches for is the Capital Asset Pricing Model. It says the return investors demand on a stock equals the risk-free rate plus beta times the equity risk premium. Written out, cost of equity is the risk-free rate, plus beta multiplied by the difference between the expected market return and the risk-free rate. Three inputs. That is the whole thing. The trouble with a crypto treasury company is that all three inputs get slippery, and one of them, beta, goes somewhere you have never seen it go for a normal business.
Why beta blows up here
Beta measures how much a stock moves relative to the broad market. A beta of 1 moves with the index, 0.5 is half as jumpy, 2 is twice as jumpy. Most large companies sit somewhere between roughly 0.7 and 1.5. A Bitcoin treasury company does not sit there, for two reasons that stack on top of each other.
First, the underlying asset is more volatile than equities to begin with. Bitcoin has historically swung several times harder than a stock index, so even a company that simply held coins one for one would carry a beta well above what you would estimate for an operating business. Second, and this is the part people skip, most of these companies are levered. They raise debt or issue convertible notes and use the proceeds to buy more of the asset. Leverage does not just add risk, it multiplies the equity's sensitivity to the asset. If a company holds Bitcoin worth twice its equity value and funds the rest with borrowing, the equity moves roughly twice as hard as the Bitcoin does. And the Bitcoin was already moving several times harder than the market.
Stack those and you can end up estimating a beta of 4, 5, or higher. I have seen regressions on these names come back in that range, and the number is not a mistake. It is what leveraged exposure to a volatile asset actually produces. The problem is what that does to the rest of the formula.
Running the numbers, and watching them wobble
Take a plausible set of inputs. Risk-free rate around 4 percent, using a long government bond yield. Equity risk premium historically in the neighborhood of 5 percent, though reasonable people argue for anywhere from 4 to 6. Now plug in a beta.
- Beta of 1: cost of equity is about 4 plus 1 times 5, so 9 percent.
- Beta of 2: about 4 plus 2 times 5, so 14 percent.
- Beta of 4: about 4 plus 4 times 5, so 24 percent.
- Beta of 5: about 4 plus 5 times 5, so 29 percent.
So the hurdle rate for this business, the return an equity investor should demand before the stock is worth buying, lands somewhere around a quarter or more. That is a brutal discount rate. Discount a stream of future value at 25 or 30 percent and almost nothing survives. It tells you the market is pricing this equity as something that has to compound very hard just to justify its own risk.
Now watch how much the answer depends on choices you made almost casually. Bump the equity risk premium from 5 to 6 percent with a beta of 4, and cost of equity jumps from 24 to 28 percent. Nudge beta from 4 to 5 and you add another 5 points. Change the risk-free rate by a point and everything shifts by a point. Three inputs, each defensible across a range, and the output swings by ten percentage points or more depending on which corner of the range you pick. That sensitivity is the real lesson. The formula looks precise and the estimate is anything but.
A workflow that keeps you honest
When I run this for myself, I do not compute one number. I compute a grid, because a single point estimate for this kind of company is close to false confidence. Here is the rough sequence I follow.
- Estimate a raw beta by regressing the stock's returns against a broad index over a decent window, ideally a couple of years of weekly data so you are not fooled by one wild month.
- Sanity-check it from the bottom up. Take the asset's own beta, then adjust for how much leverage the company carries. If the levered figure and the regression figure roughly agree, you have something. If they are far apart, find out why before you trust either.
- Build a table. Beta down the side, equity risk premium across the top, and fill in the cost of equity for every combination. Now you can see the range instead of a single seductive number.
- Pick the corner that matches your actual view, and write down why. If you are cautious, you use a higher beta and a higher premium. If you think the leverage will come down, you say so explicitly.
The failure mode I watch for is anchoring on the regression beta as if it were a fact. Beta estimated from past returns is backward looking, and for a company that keeps changing its leverage by raising fresh capital, the past two years may describe a different company than the one you are pricing today. A convertible note issued last quarter changes the equity's sensitivity going forward in a way the historical regression cannot see. So I treat the regression as one input into judgment, not the answer.
There is also a deeper caveat worth naming. CAPM assumes beta captures the risk that matters, and for a Bitcoin treasury company a lot of the real risk is not market beta at all. It is the specific risk that the company is forced to sell into a drawdown to meet an obligation, or that the premium the stock trades over its net asset value collapses. CAPM does not price that. So the cost of equity you get is a floor for the conversation, not the end of it.
If you take one thing from the arithmetic, let it be the sensitivity rather than any particular percentage. The cost of equity for a levered crypto treasury company is high, it is genuinely uncertain, and it moves several points on inputs you could reasonably argue either way. Report it as a range, show your inputs, and be suspicious of anyone who hands you a single clean number for a business this volatile.