Baron Asset Fund's Q2 Beat Was One Stock — Its 'Cheap' AI Bets Don't All Hold Up

Baron Asset Fund posted a quarter most funds would frame as a win. Its institutional shares returned 18.98% in the three months through June 30, 2026, beating the Russell Midcap Growth Index's 14.55% by 443 basis points. Read the attribution, though, and the headline changes meaning. One holding — SpaceX, at 34.3% of net assets — contributed 13.97 percentage points of that return, roughly three-quarters of the whole quarter's gain. Strip it out and the rest of the fund added about five points, close to what the broad market did on its own.

The part of the letter that should matter to a retail investor reading it as a map of where an active manager sees opportunity is a separate, quieter claim. Again and again, the fund frames its beaten-down software and information-services holdings — CoStar, Gartner, Guidewire, Verisk — as cheap because investors wrongly fear AI will disrupt them. That pricing story is the bookend to the SpaceX bet, and it deserves the same skepticism the attribution math just got. Cheapness is not a fact; it's a claim that only survives a comparison, and the two names carrying the claim do not survive it equally.

Gartner first, because this is the case where the "misplaced fear" argument actually checks out on the data. Against the analytics complex it trades at a genuine discount: a trailing P/E near 15 and a forward P/E near 13, EV/EBITDA around 10 and price/sales under 2, versus MSCI at about 31 times trailing earnings and 24 times EBITDA, and FactSet at roughly 19 and 13. Cheap on every earnings multiple, not just on a revenue basis. And the report card is improving, which is what separates a value situation from a value trap: Gartner beat the last two quarters by a wide margin, with second-quarter EPS of $4.37 against a $3.76 consensus, and revenue of about $1.68 billion versus a $1.65 billion estimate. Margins are the strongest in this corner of the book — a roughly 19% EBITDA margin and about 25% return on invested capital. The weak spot is growth, not valuation: revenue rose less than 1% year over year. Momentum has turned back up off the lows, with the stock above both its 50- and 200-day averages even as it stays down more than a quarter on the year. So Gartner earns the label: cheap, profitable, and getting better against rising estimates.

CoStar is the other half of that same sentence in the letter, and it does not survive the same test. The stock is down about 54% this year, from a 52-week high near $91 to roughly $31. That reads like a fire sale. But "down a lot" and "cheap" are different statements, and the numbers separate them. On earnings and EBITDA, CoStar is not cheap at all: a trailing P/E of about 169, roughly 37 times EBITDA (against Verisk's 18), a P/E on forward earnings that is negative, and a return on invested capital near half a percent. Its profitability has collapsed — a roughly 2% operating margin and just $227 million of free cash flow over the trailing year on a $12.5 billion market value — because the company is pouring money into Homes.com. The discount exists almost entirely on a revenue basis, and even there the latest quarter was mixed: an earnings beat alongside a revenue miss. What fell 54% is an expensive growth asset that repriced, not a business trading below what its earnings justify.

That is the useful read of this letter. The fund managed its headline beat with one concentrated position, and it presents the "cheap on AI fear" sleeve as a single idea when the factor stack splits it in two. Gartner holds up — a low multiple backed by rising estimates, the improving report card that makes a discount real. CoStar does not — the multiple only looks low if you price it on sales, and the marginal revision is the wrong direction. In the end the letter's two claims deserve different amounts of faith. The win is concentration, not proof the whole book works, and next quarter it is the software and information-services sleeve, not the rocket, that has to prove it is actually cheap.