{
  "docId": "019dd923-5e88-73ef-bd5d-cdf676789001",
  "docSlug": "1873217bba818632",
  "documentTitle": "2025 Bond Cap Artificial Intelligence AI 2025",
  "authorId": "BondCap",
  "authorName": "BOND",
  "documentKindSlug": "consulting-deck",
  "documentKindLabel": "Consulting deck",
  "sourceTypeSlug": "vc_research",
  "sourceTypeLabel": "VC research",
  "presentationDate": null,
  "orientation": "landscape",
  "aspectRatio": 1.333,
  "pageNumber": 185,
  "pageCount": 340,
  "prevPage": 184,
  "nextPage": 186,
  "slideType": "key_messages",
  "function": "summarize",
  "density": "overcrowded",
  "nDataPoints": 2,
  "notes": "The slide contrasts the capital-intensive nature of training LLMs with the rapid commoditization and efficiency gains in inference.",
  "elementsJson": [
    "paragraph",
    "footnote"
  ],
  "metadataConfidence": 0.95,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5d-cdf676789001/185",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5d-cdf676789001",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5d-cdf676789001.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5d-cdf676789001#slide-185",
  "components": [
    {
      "bbox": {
        "h": 0.05,
        "w": 0.7,
        "x": 0.15,
        "y": 0.75
      },
      "kind": "callout",
      "text": "What used to cost dollars can now cost pennies. And what cost pennies may soon cost fractions of a cent...",
      "attrs": null,
      "subkind": "primary",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "5278fcc9-dc0f-4972-82a7-71b30c65aaaa",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "callout",
      "text": "Inference represents a new cost curve, and – unlike training costs – it’s arcing down, not up.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-4e54-704a-a059-8783ed9b920a",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "cost-per-token",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-4e54-704a-a059-8a9a9c78ea5c",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.12,
        "w": 0.7,
        "x": 0.15,
        "y": 0.5
      },
      "kind": "paragraph",
      "text": "At the same time, the cost of applying/using these models – known as inference – is falling quickly. Hardware is improving – for example, NVIDIA’s 2024 Blackwell GPU consumes 105,000x less energy per token than its 2014 Kepler GPU predecessor. Couple that with breakthroughs in models’ algorithmic efficiency, and the cost of inference is plummeting.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "05fbd278-930f-4cf3-9640-1640990e6485",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.1,
        "w": 0.7,
        "x": 0.15,
        "y": 0.27
      },
      "kind": "paragraph",
      "text": "Training the most powerful large language models (LLMs) has become one of the most expensive / capital-intensive efforts in human history. As the frontier of performance pushes toward ever-larger parameter counts and more complex architectures, model training costs are rising into the billions of dollars.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "22bd87d4-f070-413b-b6d6-f1e969b149f7",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.7,
        "x": 0.15,
        "y": 0.41
      },
      "kind": "paragraph",
      "text": "Ironically, this race to build the most capable general-purpose models may be accelerating commoditization and driving diminishing returns, as output quality converges across players and differentiation becomes harder to sustain.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "334d053e-c0b6-4aaf-a73c-30efce1e2a67",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.08,
        "w": 0.7,
        "x": 0.15,
        "y": 0.65
      },
      "kind": "paragraph",
      "text": "Inference represents a new cost curve, and – unlike training costs – it’s arcing down, not up. As inference becomes cheaper and more efficient, the competitive pressure amongst LLM providers increases – not on accuracy alone, but also on latency, uptime, and cost-per-token*.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "eb5a0abf-ad70-4c4b-952e-0cb1d693a6e9",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.7,
        "x": 0.15,
        "y": 0.19
      },
      "kind": "paragraph",
      "text": "To understand where AI model economics may be heading, one can look at the mounting tension between capabilities and costs.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "f6f8474e-305e-4ef9-aaae-b606db6d6baa",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.9,
        "x": 0.05,
        "y": 0.83
      },
      "kind": "source-note",
      "text": "*Cost-per-token = The expense incurred for processing or generating a single token (a word, sub-word, or character) during the operation of a language model. It is a key metric used to evaluate the computational efficiency and cost-effectiveness of deploying AI models, particularly in applications like natural language processing.",
      "attrs": {
        "numbered": true
      },
      "subkind": null,
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "1470e925-ae31-46a8-981b-c592802aeb50",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [
    {
      "name": "Pyramid Principle",
      "slug": "pyramid-principle",
      "agent": "Architect",
      "layer": "slide",
      "matchId": "d013541c-f6ae-4a01-bdb4-bd724a21f5b6",
      "evidence": "The slide presents a series of paragraphs that break down the relationship between capabilities and costs, indicating a structured argument.",
      "confidence": 0.7
    }
  ],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 340,
      "from": 31,
      "beatId": "08b5f3f3-e96d-4460-a176-750933765cf7",
      "arcName": "Monroe's Motivated Sequence",
      "arcSlug": "monroes-sequence",
      "beatName": "Action",
      "beatSlug": "monroes-sequence-action",
      "evidence": "The deck concludes by encouraging action, providing guidance on navigating the AI landscape.",
      "position": 4,
      "confidence": 0.8,
      "parentBeatName": "Resolution",
      "parentBeatSlug": "resolution"
    }
  ],
  "loops": [],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}