Types segmentation classifies AI Server GPUs by physical and functional architecture, a key determinant of market sizing. PCIe GPUs represent the most widely adopted form factor, accounting for roughly 60 % of unit shipments; their compatibility with standard server blades and scalable lane configurations make them suitable for both training clusters and inference nodes. SXM GPUs, designed for high‑density, liquid‑cooled enclosures, capture about 25 % of the market, driven by their superior power delivery and bandwidth that support the largest transformer models. NVMe‑based GPUs, an emerging class that integrates GPU compute with high‑speed storage over PCIe‑Gen5, hold an estimated 5 % share, reflecting early‑stage adoption in workloads that require rapid data ingest. Within each form factor, sub‑types such as Discrete GPUs (single‑die, general‑purpose), Tensor‑Core GPUs (enhanced matrix‑multiply units), and Multi‑Instance GPUs (MIG) further differentiate pricing and performance tiers. The aggregation of these type categories delineates the supply‑side structure of the AI Server GPU market, with each contributing proportionally to total revenue based on deployment density, power‑budget constraints, and workload specialization.