| Episodes | 78 |
| Mentions | 100 |
| Cited here | 61 |
| First — last | #22 — #717 |
| Top guests | Jeff Keyzer, Bunnie Huang, Andrew Witte |
| Related | digi key · component sourcing · contract manufacturer · bill of materials · microcontroller |
A supply chain in electronics manufacture is the sequence of stages — raw-material extraction, refining, component fabrication, assembly and distribution — through which materials become finished products, and a disruption at any tier propagates downward to the end product.[2][631] Because inventory is held at each tier, an upstream event may not reach a buyer’s price or availability for twelve to eighteen months.[688] The chain’s arrangement is treated as a competitive asset in its own right: supplier identities and bill-of-materials costing are commonly withheld even from otherwise open hardware releases, on the basis that the sourcing arrangement rather than the schematic is what is hard to copy.[298]
Chain structure and propagation
Raw-material extraction is concentrated in a small number of countries, and a disruption at that level propagates through refiners, component makers and assemblers to the finished product.[2] Semiconductor manufacturers are not the base of the chain: they depend in turn on wafers, process chemicals and energy, so an upstream disruption reaches a designer through several intermediate layers.[631]
The propagation is delayed by inventory. A stoppage at a component factory does not appear at the buyer immediately, because pipeline and distributor inventory continues to ship; the shortage surfaces weeks or months later when that inventory drains.[482] A buyer of assembled product sits several tiers from the disruption, so an event today may not reach purchase price or availability for twelve to eighteen months, with inventory held at each tier — often around six months’ worth — setting the delay.[688]
Concentration at any tier transmits shocks. A supplier whose output is concentrated on one customer inherits that customer’s demand swings directly; suppliers whose business was around ninety percent automotive shut down when the vehicle factories stopped ordering.[536] Chains also cross borders by process stage: in North American automotive manufacture, engine blocks are cast in Canada and wiring harnesses built in Mexico, so tariff and border policy applies to a product several times before final assembly.[694]
Part identity is not stable across the chain. Corporate splits, mergers and acquisitions leave the same physical component listed under two distinct entries, and a component management system must track subsequent part merges to keep one canonical identity.[542]
Sourcing channels
Buying components through the manufacturer’s authorised distribution channel rather than open marketplaces is the basic defence against out-of-tolerance and counterfeit parts; a designer relying on 0.1 percent resistors meeting their stated grade sources them from a named manufacturer through a franchised distributor.[22] A component supplier’s institutional form is itself a sourcing criterion: a purchaser designing a part into a product assesses whether the supplier can be relied on to keep supplying it, and treats a hobby-scale vendor as a supply risk regardless of the part’s technical merit.[44]
Buying direct from a component manufacturer is not always superior to distribution. A manufacturer’s own web store may accept and charge for an order without displaying stock status, disclosing a months-long backorder only afterwards; catalogue distributors that publish live stock figures avoid this failure, which is one reason parts are often ordered through distribution rather than direct.[74] Some semiconductor vendors appoint a single authorised volume distributor even where their parts are also listed by broad-line catalogue distributors; Xilinx volume business ran through Avnet, partly to simplify export compliance.[466]
Intermediaries fill structural gaps. For small orders of low-value parts, freight and per-supplier shipping charges can exceed the cost of the goods, so a regional reseller’s economic function is consolidating multiple origins into one shipment rather than adding value to the parts themselves.[189] Display and component makers do not transact directly with small buyers at repair-shop quantities, so the independent repair trade is supplied by the overrun between a contracted build quantity and the parts shipped against it.[311] For a function whose dedicated part is obsolete, the alternatives are sourcing secondhand devices desoldered from scrap boards, whose provenance and condition are unknown, or re-engineering the function from currently available parts, typically with a higher component count.[263]
Pricing, volume and allocation
Unit price for the same commodity passive can differ by roughly two orders of magnitude between a small buyer ordering singles from a catalogue distributor and a high-volume manufacturer buying direct.[70] The gap between single-piece and volume price is steepest for parts that turn over slowly in distribution, because the distributor’s holding cost is spread over few sales; FPGAs with single-piece catalogue prices around 150 dollars are the standard example.[466] A customer’s purchase volume determines the commercial terms and the level of engineering support obtainable from chip vendors, and rising volume opens access to vendors who would not previously engage.[474]
Volume purchasing also creates arbitrage structures. Buying a module from a vendor who purchases the underlying silicon in volume gives a low-volume customer access to pricing they could not obtain themselves.[466] A platform vendor’s volume purchasing can put a complete board below the single-unit distributor price of its main chip; a Zynq-based board sold at 55 dollars against a single-piece device price of about 70 dollars.[282] Small per-part differences multiply through production volume and margin: the roughly two-dollar premium for a gigabit rather than a 10/100 Ethernet PHY was decisive on a board competing at a 35-dollar price point.[265]
