Rishi Maheshwari — Dallas VC - The Inference Market Map - May 2026
FreeFreeFree tier
About Rishi Maheshwari — Dallas VC - The Inference Market Map - May 2026
A detailed market map analyzing the AI inference serving layer, published by VC Rishi Maheshwari in May 2026. The article examines approximately 80 companies organized into four horizontal sub-layers (Chips, Engines, Clusters, and Gateway) plus modality verticals (voice, image, video, embeddings, code). It includes funding figures, ARR data, and margin analysis, arguing that inference is where AI recurring revenue concentrates. Key companies discussed include NVIDIA, Cerebras, Fireworks, Baseten, fal.ai, Vapi, and others.
Key Features
Detailed analysis of four inference serving sub-layers: Chips, Engines, Clusters, Gateway
Coverage of modality verticals: voice (ElevenLabs), image (Black Forest Labs), video (Runway), embeddings (Voyage), code (Poolside)
Funding and ARR figures for key companies (e.g., Cerebras IPO at $56B, Fireworks at $15B, Baseten at $11B, fal.ai at $8B with $400M ARR)
Discussion of chip dynamics: NVIDIA dominance, AMD as second source, hyperscaler custom silicon, transformer-specific ASICs (Etched, MatX), South Korean edge players
Explanation of margin accrual across the stack, with a take per layer
Organized map with about 80 named companies and revenue estimates
Pros & Cons
Pros
- Comprehensive and detailed overview of a rapidly evolving market
- Includes specific financial data (funding, ARR, valuations) from credible sources
- Written by an active VC with deep industry knowledge
- Provides a clear hierarchical framework (layers) for understanding the stack
- Covers both major players and emerging companies across modalities
Cons
- Focused exclusively on inference serving, not covering training or other AI layers
- Opinionated analysis; margin predictions may not hold over time
- Dated as of May 2026; the market moves quickly
- Not a hands-on tool or software; it is a static report
- Assumes a certain architecture (transformer) dominance for high-conviction bets
Best For
Understanding the AI inference market landscape for investment researchCompetitive analysis for AI infrastructure startups and incumbentsStrategic planning for companies building or using inference serving platformsEducational resource for students and professionals studying AI economicsIdentifying trends in model deployment, hardware, and cloud layer
FAQ
What is the inference market map about?
It is a detailed analysis of the AI inference serving layer, organized into chips, engines, clusters, and gateways, plus modality verticals. The author examines about 80 companies and argues that inference is where the majority of AI compute spend and recurring revenue flows.
Who wrote this market map?
Rishi Maheshwari, a Dallas-based venture capitalist and investor.
Which companies are covered in the map?
Companies mentioned include NVIDIA, AMD, Cerebras, Groq, SambaNova, Fireworks, Baseten, fal.ai, Vapi, ElevenLabs, Black Forest Labs, Runway, Voyage, Poolside, Etched, MatX, Rebellions, FuriosaAI, Hailo, and many others.
What is the key argument of the article?
That inference serving is the layer where the majority of AI compute spend and margin accrues, and that companies that get this layer right will compound for the next decade.
Is this a free resource?
Yes, the article is published on Substack and is freely accessible without a paywall.