CoStar Is Buying Vexcel for $730M to Read Buildings From the Air

On October 27, 2026, CoStar Group, Inc. reported under Item 1.01 of a Form 8-K that two of its subsidiaries had entered into an Agreement and Plan of Merger to acquire Vexcel Group, Inc. for $730 million in cash, subject to customary working capital and other post-closing adjustments. Closing depends on the Hart-Scott-Rodino waiting period expiring, on certain foreign regulatory approvals, and on the absence of a material adverse effect at Vexcel. It does not need a CoStar shareholder vote or financing. Either side can walk if the deal has not closed by July 27, 2027, with extensions in specified cases. The press release, furnished as Exhibit 99.2 to the same filing, says CoStar will pay from cash on hand and its existing revolving credit facility and expects to close in the first quarter of 2027.

The release describes what CoStar is buying, and every figure here is a company claim rather than a PSV finding. Vexcel captures more than 60 million images a year across more than 45 countries and 320 major metros, covering 99.6 percent of the U.S. population. It refreshes major U.S. metros up to three times a year and flies disaster zones within 24 hours. It designs and manufactures the UltraCam line of aerial cameras and runs the aircraft and processing itself, delivering resolutions of 7.5 to 15 centimeters, which CoStar calls roughly 16 times finer than typical satellite imagery. Vexcel’s customers use the data for insurance underwriting, catastrophe response, infrastructure planning and mapping. CoStar says Vexcel generated more than $140 million of revenue over the last twelve months, up more than 30 percent, on recurring multi-year contracts, at an Adjusted EBITDA margin of approximately 20 percent. Vexcel will keep its brand and its chief executive, Erik Jorgensen.

The rationale is in founder and chief executive Andy Florance’s own words in the release. He describes CoStar researchers driving, since 1989, tens of millions of miles to photograph properties, then says that combining high-resolution aerial images with AI that can recognize and extract features from properties could let CoStar capture hundreds of data fields per property across the roughly half a billion properties Vexcel has captured. He calls it a new, ultra-low-cost proprietary data source. In the earnings release furnished the same day as Exhibit 99.1, he ties Vexcel and the August purchase of Zonda together as proprietary datasets that he calls real differentiators in the world of AI. CoStar also lists, among its own risk factors, its ability to extract property data from aerial imagery using AI at the anticipated scale, cost and pace.

Two consequences follow for an operator. First, more of the property data in the market’s largest commercial database may soon be inferred by a model from a photograph rather than observed by a person, and those are different kinds of facts with different error patterns. Second, concentration keeps rising. CoStar already owns CoStar, LoopNet, Apartments.com, Homes.com, STR, Ten-X, Matterport and Zonda. On the company’s rounded figures, $730 million is about 5.2 times Vexcel’s trailing revenue and about 26 times its implied Adjusted EBITDA, against roughly 4.7 and 20 times for Zonda’s $800 million on its 2025 figures; that is PSV arithmetic, not a CoStar disclosure, and the inputs are stated as more than and approximately. For scale, CoStar reported third quarter revenue of $952 million and Adjusted EBITDA of $229 million, and raised its 2026 Adjusted EBITDA guidance to $820 million to $840 million.

The tool is a provenance check on any property record that feeds an underwriting or valuation decision. Inputs are the record itself, the field list it draws on, and whatever the vendor discloses about how each field was produced: researched, reported by an owner, measured from imagery, or inferred by a model. The output is a short table that marks each decision-relevant field, such as building area, roof condition, parking count, land area or construction status, by source type, with the primary record that would confirm it: the survey, the county assessor, the rent roll, the permit file or a site visit.

The reviewer is the analyst who owns the deal file, and the approval gate is that no inferred field moves a price, a loan size or an investment committee figure until it matches a primary record. One boundary matters here. CoStar’s terms restrict using its data in outside AI tools without written permission, so this check runs on what the license allows a person to read, and counsel decides what may go anywhere else. PSV has not seen any Vexcel-derived field in a CoStar product, and nothing in the filing describes one yet.