For each funding round’s value R, distribute the sum evenly among all participating investors N, and then sum these values over all rounds to obtain the total investment F by each funder j in the target company. Estimate.
A limitation of this methodology is that it assumes that the contributions of investors participating in the round are equal. This could lead to an underestimation of the influence of the lead investor and an overestimation of the contribution of subsequent investors. As a result, the funding structure is assumed to be more polarized than this model suggests. Common estimates suggest that lead investors may commit up to 20-50 percent of a given round. For visual clarity, this visualization also excludes investments from regions with only a small stake in EU companies.
For computing infrastructure, we tracked startup computing providers from public sources to see what kind of hardware infrastructure each computing provider uses. The main limitation of this methodology is the lack of information about the relative weight of different chips in a computing cluster when there are multiple providers. Additionally, companies may have additional alternative computing providers for which information is not available. Given the path dependencies between hardware and large-scale AI models, these potentially missing compute providers are expected to use the same hardware as other compute providers in the target enterprise. Masu.
Funders were geographically located based on the location of their main headquarters, and classification of investor types was based on contextual information about the nature of the company.
Source data EU AI startup market analysis
Acquired by AMD in July 2024
