←SILCSilicomOur view: UnattractiveFull analysisעברית
Updated 16 Nov 2026
The company · 1 · Who is this company?

Silicom makes custom network cards for servers

Network cards

built around a bought-in chip

Its own software

written for each card

Subcontractors

build them

Silicom designs custom network cards the chip giants skip. The customer then tests each card for a year or more before ordering in volume.

The company · 2 · What exactly does it sell?

America is four-fifths of sales

Each $1 of sales, first half of 2026
82%8%6%4%
United StatesEuropeAsia-PacificIsrael

Silicom does not split sales by product, only by where customers are.

The company · 3 · How does it make money?

Paid per card, after long testing

$96M to $98Mfull-year sales forecast
Over 400designs customers picked, still live
$45.0Mcash and bonds, down $10M

A fixed engineering payroll of about $20M a year means recovering sales add profit quickly.

Is it a good business? · 4 · What gives it an edge?

A year of testing locks each card in

12 months or more

or more of customer testing per design

Over 400

designs customers picked, still live

Nearly $10M

one existing customer's expected 2027 spend, tripled by a new win

Replacing an approved card means repeating that test at a cost to the customer's own schedule.

Is it a good business? · 5 · What is the catch?

Chip giants sell the same job their way

NVIDIABlueField cards inside standard servers
Broadcomchips that let buyers build their own
Marvellanother chip supplier

Listed rivals: NVDA · AVGO · MRVL

Large customers can also design the card themselves.

Is it a good business? · 6 · What does its record show?

Sales halved and profit went negative

Sales, $M
128.5
150.6
124.1
58.1
61.9
Dec 2021Dec 2022Dec 2023Dec 2024Dec 2025
Sales$128M → $61.9Msales, 2021 to 2025
Profit10% → −20%operating profit per $1
Growth+6.6%sales growth in 2025

Engineering costs did not move as sales fell, so recovering sales widen profit per dollar.

Is it a good business? · 7 · How much room is there to grow?

A growing market, an unmeasured niche

$848Msmart network cards sold in 2024
$270Mthe same market in 2020
26%a year of growth

Nobody splits out the custom slice Silicom says the giants leave to it.

Is it a good business? · 8 · Who is in charge?

A family-run company with insiders selling

2022Liron Eizenman took charge
About 5%owned by insiders, mostly the chairman
15insider sales in a year, none under a plan, no buys

His father co-founded Silicom and chairs the board, and the finance chief has served since 2005.

The story right now · 9 · Why did the shares nearly triple?

Buyers bet on AI network cards

+197%the shares over a year, to late September
$3M to $4MAI sales this year

AI servers move huge amounts of data, so each needs fast network cards. Silicom hopes to sell custom ones, but its AI sales are still small and no AI deal is firm yet.

The story right now · 10 · Why do wins take so long?

A design a customer picks is not yet an order

  1. Design a cardFor one customer.
  2. A year of testingOr more.
  3. Volume ordersIf it passes.
  4. No firm AI orderSo far.

We would be early rather than wrong if the designs customers picked in 2026 become firm orders through 2027.

The story right now · 11 · What changed lately?

Sales up 71% and a small adjusted profit

$26.7Msales, up 71%
$96M to $98Mraised full-year forecast
$0.3Mstill lost on the reported measure
4 quartersin a row of faster growth

Solid quarter

The release gave no AI sales figure or named customer.

The story right now · 12 · What is everyone arguing about?

Free AI upside, or a small AI line?

Many investors think

A cheap turnaround with free AI upside

We think

AI is about 4% of revenue, with no firm win

The release calls AI a game-changer; on the call the number is $3M to $4M.

The price and our view · 13 · What is a share worth?

No single value here: it is priced as a bet

1.3×if it works
88%kept if it fails

Multiples of the company's value when the bet was last struck, 6 Nov 2026.

We put no single value on a share here, because the business does not yet earn a lasting profit. Instead we ask what the shares pay if the plan works, and what is left if it fails.

The price and our view · 14 · How could we be wrong?

Wins that may never become orders

Wins that never become ordersBiggest

No firm AI deal.

Still losing moneyShowing now

On the reported measure.

Customers who vanishWatch

Sales halved in 2024.

The first AI orderUnknown

Size and term withheld.

The niche's true size is unmeasured, and insiders have sold.

The price and our view · 15 · How will we know?

February must name an AI customer

AI sales near $10M forecastWe turn positive
A named customer and termGood sign
Wins stay unconvertedBad sign
FEB 2027-

Fourth-quarter and full-year 2026 results, February. Companies report every three months; each report is a test of the story.

Our view moves on these business events, not on a price line, since we put no single value on a share here.

The price and our view · 16 · So, what do we think?

A real recovery, priced as if the AI story is proven

UnattractiveMedium conviction

Silicom designs custom network cards for servers, and growth has sped up four quarters running, though it still loses money. The shares nearly tripled on AI cards worth only a few million dollars this year, with no firm deal.

↑ ↓ to move · C for contents