Back
Menlo Ventures - 2025 Mid-Year LLM Market Update
webA venture capital market report useful for understanding the commercial trajectory of frontier LLMs; relevant for tracking capability diffusion, deployment trends, and competitive dynamics among major AI labs, though not focused on safety research.
Metadata
Importance: 42/100blog postanalysis
Summary
Menlo Ventures' mid-2025 report analyzes the enterprise LLM market, finding that Anthropic has surpassed OpenAI in enterprise usage, model API spending has more than doubled to $8.4B, and AI spend is shifting from training to inference. Key findings include that enterprises switch models for performance rather than price, and open-source adoption has plateaued.
Key Points
- •Enterprise LLM API spending more than doubled in 6 months, from $3.5B to $8.4B, signaling rapid scaling of production inference deployments.
- •Anthropic has surpassed OpenAI as the leading enterprise LLM provider by usage, representing a major market share shift.
- •Open-source model adoption in enterprises has flattened, as frontier closed-source models maintain a performance edge.
- •Enterprises prioritize model performance over cost when switching providers, challenging assumptions about commoditization.
- •AI compute spend is visibly shifting from pre-training/development to inference, reflecting maturation of production AI workloads.
Cited by 2 pages
| Page | Type | Quality |
|---|---|---|
| Open vs Closed Source AI | Crux | 60.0 |
| Anthropic Valuation Analysis | Analysis | 72.0 |
3 FactBase facts citing this source
Cached Content Preview
HTTP 200Fetched Apr 7, 202613 KB
2025 Mid-Year LLM Market Update: Foundation Model Landscape + Economics | Menlo Ventures
Skip to Main Content
All Perspectives
Perspective Portfolio Report
2025 Mid-Year LLM Market Update: Foundation Model Landscape + Economics
A new enterprise LLM leader has emerged as usage and spend surge.
July 31, 2025
Tim Tully , Joff Redfern , Deedy Das , Derek Xiao
Facebook
Linkedin
Twitter
Envelope
Copy link
Copied to clipboard!
Table of Contents
Foundation models are not just powering generative AI. They are shaping the future of computing. As their capabilities and economics evolve, so will the systems, applications, and industries built on top of them.
When we released Menlo Ventures’ 2024: The State of Generative AI in the Enterprise report last November, several critical questions about this foundational layer remained unanswered:
Would demand for LLM APIs keep pace with the growth of consumer applications?
How smart will these models get, and how fast will they get there?
Would open-source models catch up to closed-source frontier models in performance, and if so, how would that impact enterprise adoption?
And most importantly, where might long-term value accrue?
Six months later, the data tells a clearer story:
Model API spending has more than doubled in this brief period—jumping from $3.5 billion (of a total $13.8 billion generative AI spend we estimated last year) to $8.4 billion . 1 Enterprises are increasing production inference rather than just model development, marking a shift from previous years.
Code generation has become AI’s first breakout use case. Beyond pre-training, foundation models are now scaling along a second axis: reinforcement learning with verifiers. And while open-source continues to advance, the slowdown in frontier breakthroughs from Western labs has tempered what had previously been a rise in enterprise adoption. As a result, enterprise dollars are now consolidating around a few high-performing, closed-source models, giving us a new market leader in Anthropic *.
To capture the state of the current LLM market, we surveyed over 150 technical leaders 2 across startups and enterprises on the modern AI stack’s foundation layer: who’s gaining share, what’s running in production, and the selection criteria shaping the entire stack.
Here is what we learned:
Anthropic Surpasses OpenAI in Enterprise Usage
By the end of 2023, OpenAI commanded 50% of the enterprise LLM market, but its early lead has eroded. Today, it captures just 25% 3 of enterprise usage—half of what it held two years ago.
Anthropic is the ne
... (truncated, 13 KB total)Resource ID:
3cb52ce5c601c885 | Stable ID: sid_luN9jW1GXz