▶ See it in 3D The Memory Stack region
Ordinary server memory — slower than HBM but far bigger and much cheaper. Made by the same three giants, plus a few firms (Rambus, Montage) that sell the little buffer chips every memory stick needs.
The CPU's main memory. 8–20× less bandwidth than HBM but 10–40× the capacity at a fraction of the cost. The GPU reaches it only across PCIe.
Same three DRAM makers, plus the interface layer: Rambus and Montage sell the RCD/MRDIMM buffer chips that scale with every server. A commodity-cycle business with an AI-content kicker.
Commodity by design — dozens of fabs, a deep spot market — but the cycle has flipped: HBM is eating the same wafers, so conventional DRAM went into outright shortage in 2026. Contract prices rose ~90–95% in Q1 2026 and another ~60% in Q2; DDR4 spot hit a record $42.50 in August. Nothing gates AI here structurally, but for the memory makers' P&L this is now the bigger story — see The DRAM Cycle.

#1 HBM maker (~62% share). Qualified HBM4 at NVIDIA first and holds an estimated 60–70% of Rubin's HBM4 allocation, with a TSMC-made logic base die. Sold out through 2027; its CEO calls 2027 potentially the tightest supply year the industry has seen. The bellwether of the whole cycle.
#2 in HBM. Began HBM4 mass production in Feb 2026 and is qualified at NVIDIA — with an in-house 4nm base die, the one turnkey memory+logic+packaging offer. P5 Pyeongtaek pulled forward to late 2027; ~$73B cumulative capex through 2026.
#3 HBM, the only US-HQ DRAM maker. Qualified on Rubin HBM4; sold out through 2027 and meeting only ~50–65% of what customers ask for. New Idaho capacity doesn't land until 2027–28 — so the squeeze is structural, not seasonal.
Memory-interface IP + chips (CXL, MRDIMM, PCIe). Licensing royalties scale with every DDR5/CXL device.
CXL memory controllers + DDR5 RCD/MRDIMM buffers. China's CXL silicon play.