---
title: "Wafer Raises $40M Series A to Automate AI-Inference Optimization"
description: "Marathon and Chemistry co-led Wafer’s $40M Series A, with AMD Ventures joining a bet that model-specific optimization becomes an independent infrastructure layer."
date: 2026-09-01T00:00:00.000Z
tags: ["2026-vc-news", "startup-funding", "venture-capital", "artificial-intelligence", "enterprise-saas"]
source: https://venturecapitaltracker.com/2026-wafer-40m-series-a-ai-inference
---

# Wafer Raises $40M Series A to Automate AI-Inference Optimization

> Marathon and Chemistry co-led Wafer’s $40M Series A, with AMD Ventures joining a bet that model-specific optimization becomes an independent infrastructure layer.

Wafer raised a **$40 million Series A** to automate optimization of AI models for inference hardware. Marathon and Chemistry co-led; Wing Venture Capital, AMD Ventures and Outset joined, while Fifty Years and Y Combinator returned.

AMD’s participation is the strategically important fact. Wafer is pitching a software layer that can make models run efficiently across changing chips, placing it inside the contest to broaden AI infrastructure beyond Nvidia’s default stack.

## The financing

| Field         | Detail                     |
| ------------- | -------------------------- |
| Round         | **$40M Series A**          |
| Co-leads      | Marathon; Chemistry        |
| New investors | Wing; AMD Ventures; Outset |
| Returning     | Fifty Years; Y Combinator  |
| Valuation     | Not disclosed              |

## What Wafer sells

Inference optimization is usually a labor-intensive combination of quantization, compilation, kernel tuning and deployment testing. Wafer aims to automate that work so model developers can target performance, latency and cost goals without maintaining separate optimization teams for every hardware configuration.

The commercial question is whether this becomes a durable independent layer or a feature absorbed by chip vendors, cloud platforms and model-serving providers. AMD Ventures gives Wafer distribution and hardware context, but it also sharpens that platform-dependence risk.

## What is not disclosed

Wafer did not publish revenue, customer count, benchmark methodology, gross margin or valuation. Performance claims need workload-specific comparisons: model, precision, batch size, latency target, power use and chip generation can all change the result.

A $40 million Series A gives Wafer room to build engineering and go-to-market teams. The investment thesis works only if customers value cross-hardware portability enough to pay a neutral optimizer rather than accept the tools bundled with their preferred cloud or accelerator.

For the wider infrastructure wave, see [VAST’s ~$446M world-model financing](/2026-vast-rmb3b-series-b-world-models) and the [September 1 VC roundup](/2026-september-1-vc-news-ai-defense-climate).

**By:** [Venture Capital Tracker](https://venturecapitaltracker.com/editorial-policy)

**Editorial note:** AI tools assisted with research, structure, or drafting. Venture Capital Tracker retains human editorial responsibility for factual accuracy, relevance, and source quality before publication.