BRENNEN

Industries / Technology / Electronics and semiconductors / Electronics distribution

Design support

Design support is the engineering help a distributor gives a customer's designers, so that the parts it sells are chosen for a new product in the first place.

5

Milestones achieved, from 1976 to 2021.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1976

    Chips sold with tools

    A microprocessor was of little use without software and test gear to program it, so distributors began selling those too. In July 1976 the trade magazine Electronic Design reported that Ryan-McFarland's Basic compiler for Intel's 8080 microprocessor was available from the distributor Hamilton/Avnet for $325; Avnet's own history says it placed the first distributor order with Intel, in 1969, and that Intel later let it sell development and demonstration tools to engineers.

    Lessononce parts are programmable, selling them includes helping people design with them.

    worldradiohistory.commy.avnet.com

    Credit

    Hamilton/Avnet; Intel

  2. 2013

    Credit for a design win

    A distributor that gets a maker's part designed into a customer's product can register that design win with the maker and be rewarded on the orders that follow, a practice the trade calls demand creation. In TTI's MarketEYE in November 2013, Murata's North American distribution director said Murata grants such a registration for 12 months, renewable, and that these programs often break down when a product is designed in the US but built in Asia, so Murata had begun transferring design wins between regions.

    Lessona reward for design work holds only if it follows the design to the factory that builds it.

    ttiinc.com

    Credit

    Murata; TTI

  3. 2018

    A distributor buys engineers

    In January 2018 Arrow Electronics agreed to buy eInfochips, a product engineering firm in Ahmedabad, India, founded in 1994 by Pratul Shroff. Arrow's chief executive said it added more than 1,500 solution architects, engineers and software developers for connected devices; Mint put the price at $300 million to $325 million, citing an unnamed banker.

    livemint.comvccircle.com

    Credit

    Arrow Electronics; eInfochips

  4. 2020

    Footprints on the part page

    A footprint is the pattern of copper pads a part is soldered to, and a symbol is how the part is drawn in a circuit diagram. In November 2020 DigiKey put more than 1.5 million SnapEDA symbols, footprints and 3D models up for download on its own part pages, in formats for Altium, KiCad, Eagle, OrCAD and other design programs, after working with SnapEDA since 2017; by September 2022 SnapEDA said its models reached more than 30 distribution partners, DigiKey and Mouser among them.

    eenewseurope.comengineering.com

    Credit

    DigiKey; SnapEDA (Natasha Baker)

  5. 2021

    Design credit under review

    In 2020 the Electronic Components Industry Association (ECIA), with two other trade groups, ERA and GEDA, ran a two-part survey of how design registration worked across the authorized sales channel, and in 2021 a work group agreed what a better process should look like. ECIA's Don Elario said the channel "needs a more efficient, smart process that improves the accuracy and speed of information."

    Lessona reward that rests on hand-filed claims is slow to check, and the industry's own body said so.

    emsnow.com

    Credit

    ECIA's Global Industry Practices Committee

Watch for

  1. AI design with live stock

    CELUS, a Munich company, says its AI design assistant takes an engineer's requirements, proposes a hardware architecture and matching reusable circuit blocks, and now shows Mouser's price, stock and order quantity for each line of the resulting parts list, with a button that moves the list into a Mouser shopping cart. Design News reported the launch with quotes from both companies and cites no outside test of the proposals.

    Achieved whenan outside group publishes how often the assistant's parts lists meet the stated requirements and can be ordered, on designs it did not pick.

    designnews.com

    Claimed by

    CELUS and Mouser Electronics, November 2026

Open

  1. 01

    Real support questions, scored

    A test of AI assistants on the questions designers send to distributors' applications engineers: how to use a part, what a datasheet line means, why the circuit around a part misbehaves. Take 300 such questions that already have an accepted answer from a staff engineer or an expert, put each to several AI models, and have two engineers score the answers blind against the accepted answer and the datasheet.

    The brief
    Credit · unclaimed
  2. 02

    Part picks checked against datasheets

    200 part requests of the kind a designer gives a distributor, such as "a 3.3 V regulator for 500 mA in a SOT-23-5 package, rated to 125 °C", put to AI assistants that can search distributors' catalogs. An answer passes only if the maker's datasheet meets every stated limit and a distributor lists the part as active and in stock on the day.

    The brief
    Credit · unclaimed

Open 01 · the brief

Real support questions, scored

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A test of AI assistants on the questions designers send to distributors' applications engineers: how to use a part, what a datasheet line means, why the circuit around a part misbehaves. Take 300 such questions that already have an accepted answer from a staff engineer or an expert, put each to several AI models, and have two engineers score the answers blind against the accepted answer and the datasheet.

Why it is still open

Published electronics tests for AI models are built from curated problems, such as EEE-Bench's 2,860, or from set tasks, such as TMIQ's instrument-command matching, not from real customer questions. DigiKey answers such questions by phone, chat and its public TechForum, where one of its applications engineers wrote in 2025 that he had never visited a customer.

arxiv.orgarxiv.orgforum.digikey.com

Built with what exists

DigiKey's TechForum, which has sections in Japanese, Korean, German, Chinese, Hebrew and Spanish, and Electrical Engineering Stack Exchange, checking each site's terms before copying any text; makers' datasheets; AI models with web search, such as Claude Fable 5.2 (see the Note "Three working devices"); a scoring rule that asks whether the answer reaches the asker's real goal, as Ai2's XYBench does (see the Note "Answering the question behind the question").

  1. First step

    Pull 50 solved topics from three TechForum categories (Semiconductor, Power and Passives) and write down the accepted answer to each.

  2. Done looks like

    300 questions in at least three languages, each linked to its accepted answer; for each model, the share of answers both engineers rate correct and the share that cite the right datasheet section; all questions, answers and scores published.

Open 02 · the brief

Part picks checked against datasheets

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

200 part requests of the kind a designer gives a distributor, such as "a 3.3 V regulator for 500 mA in a SOT-23-5 package, rated to 125 °C", put to AI assistants that can search distributors' catalogs. An answer passes only if the maker's datasheet meets every stated limit and a distributor lists the part as active and in stock on the day.

Why it is still open

HWE-Bench (March 2026) gives models a fixed set of 2,914 datasheets and scores whole schematics, and its best model passed 8.15%; it does not test picking from a live catalog of millions of parts. A 2026 research workflow chooses power-converter parts through distributors' interfaces and checks them against datasheets, but reports two worked designs, not a pass rate over many requests. Anthropic says Claude Fable 5.2 ran its own checks that distributors had its parts in stock, but no outside test of AI part picking against stock and datasheets was found.

arxiv.orgarxiv.organthropic.com

Built with what exists

DigiKey's Product Information V4 interface, which searches by part number, description, maker or category, and other distributors' search tools, for stock and parameters; makers' datasheets; AI models with web search, such as Claude Fable 5.2 (see the Note "Three working devices"); a script that pulls each answer's datasheet limits into a table for a person to check.

  1. First step

    Write 30 requests from makers' public reference designs, where the maker's own chosen part is a known good answer, and confirm that DigiKey's parametric search finds at least one in-stock match for each.

  2. Done looks like

    200 requests across regulators, amplifiers, transistors, connectors and passive parts; for each tool, the pass rate, the share of answers that break a stated limit and the share not in stock, each check tied to a datasheet page; a second engineer re-scores 10% of the answers.