When supply is short, allocation replaces price as the rationing mechanism. Memory and other high-volume semiconductor output is allocated in periodic lots, and a customer large enough to commit an entire period’s production can do so, leaving smaller buyers with nothing for that period regardless of price offered.[193] Distributors and vendors ration on expected volume and end use, so a buyer taking a thousand units into a product is served ahead of a reseller taking three hundred at a time.[207] Under allocation a supplier serves its largest committed customers first, and an order for ten thousand units is deprioritised behind million-unit accounts irrespective of the smaller buyer’s willingness to pay.[517] Component suppliers whose parts are in demand triage inquiries by order size, and a vendor able to sell 50,000 units elsewhere will decline to engage with a 3,000-unit buyer.[537]
High-volume semiconductor procurement runs roughly six months ahead of production, so the order book behaves like a six-month first-in-first-out buffer and a buyer’s place in it is set half a year before the parts are needed.[541] A dominant buyer can exploit that structure: a recorded pattern has the buyer inducing a component supplier to fund additional capacity against a promised order and then demanding a price reduction once the capacity exists, because the supplier’s new capacity has no alternative customer; a twenty percent reduction demanded after a capacitor vendor expanded for an automotive customer is a documented instance.[193] Sourcing advantages that are hard to copy include volume pricing unavailable to smaller buyers and exclusivity agreements under which a vendor supplies a given part to only one customer.[298]
Geography and manufacturing clusters
The principal sourcing advantage of the Pearl River Delta is density rather than labour cost: components and vendor capabilities of almost any kind can be found inside a single region, which shortens the search and the logistics for every line item.[113] A regional manufacturing cluster is an accumulated stock of infrastructure, parts availability and equipment built over roughly two decades, so it cannot be replicated elsewhere on the timescale of a wage change.[111] Relocating assembly away from a components cluster removes the ability to recover from a stock-out locally; where the market is nearby a stopped surface-mount line can be restarted by buying a reel the same day, and where it is not, the line waits for a shipment.[175] Manufacturing near a dense supplier base also shortens engineering support loops: a component problem that a datasheet cannot resolve can typically be answered by a local manufacturer’s representative attending in person within a day or two.[279]
Volume determines where manufacture pays. Below roughly one thousand units the labour content of assembly is a small share of total cost, so the shorter feedback loop, faster design-build-test cycle and stronger intellectual-property position of local manufacture usually dominate; offshore manufacture becomes economically compelling from around five thousand units upward, where labour and factory markup scale with volume.[113] Contract factories earn on volume, so orders below roughly five thousand units struggle to obtain a factory’s attention and priority.[113]
Factory organisation differs by region. Chinese contract manufacture tends to be vertically integrated, with surface-mount assembly, injection moulding, mould making, final assembly, test and packaging under one roof, while United States and European manufacture is typically horizontal and requires the buyer to assemble several specialist vendors into a working chain.[451] A horizontal supplier base costs more effort to assemble because there is no single accountable vendor, but a maintained database of which factories work well together and which already hold payment terms with each other converts that effort into reusable routing knowledge.[451]
Regional calendars and policy must be built into schedules. National holiday periods in a manufacturing region suspend production and shipping and must be built into delivery schedules by buyers ordering into that region.[391] Electronics assembly is the last step of a chain whose inputs are themselves imported, so relocating assembly under a tariff regime only shifts the tariff exposure upstream to bare boards, laminate, process chemistry and semiconductor inputs unless those tiers are built out at the same time.[699] Consumer electronics has historically been produced through original design manufacturers who handle both design and manufacture at high volume, leaving the brand with software and marketing; rising tariffs have made that arrangement less durable and pushed such companies to evaluate domestic assembly.[699]
On factory visits, Jeff Keyzer’s practice was to carry tools and consumables in hand luggage rather than checked baggage, because checked baggage can be delayed or lost and resupplying locally from an electronics market costs half a day of travel and negotiation.[279]
Inventory strategy
Just-in-time inventory is defined by the deliberate decision not to hold stock, with parts arriving as they are consumed.[197] Kanban, the scheduling method associated with it, was developed for vertically integrated automotive manufacture, where the supplying operations sit inside the same company and can be scheduled directly; a buyer purchasing catalogue parts from an external distributor controls none of the upstream steps and cannot apply the method in the same form.[197]
Just-in-time supply optimises working capital at the cost of resilience, and a local buffer stock is the counterweight; short disruptions of only a few weeks produced effects that persisted long afterwards.[511] Buying ahead is the main defence against allocation, so access to capital rather than engineering skill determines who can weather a shortage; a company unable to fund a forward purchase of its critical parts has no equivalent option.[573] On the hardware programme Ariel Briner described, critical integrated circuits were bought up to two years before the designs that would use them — a defensible response to scarcity that converted a supply risk into a cash-flow burden by tying up capital in parts for products that did not yet exist.[614]
