Where are all of the successful accelerator participants?

Accelerators are practically everywhere in the U.S., and a look at Pitchbook data on May 13, 2024, seems to confirm that. For the five years of 2019-2023, Pitchbook tagged 18,808 different companies as having received “accelerator/incubator funding.” Conceptually, they were all startups when they received that funding and will be at widely varying degrees of evolution today (the status for 1,730 of them, for instance, was listed as “out of business”). Only 765 were classified as being one of four statuses that might be most easily considered as positive exits:  1. publicly held; 2. in IPO registration; 3. an operating subsidiary of another firm through a merger or acquisition; or 4.  integrated into another firm through a merger or acquisition and no longer tracked separately. 

Additionally, only 619 of those 765 successes are headquartered in the U.S. presently—meaning one in five are international firms with a U.S. presence.

Figure 1 shows the state-by-state distribution of the U.S. accelerator/incubator-funded successes from the past five years. Clicking on a state’s record in the circle shows the data may be broken down by city for those locations with two or more companies. All other successful accelerator participants in a state were single hits for their respective cities and are grouped as N.E.C. (not elsewhere classified) or simply in the state total.

For instance, Ohio has six companies included among the 619, but no single city is called home by two of the six, so the state entry on the graphic simply has a state total. California, on the other hand, has more than a dozen cities with two or more firms, with one-third of all companies currently listing the Golden State in their address.

Figure 1: Selected accelerator/incubator-funded firms, 2019-2023

There are several limitations to interpreting the results of identifying these firms. Most importantly, the address listed in Pitchbook for each record is not necessarily the firm’s original address. Also, the dataset does not identify the location of the accelerator. There are also opportunities for—and evidence of—coding errors in the tags based on a quick, deeper examination of a sampling of results.