Industries / Technology / Electronics and semiconductors / Electronics distribution
Ordering
Buyers find parts, compare prices and place orders, by catalog, by computer link, on the web and now through AI tools, while makers settle prices with the distributors that resold their parts.
6
Milestones achieved, from 1972 to 2018.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1972
Any quantity, by mail
DigiKey began in Thief River Falls, Minnesota, as a mail-order catalog business. It still calls itself a "broken pack" distributor: it takes orders of any size, including small custom jobs that many suppliers cannot handle, and by 2021 it had revenue of $4.7 billion, the Star Tribune reported.
Credit
Ronald Stordahl, DigiKey's founder
1998
Orders move to the web
RS Components, founded as Radiospares in a London garage in 1937 to supply radio repair shops, launched its first website that took orders in the UK in 1998. The group's online revenue passed £1 billion in 2017.
Credit
RS Components, now RS Group
2000
Shared order messages stall
RosettaNet, a consortium of computer and electronic parts firms, wrote standard computer-to-computer messages, called Partner Interface Processes, for steps such as purchase orders and price protection. After two years its members had agreed 15, but Computerworld found Arrow Electronics in production with just one trading partner and Cisco with two distributors, and the consortium's chief executive said the work was "harder work than people anticipated".
Lessona shared message format is quick to agree on and slow to wire into each partner's own systems.
Credit
RosettaNet; Arrow; Cisco
2015
Part search joins design tools
Altium, a maker of circuit board design software, agreed to buy Octopart, a search engine that shows many distributors' stock and prices for a part on one page. Altium said Octopart held data on more than 30 million parts and had more than 500,000 users a month. In August 2024 the chipmaker Renesas completed its purchase of Altium for about A$9.1 billion, so the search engine now belongs to a maker of parts it lists.
Credit
Octopart, led by co-founder Sam Wurzel; Altium; Renesas
2016
A new system slows deliveries
Avnet, one of the largest distributors, moved its Americas business onto a new enterprise resource planning (ERP) system, the software that takes and fills orders, in the last quarter of its 2016 fiscal year. Its annual report said Americas sales fell 6.9% that year, partly because it could not deliver to and serve customers as before after the change. The next year it chose a new global system and cut the remaining life of the Americas one to 24 months, a $16.0 million charge.
Lessonone system carries every order, so changing it at once can cost a region its deliveries.
Credit
Avnet
2018
Sales booked at shipment
Distributors often sell a maker's chips below the standard price to win business and then claim the difference from the maker, a step called "ship and debit". Under the older US rules some makers, Lattice Semiconductor among them, waited to record such sales until distributors reported reselling the parts. A new revenue rule, ASC 606, made them record the sale when the distributor takes the goods and estimate the claims still to come; Lattice changed over on 31 December 2017, the first day of its 2018 fiscal year.
Credit
Financial Accounting Standards Board (ASC 606)
Watch for
AI assistants price BOMs
Luminovo, a German maker of quoting and purchasing software, says AI assistants such as ChatGPT, Claude and Microsoft Copilot can now read a company's bills of materials (BOMs, the parts lists for a product), quotes and supplier contracts in its platform, with live prices and lead times, through the Model Context Protocol, an open way to link assistants to business software. It says the contract maker Omega EMS now quotes a complex BOM in about 20 minutes instead of half a day.
Achieved whena user or outside tester publishes a timed comparison over a stated set of BOMs, with errors counted line by line.
Claimed by
Luminovo, November 2026
Checked data behind BOM AI
Accuris, a parts-data firm, launched AI tools that score BOMs for risk and answer questions such as which parts contain PFAS chemicals. It says they rest on records for 1.3 billion electronic components "verified at 98%+ accuracy".
Achieved whenan outside audit of a random sample of its records, or of the agent's answers, confirms the stated accuracy.
Claimed by
Accuris, July 2026
Open
01
A BOM quote test
A public test of 50 bills of materials from open hardware projects. People first work out the right orderable part, packaging and price at a fixed quantity for every line, from a dated price and stock snapshot. AI agents then quote each BOM, and every line is scored.
The briefCredit · unclaimed02
Orders that follow the rules
A test in which an AI agent turns a priced BOM into orders under a written buying policy: approved makers only, minimum quantities and order multiples respected, a budget cap, and no non-cancellable lines without sign-off. Orders go to a sandbox, so nothing is bought or paid for, and each is checked against the policy.
The briefCredit · unclaimed
Open 01 · the brief
A BOM quote test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test of 50 bills of materials from open hardware projects. People first work out the right orderable part, packaging and price at a fixed quantity for every line, from a dated price and stock snapshot. AI agents then quote each BOM, and every line is scored.
Why it is still open
Vendors describe AI that cleans and prices BOMs, but publish no test set or match rate: Accuris states the accuracy of its data, not of its BOM matching, and Luminovo gives time saved. A search for a public BOM-matching benchmark found none. Free tools already price BOMs that carry exact part numbers, such as KiCost, so the test should weight the harder lines that give only a value and a package size.
Built with what exists
KiCad projects on GitHub; DigiKey's and Mouser's product APIs for the price and stock snapshot; KiCost and Kitspace's 1-click BOM as baselines; an AI agent driving a browser with Browser Use (see the Note "Deleting 80% to start over").
First step
Pick 10 KiCad projects whose BOMs mix exact part numbers with bare values, and write the answers for one of them.
Done looks like
50 BOMs, about 2,000 lines, with answers published; for each agent, the share of lines with an acceptable part, the right packaging and a price within 2% of the snapshot, set beside KiCost and one timed person on the same set.
Open 02 · the brief
Orders that follow the rules
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A test in which an AI agent turns a priced BOM into orders under a written buying policy: approved makers only, minimum quantities and order multiples respected, a budget cap, and no non-cancellable lines without sign-off. Orders go to a sandbox, so nothing is bought or paid for, and each is checked against the policy.
Why it is still open
Vendors began selling agents that draft distributors' quotes and orders from customers' emails in 2026 without published error rates, for example Proton.ai in July 2026. DigiKey already runs a sandbox for testing ordering code "while not actually making production requests". Contract2Plan, a January 2026 research system, checks purchase plans against contract terms such as minimum order quantities, but on 500 synthetic cases and without an agent placing orders with a distributor. A search found no public test of whether parts-buying agents keep to a policy.
Built with what exists
DigiKey's ordering API sandbox; a mock catalog of real offers from several sellers taken from a dated price snapshot, with planted traps (a maker not on the approved list, a full-reel minimum, an order multiple of 10, a non-cancellable line); an AI agent; a policy checker written in Python.
First step
Write a one-page buying policy and 10 trap tasks, and register a DigiKey sandbox app.
Done looks like
100 order tasks published with their traps; for each agent, the count of policy breaches and of correct orders; a pass is no breaches on all 100.