{"data":{"slug":"nvidia-b200","etype":"chip","name":"NVIDIA B200","cat":"GPU","subtitle":"Blackwell: two reticle-limit dies fused into one GPU — the 2025 workhorse.","summary":"Blackwell hit the reticle limit and kept going: the B200 joins two 104-billion-transistor dies with a 10 TB/s die-to-die link so they behave as one 208-billion-transistor GPU. It introduced FP4 precision for inference, moved to CoWoS-L packaging, and in its rack-scale GB200 form made liquid cooling mandatory at the frontier. Every major cloud deployed it through 2025.","attrs":[{"g":"Overview","l":"Designer","v":"NVIDIA"},{"g":"Overview","l":"Architecture","v":"Blackwell"},{"g":"Overview","l":"Announced","v":"March 2024 (GTC); volume through 2025"},{"g":"Overview","l":"Status","v":"Shipping"},{"b":"Vendor spec","g":"Manufacturing","l":"Foundry / process","v":"TSMC 4NP"},{"b":"Vendor spec","g":"Manufacturing","l":"Transistors","v":"208 billion (2 × 104B dies)"},{"b":"Vendor spec","g":"Manufacturing","l":"Die-to-die link","v":"NV-HBI — 10 TB/s"},{"b":"Vendor spec","g":"Manufacturing","l":"Packaging","v":"CoWoS-L, 8 HBM3E stacks"},{"b":"Vendor spec","g":"Memory","l":"Memory","v":"192 GB HBM3E (8 × 24 GB stacks)"},{"b":"Vendor spec","g":"Memory","l":"Memory bandwidth","u":"TB/s","v":"8"},{"b":"Vendor spec","g":"Compute","l":"BF16 tensor (dense)","n":"NVIDIA datasheets often headline throughput with structured sparsity; figures here are dense unless noted.","u":"PFLOPS","v":"2.25"},{"b":"Vendor spec","g":"Compute","l":"FP8 tensor (dense)","u":"PFLOPS","v":"4.5"},{"b":"Vendor spec","g":"Compute","l":"FP4 tensor (dense)","n":"FP4 with second-generation Transformer Engine is new in Blackwell","u":"PFLOPS","v":"9"},{"b":"Vendor spec","g":"Power","l":"TDP","u":"W","v":"up to ~1,000 (HGX, configurable); 1,200 in GB200 configurations"},{"g":"Physical","l":"Form factors","v":"SXM (HGX B200), GB200 superchip"},{"b":"Vendor spec","g":"Physical","l":"Interconnect","v":"NVLink 5 — 1.8 TB/s"},{"b":"Reported","g":"Price","l":"Price","v":"~$30,000–40,000 per GPU indicated by NVIDIA leadership in 2024"}],"kf":{"node":"TSMC 4NP","year":2024,"class":"GPU","tdp_w":1000,"mem_bw":8.0,"memory":"192 GB HBM3E","designer":"NVIDIA"},"aliases":["B200","Blackwell"],"ticker":null,"exchange":null,"hq":null,"founded":null,"website":null,"status":"Shipping","region":null,"asof":"2026-01-15","sources":[{"url":"https://www.nvidia.com/en-us/data-center/technologies/blackwell-architecture/","label":"NVIDIA Blackwell page"},{"url":"https://www.nvidia.com/en-us/data-center/dgx-b200/","label":"DGX B200 datasheet"}]},"meta":{"found":true,"count":1,"served_from":"site cache (no live upstream calls)","as_of":"2026-10-14T06:50:45+00:00","disclaimer":"Research and information only — not investment advice. Structured records carry basis labels (vendor spec / reported / estimate); market quotes are delayed/indicative.","attribution":"Records compiled from vendor documentation, official filings and reputable public reporting; market quotes via financial data providers. See /data-sources."}}