Silicom’s first AI inference order signals transition from testing to revenue, boosting prospects for hyperscaler deployment

Silicom Ltd secures its first production order for AI inference products, indicating a shift from prototype testing to commercial deployment and potential rapid growth in AI infrastructure revenue, with early signs pointing to major hyperscaler involvement.

Silicom Ltd is trying to turn a niche networking business into a bigger AI infrastructure play, and the latest signal is its first production order for an AI inference product. According to company statements relayed by industry outlets, deliveries are due in 2026 and expected AI inference revenue for that year has moved into the multi-million-dollar range, a notable step beyond earlier testing and evaluation work.

The order matters because it suggests the company has crossed from proving its technology to getting paid to ship it. Silicom says the customer has moved ahead of the original timetable, while chief executive Liron Eizenman said the milestone validates the product’s ability to tackle performance bottlenecks in advanced AI inference systems. The company has not disclosed the customer or the size of the contract.

AI inference, the process of using trained models to make predictions or decisions, is becoming an increasingly important part of the broader AI build-out. That has given Silicom a possible second growth engine alongside its core networking and data infrastructure business. The company is also said to be working on additional inference-related engagements, with expectations that revenue from the segment could accelerate in 2027 and beyond.

There is also a potentially larger prize if early trials scale up. One report said a customer targeting major hyperscalers has selected Silicom’s inference-specific solution for proof-of-concept work scheduled for the second half of 2026, after placing initial orders for the first half of the year. If that effort succeeds, the customer could move to a first full deployment requiring tens of thousands of units, each priced at a multi-thousand-dollar average selling price. For now, that remains a possibility rather than a forecast, but it helps explain why investors are watching the stock closely.