AI Infrastructure Tools

3/936 → 14/1288; +0.77 percentage points of formation share.

Underneath the "AI Infrastructure Tools" label sits a clear generational split: an earlier wave of companies built the raw plumbing for multimodal data, Sieve, Eventual, DataSuite, and Tenyks all exist to wrangle terabytes of video, image, and sensor data into something trainable, while a newer wave, Instant, Beam, Unlayer, and Writesonic, assumes that data problem is basically solved and instead builds the layer apps and agents run on top of, from backends to inference clouds to visibility tracking in AI search itself. That shift matters because this formation is accelerating at 3.39x while the two categories most people would assume it maps onto, general "Infrastructure" at 1.48x and "Developer Tools" at a shrinking 0.48x, are barely moving or contracting, meaning the growth is concentrated specifically in AI-native plumbing rather than tooling broadly. The direction of travel inside the group is also telling: companies like Decohere pushing toward real-time generation without long GPU waits, and Writesonic pivoting fully into tracking brand presence across ChatGPT and Gemini, show the frontier moving from "help us build models" to "help us operate in a world where models already exist and are the interface." That pivot became viable once foundation models good enough to generate real-time video and to serve as default search surfaces actually shipped, which is why so many of these companies only make sense starting around 2022 and 2023, right as multimodal generation and LLM-based search became usable at product scale rather than research demos. The result is a formation less about training the next model and more about the operational layer sitting between finished models and paying users.

No comparable outside activity is connected yet.

These share a word with “AI Infrastructure Tools” but are tracked as separate categories. The real overlap below is the actual share of AI Infrastructure Tools’s companies also found in that category — not a guess.