[KG] Blackwell — shift

{
  "entity_slug": "blackwell",
  "entity_name": "Blackwell",
  "entity_type": "product",
  "title": "Blackwell: Nvidia's AI Workhorse Faces Succession Pressure",
  "narrative": "Nvidia's Blackwell microarchitecture, powering the B100, B200, and GB200 GPUs, is the current backbone for major AI workloads—OpenAI, DeepSeek, and Perplexity AI all rely on it. Perplexity claims 3x inference throughput for 70B models, while NVFP4 integration cuts JAX training by 1.8x in MaxText. Yet the graph reveals mounting tension: Blackwell competes directly with its predecessor Hopper and the H100, but the real threat comes from Nvidia's own next-gen Vera Rubin NVL72, which cuts agentic AI cost 10x. A recent security report flags a 61% regression in confidential compute via Blackwell NVLink. With US export policy shifting—clearing H200 sales to China—and Vera Rubin looming, Blackwell's dominance window is narrowing. The question: can Blackwell maintain its deployment velocity before Nvidia cannibalizes its own crown jewel?",
  "key_points": [
    "Adopted by OpenAI, DeepSeek, Perplexity AI, and Cursor AI",
    "Competes with Hopper and H100; Vera Rubin threatens 10x cost reduction",
    "NVFP4 boosts JAX training 1.8x; Perplexity claims 3x inference throughput",
    "NVLink confidential compute regression of 61% reported",
    "US cleared H200 sales to China, altering competitive landscape"
  ],
  "angle": "shift",
  "neighborhood_size": 11,
  "generated_at": "2026-06-13T03:41:18.170997+00:00"
}