{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-11-01-019",
    "source_story_id": "tmp-brief-speculative-energy",
    "edition_id": "mp-2026-11-01-morning-0114",
    "edition_url": "https://themachinepress.com/edition/2026-11-01",
    "position": 21,
    "story_type": "ticker",
    "section": "chips-infrastructure",
    "editorial_classification": "editorial",
    "headline": "A Rejected Draft Can Turn an Energy Saving Into a Loss",
    "slug": "a-rejected-draft-can-turn-an-energy-saving-into-a-loss",
    "dek": "Production-style H100 and H200 measurements found speculative decoding saved energy only when acceptance stayed high, with the benefit shrinking as batches grew.",
    "summary": "Production-style H100 and H200 measurements found speculative decoding saved energy only when acceptance stayed high, with the benefit shrinking as batches grew.",
    "body_text": "Verification dominated dense-model energy, and large batches already amortized weight reads. At high batch size, low-acceptance or compute-bound drafting could raise energy per accepted token even while meeting latency targets.",
    "why_it_matters": "Production-style H100 and H200 measurements found speculative decoding saved energy only when acceptance stayed high, with the benefit shrinking as batches grew.",
    "limitations": [],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-11-01-019/a-rejected-draft-can-turn-an-energy-saving-into-a-loss",
    "json_url": "https://themachinepress.com/story/mp-2026-11-01-019.json",
    "first_published_at": "2026-11-01T09:00:00.000-05:00",
    "modified_at": "2026-11-01T09:00:00.000-05:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-11-01-019-001",
        "text": "Production-style H100 and H200 measurements found speculative decoding saved energy only when acceptance stayed high, with the benefit shrinking as batches grew.",
        "source_ids": [
          "source-2026-11-01-021"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-11-01-021"
    ],
    "tags": [
      "energy",
      "speculative decoding",
      "GPU serving"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-11-01-021",
      "title": "arXiv preprint 2610.40287, original v1",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.40287v1",
      "canonical_url": "https://arxiv.org/abs/2610.40287v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-28T20:00:00.000-04:00",
      "accessed_at": "2026-11-01T08:20:00.000-05:00",
      "supports_claim_ids": [
        "claim-mp-2026-11-01-019-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "A Rejected Draft Can Turn an Energy Saving Into a Loss",
    "publisher": "The Machine Press",
    "published_at": "2026-11-01T09:00:00.000-05:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-11-01-019/a-rejected-draft-can-turn-an-energy-saving-into-a-loss"
  }
}