Cash position is therefore part of supply readiness. Crowdfunding readiness cannot be judged from a claim that all parts are purchasable; the substantive tests are whether suppliers are contracted, whether payment terms and cash flow can carry the build, and whether the stock counted on has been secured, since uncommitted distributor stock can be bought out by another customer at any time.[393]
Design for supply continuity
A contract manufacturer’s supply-chain function works towards continuity by ensuring commodity parts carry several qualified alternates in the design, so no jellybean line item is single-sourced.[279] Single-sourced specialised components must be settled early in a development programme, because securing volume for them can require a supplier to stand up an assembly line or a new factory, which takes far longer than the design work it supports.[279] Choosing a part that has no pin- and feature-compatible relatives within its family eliminates any second-source or substitution path, leaving the design dependent on that single part number.[269] Semiconductor lead time sets a hard floor on a project schedule that no amount of resource can compress — eight weeks is a working minimum for a chip order — and dependent tasks must be sequenced behind it.[298]
Institutional customers can force multi-sourcing. A sole supplier of a technology to a defence customer can be required to standardise it and license it to competitors, so that the customer is not dependent on a single source.[443] Inside a company large enough to buy components by the reel and to employ supply-chain specialists, part selection is governed by the approved and stocked list, and a designer’s freedom to substitute a preferred part at layout is correspondingly small.[445] Adding a new part to a component library carries a recurring cost beyond symbol and footprint creation, because the part must be checked and its sourcing vetted; where a company employs a part librarian, that role is also where supply-chain screening happens.[408] Scaling a design beyond prototype requires both a documented, reliable datasheet for each part and an understanding of how that part is supplied; a component with no dependable datasheet is treated as a liability rather than a cost saving.[330]
Architecture decisions carry sourcing consequences. Absorbing discrete support functions into the main silicon reduces both bill-of-materials cost and the number of separate parts that must be sourced, so integration is a supply-chain simplification as well as a cost reduction.[648] Building a product on top of a finished development board converts that board into a purchased component and inherits its availability risk.[186] Development boards are priced as marketing vehicles at lower margins than production hardware and are built to a forecast of engineering demand; a single customer designing them into a product can consume the entire year’s build and deny the boards to their intended audience.[422] Designing a complete third-party board into a product as a plugged-in module is a sourcing risk and stops being economical once volumes reach the thousands, at which point the same function is integrated onto the product’s own board.[349] Prototyping with a certified radio module and moving to a chip-down implementation for production is a recognised path, taken for supply-chain reasons as well as cost; the reverse trade is that a pre-certified module buys back engineering and certification time that a small team may not have.[614]
More structural responses exist. Commissioning a custom integrated circuit can be justified purely as supply mitigation: a company dependent on a sole-sourced chip initiated its own design to remove that dependency, with the decision made by comparing the design investment against the exposure.[503] Open hardware release gives a company adopting a board the option to take control of its own supply, because the design detail needed to build or re-source it is available.[723]
At the factory level, incumbency substitutes for new sourcing work. A factory that already builds a product class holds the sourcing relationships and tooling for its inputs, so a variant of an existing product can be quoted and run at a few thousand units without new supply-chain work; the incremental effort is design change, not procurement.[336] Where a factory already runs the product class, a credible forecast of a couple of thousand units is enough to have a line duplicated for a variant within days.[336] An effective sourcing method for a new product is therefore to locate an existing product closely resembling the intended one and trace the factory that builds it, because that factory has demonstrated the capability, holds the tooling and team, and can usually modify the existing design to a new specification.[715]
Organisation and division of labour
Hardware programmes are built under several division-of-labour models: specifying the product and contracting an original design manufacturer to design and build it, joint development of schematics with the manufacturer, or completing the design in house and using the manufacturer only for layout and fabrication.[357] For a small hardware company, engaging a tier-one contract manufacturer is generally the wrong choice; a local box-build house where the company’s engineers can stand on the floor bringing up boards and designing test fixtures produces a working production process, because production competence is acquired by doing it rather than by delegating it.[402] Supply chain, tooling and inventory management are the slowest-changing parts of hardware development and change far more slowly than component technology; they are constraints to be operated within rather than problems to be designed away.[402]
