Skip to content
⌘ K
Log inSign up

By Intermission· 1,004 words

ResearchEvidenceQuestion

Where will durable AI profit pools accrue as models commoditize?

Evidence(7)

  1. 2025: The State of Generative AI in the Enterprise | Menlo Ventures

    [1]

    Menlo estimates generative AI spending reached $37 billion in 2025, including $19 billion for applications and $18 billion for infrastructure. Application spending comprised $8.4 billion in horizontal AI, $7.3 billion in departmental AI, and $3.5 billion in vertical AI. Coding topped $4 billion, while 76% of use cases were purchased.

    1

    Applications captured slightly more than half of measured spend, supporting durable pools in workflow products, departmental software, vertical expertise, and distribution. Yet the estimate excludes chips, inference, serving, and embedded features, so it understates upstream monetization. Its survey-based, U.S.-only methodology and venture perspective make the dollar splits directional rather than audited market shares.

  2. Full Report: Artificial Intelligence markets | OECD

    [2]

    The OECD finds model markets dynamic: language-model developers rose from 9 to 47, while active text-to-text models rose from 22 to 453. Quality-adjusted prices fell nearly 80% between January 2024 and April 2026. It identifies concentration in chips, cloud, data, skills, and applications, reinforced by integration, bundling, and switching costs.

  3. Generative AI at Work* | The Quarterly Journal of Economics | Oxford Academic

    [3]

    The available article page contains navigation, a PDF link, and view controls, but no abstract, methods, results, tables, or figures. Consequently, the study’s sample, intervention, productivity estimates, and limitations cannot be verified from this source. The evidence boundary supports only a documentation note, not substantive conclusions about AI profit pools.

  4. hai.stanford.edu

    [4]

    The report records generative AI reaching nearly 53% population-level adoption, while organizational adoption reached 88% within three years. Model performance is converging, while industry produced 91.2% of notable models. Compute capacity reached 17.1 million H100 equivalents, NVIDIA accounted for over 60%, and inference can already exceed training energy within months.

  5. Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of Artificial Intelligence on Knowledge Worker Productivity and Quality | Organization Science

    [5]

    A preregistered experiment assigned 758 knowledge workers to no AI, GPT-4, or GPT-4 with prompting. Across 18 tasks inside AI’s frontier, users completed 12.2% more tasks and worked 25.1% faster with higher quality. On a task beyond the frontier, AI users produced correct solutions 19% less often than the controls.

  6. nvda-20260125

    [6]

    NVIDIA reported fiscal 2026 revenue of $215.9 billion, up 65%, while Data Center revenue grew 68%. Its platform combines GPUs, networking, systems, CUDA, libraries, and enterprise software. Commitments reached $95.2 billion and cloud-service commitments $27 billion. Two direct customers represented 22% and 14% of total revenue during fiscal 2026, respectively.

  7. Partnerships Between Cloud Service Providers and AI Developers

    [7]

    The FTC examined three cloud–AI partnerships, drawing on information through September 2024. It reports equity, revenue sharing, cloud-spend commitments, exclusivity, discounted compute, shared data, talent, chip co-development, and product integration. It highlights input access, switching costs, and privileged information as competition risks, while disclaiming formal legal conclusions about the partnerships.

Sources

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.

Comments

LatestPopular
Write a comment
Loading comments…

Request a Thesis

Tell us what you’d like Roadstar to investigate.

New question

What would you like to know?

Context guides the research and is not shown on the finished page.