SiFive dependent on few large customers

Diving deeper into

SiFive

Company Report
Average revenue per customer therefore masks wide variance, with a small number of large customers accounting for the majority of contract value.
Analyzed 4 sources

This is the core economic reality of SiFive’s model, a few big chip programs drive the business, while the long tail mainly proves ecosystem reach. SiFive sells CPU blueprints and engineering help at the start of a chip project, so contract size depends on how ambitious the customer’s design is, how much customization it needs, and whether multiple cores or product lines are covered under one agreement. That makes a hyperscaler, top semiconductor company, or automotive platform win worth far more than dozens of smaller embedded deals.

  • The revenue mix explains the concentration. About 60% of 2023 revenue came from upfront IP licenses and 30% from custom design services, both tied to active design programs rather than unit shipments. In practice, the customers writing the biggest checks are the ones building large, complex SoCs and paying for integration, verification, and optimization work before tapeout.
  • The customer count can look broad without the dollars being broad. SiFive had more than 400 design wins, over 100 paying customers, and relationships with eight of the top ten semiconductor companies, but large customers can license multiple cores across several product lines under broader agreements. That pulls contract value toward a handful of accounts even when design win count is spread widely.
  • This is typical for CPU IP vendors competing in long chip cycles. Arm still wins many buyers because its software and tooling lower execution risk, and its Flexible Access program reduces upfront license friction. At the same time, internal chip teams at Meta and others can remove entire categories of future deals from the merchant market, which makes every marquee external customer more important to hold onto.

Over time, SiFive’s path to a steadier business is to turn early license wins into long lived royalties and to add more repeatable mid sized programs in industrial, infrastructure, and automotive. If that happens, revenue concentration should ease. Until then, growth will continue to hinge on landing and expanding a small number of very large design commitments.