Supply stability is a continuing engineering responsibility. Sustaining engineering owns supply stability alongside field-failure root cause, because keeping a released product buildable is a continuing engineering task rather than a one-time procurement act.[474] A stopped production line at a contract manufacturer is the worst operational outcome for a hardware company, because the idle line accrues cost every day while producing no revenue; it justifies dispatching an engineer immediately regardless of distance.[474]
Established chains are themselves an asset. At Simone Giertz’s product company, which had no established supplier relationships, each product took one to two years to bring to market because every sourcing step had to be built from scratch rather than reused.[592] Lead times vary by component class, and a well-planned production programme can absorb worst-case lead times of four months or more, around sixteen to eighteen weeks, without stopping the assembly line.[175]
Products spanning industries inherit incompatible cadences. A product spanning consumer electronics and fashion inherits both supply chains and their ordering rhythms: consumer electronics runs annual cycles with a handful of SKUs, while fashion runs quarterly seasons with twenty or more, and each side’s factories treat the other’s order pattern as unworkable.[537]
Disruption and shortage
The 2020 to 2022 semiconductor shortage began with mass order cancellation rather than production failure: when demand was expected to collapse, large buyers cancelled the orders already placed six months out, their capacity was reallocated, and the cancelled buyers could not regain their place when demand returned.[541] In 2020 parts remained obtainable from stock already sitting in distribution and in customers’ hands even though factories had stopped producing them; that residual stock took until 2021 to drain, which is why the shortage was felt a year after its cause.[570] Recovery from a manufacturing-region shutdown is vendor-specific rather than regional; during the 2020 disruption some board fabricators resumed normal delivery while others quoted eight-week lead times.[482]
At the peak of the shortage, distributor promise dates ran more than a year out, and designs were repeatedly revised — in one case to an eighth board revision — purely to accept whichever microcontroller could be obtained.[570] A placed order at a quoted lead time is not an allocation guarantee; a documented case had a component on a 28-week lead cancelled by the supplier in the 28th week after a larger buyer paid more, with no production shortfall involved.[570] A critical part could be out of stock worldwide with a quoted lead time of 35 weeks, leaving redesign around an alternative part as the only practical response, since replicating an integrated function such as a boost converter in discrete form is not viable.[527]
Shortage behaviour amplifies itself. Scarcity induces purchasing agents to order more than they need to protect against being caught short, which inflates the apparent demand signal that suppliers plan against and leaves an inventory overhang once production catches up.[530] Individual hoarding propagates a shortage rather than solving it: a designer who redesigns onto an available part and then buys several reels of it removes that stock from distribution and pushes the same problem onto the next buyer who had planned around it.[573]
Substitution under pressure is a distinct failure class. A component substituted for a cost saving can match the datasheet headline specification and still fail at the actual operating point: on Josh Lifton’s hardware programme, coin cells substituted to save one cent each met their paper specification but failed after a single use at the design’s current draw, discovered only 24 to 48 hours before deployment of 21,000 units.[314] A component change introduced by a supply-chain substitution after qualification can push a marginal on-chip regulator over the edge with no failure indication, only reduced efficiency; where a device has a known sensitive internal converter, feeding the core rail from an external regulator removes the exposure.[676] At the extreme, a product assembled from surplus and open-market stock has no fixed bill of materials; two production months of the same low-cost device can contain substantially different parts, because sourcing follows whatever was cheapest and available at the time.[142] Crowdfunded hardware projects more often fail on execution than on concept, and the failures cluster in timelines, quality and supplier selection, where the founders had no prior exposure to the process.[402]
Country-level exposure is managed rather than eliminated. Qualified second sources for raw materials reduce but do not remove country exposure: alternates exist for most materials, yet they carry volume and capability constraints and cannot supply everything a manufacturer would want if the primary source were cut off.[632] Under chief executive Steve Sanghi, Microchip reduced its country concentration by stages — owning no assembly or test plant in one country, performing under five percent of packaging and only two to three percent of wafer fabrication there — while remaining dependent on it for rare-earth and other raw materials.[632]
The same dynamics appear outside electronics. Heavy electrical infrastructure carries lead times measured in years rather than weeks; large transformers ran to a three-year backlog, and a grid-scale battery installation was taken partly offline by a transformer failure while the battery itself was undamaged.[724] Logistics disruption is a measurable cause of infrastructure failure: of the recorded serious United States power outages in 2014, 77 followed severe weather, 66 followed physical attack, and 17 followed fuel shortages caused by supply problems such as rail congestion.[612]