The Cost of Switching Between Great Companies

In September, I sold roughly $5,000 of Microsoft from my Roth IRA and bought Alphabet at $326.83. I liked both companies before the trade. I still like both now.

Microsoft was already one of my larger positions, and Alphabet was smaller. Shifting some capital made sense from a sizing perspective. Alphabet also looked attractive on its own: at the time, my five-year Base case was about $619 per share, implying roughly 13.6% annualized from that purchase price.

That was enough to make me interested. I was less thorough about the comparison with simply keeping Microsoft.

I mentioned this in my Q3 review. I don’t think the trade was necessarily wrong, but I would approach it differently today. Two good investments can both offer reasonable expected returns, and rotating between them adds a decision that needs to be justified.

A two-point difference is worth real money

Take a hypothetical $10,000 position in a business I expect to return 12% annually over five years. Another company looks capable of 14%, 15% or 17%. If those forecasts turn out exactly right, here’s where the money ends up, before dividends and taxes:

The calculation is $10,000 × (1 + annualized return)^5.

The extra $1,631 from a two-point advantage is meaningful. I also know how easily a model can move by two points when I’m estimating earnings and valuation five years out.

Suppose the 14% return assumes a 25x terminal earnings multiple. Keep the earnings forecast exactly the same and change only that multiple:

That’s calculated as 1.14 × (revised P/E ÷ 25)^(1/5) − 1. Three turns off the multiple would take the modeled advantage below the original holding’s 12% return, even before I reconsider revenue growth, margins or share dilution.

Of course, the original 12% forecast is uncertain too. This isn’t an argument that the existing holding is safer by default. I need to challenge both sets of assumptions, not just find one favorable difference between their headline returns.

Taxes change the comparison

The September switch happened inside my Roth IRA, so selling MSFT and buying GOOG did not generate a current capital-gains tax bill. That removed one cost from the decision, but it did not make the comparison itself any less important.

In a taxable account, realizing a gain can change the economics considerably. Here’s a simplified illustration using the same two return forecasts.

Start with a $10,000 holding that has a $6,000 tax basis. Selling it realizes a $4,000 gain. At an assumed 15% long-term capital-gains rate, that creates a $600 tax bill, leaving $9,400 to invest in the replacement.

Now hold for another five years. In both choices, assume the remaining investment is sold at the end and gains are taxed at the same 15% rate.

For the first row, I tax the ending gain against the original $6,000 basis. For the second, I pay $600 upfront and later tax the ending gain against the new $9,400 basis. Those details matter if the example is going to be comparable.

The replacement still wins on these assumptions. The projected advantage is $914 rather than the $1,631 suggested by looking only at pre-tax returns. That’s a smaller cushion against a forecast being wrong.

This is a federal-tax illustration, not my actual trade or personal tax estimate. It excludes state taxes, dividends, transaction costs, different tax brackets and possible changes to tax law.

What I would want to see before another rotation

For two healthy, established compounders, I generally look for about a 3–5 percentage-point advantage in expected annualized return before replacing one with the other. It’s a personal margin for estimation error, not a universal rule.

The spread needs to be built from comparable assumptions: the same horizon, consistent per-share earnings definitions, credible terminal valuations and an honest view of capital spending and dilution. I’d also check how the change affects concentration, account taxes and the downside case.

A nominal 17% Base case may be less attractive than a 12% one if it relies on more optimistic growth or a terminal multiple that’s hard to defend. Equally, a genuine improvement in the business or a serious deterioration in the original thesis may justify a change without waiting for a neat numerical spread.

Sometimes I don’t need to sell anything. New contributions can build the smaller position while I leave the original holding alone. That option was available in principle here, even if I chose to make the Roth rotation at the time.

For my next switch, I want the two company models and an index alternative in the same comparison before the order is placed. I want to know how sensitive the conclusion is to the inputs I trust least.

The September trade may work out very well. I just didn’t establish a strong enough comparison beforehand to know whether the switch improved my expected outcome. I can keep liking Alphabet and Microsoft while being more careful about that process next time.

The September GOOG purchase price and $619 Base case are historical inputs, not October 16 valuations. All five-year examples above are illustrative and exclude dividends. Values are rounded. The taxable illustration assumes a flat 15% rate on realized gains when switching and at the end of year five.