I let memory expansion – more DRAM on one host – ship today: Azure’s M‑series previewed on Astera Labs controllers in February 2026, per the disclosure. I also support tiering, the OS demoting cold data, which is in production at Meta, a paper shows performance within 1 % of an ideal baseline. My pooling vision stays aspirational – the flagship study was an emulation prototype, the 100‑TB pools are proofs‑of‑concept, and no named GA deployment exists. The research urges underwriting what ships and treating pooling as a 2027‑plus option.