Vertex animation texture (VAT)
A technique that bakes an animation's per-frame vertex positions into a texture, so a vertex shader can replay it with no skeleton — cheap for huge numbers of instances.
Instead of computing skinning every frame, the whole animation is precomputed and stored: rows of the texture are vertices, columns are frames, and each pixel carries a position (or offset) and a normal. The vertex shader samples the vertex's row and the current time's column and displaces the base mesh accordingly. Combined with GPU instancing it renders entire crowds, insect swarms or shattering geometry in a single draw call.
The cost: no runtime blending or retargeting — the animation is baked in — memory grows with vertices times frames, and smooth results need either many frames or interpolation between columns. It gained ground once GPUs had texture bandwidth to spare while the CPU had become the bottleneck for skinning thousands of characters.