Samsung’s AI substrate investment: why the layers beneath chips matter

A planned Sejong expansion puts the components beneath AI chips in focus. Building capacity takes years; turning it into cash depends on customers, production quality and pricing.

By The Strategic Newb Editorial Team
Event: 28 September 2026 · Published: 30 September 2026 · Updated: 30 September 2026

Samsung Electro-Mechanics has announced a 4.27 trillion won expansion of its Sejong substrate facilities, according to The Korea Times, citing the company’s regulatory filing on 28 September. New lines are planned to enter mass production in September 2028. The beginner’s question is what this spending buys: a larger place in the supply chain beneath AI processors, with the financial outcome still to be earned.

Three key points

• The announcement concerns manufacturing capacity, not a completed new factory or proven future profit.
• Substrates connect chips to the wider system; a powerful processor needs reliable connections as well as computing capability.
• The relevant horizon is roughly two years for the planned production start, and longer for assessing returns.

A physical connection behind a digital service

FC-BGA stands for flip-chip ball grid array. It is a type of package substrate: the supporting connection layer between a semiconductor and the main board. Samsung’s earlier technical explanation, published in September 2025, describes its role in carrying electrical signals and power. That older source explains the technology; it is not evidence that the newly announced lines already operate.

Think of the processor as one part of a complete machine. Customers need its connections, supporting components and surrounding equipment to work together. Buying more of one component does not solve a shortage of another. This is why investment in a less visible manufacturing stage can matter even when the best-known chip designer captures most public attention.

What the announcement does and does not establish

Newsis separately reports the same Sejong investment and September 2028 production target. These are reports of a company plan. They do not independently prove that future customer demand will absorb the output. We have not inspected the underlying filing directly; the accessible company disclosure list did not show this new document. No unnamed customer commitment or guaranteed sales volume is assumed here.

The evidence is therefore stronger for the existence of an announced expansion than for its eventual earnings. A budget tells us what management intends to spend. A production target tells us when it hopes to start. Neither tells us the eventual selling price, the proportion of output that meets specifications, or the cash left after costs.

How factory spending can become earnings

The following is economic analysis. Construction and equipment create cash outflows before the new lines generate sales. Once production starts, the factory must make acceptable products consistently. Yield is the share of manufactured units that pass the required quality checks. Poor yield means more material, machine time and labour are needed for each saleable unit, which can reduce profit even when demand looks strong.

Utilisation matters too: it measures how much available capacity is actually used. Many factory costs continue when machines are idle. Higher utilisation can spread those costs over more products, but only if customers need the output at prices that cover the full bill. Producing more units simply to fill a factory is not the same as generating healthy cash flow.

An expanding market can attract competing factories at the same time. If supply grows faster than orders, customers may gain negotiating power and prices can weaken. Conversely, reliable capacity for demanding products may help a supplier win repeat business. These are alternative scenarios, not claims that either outcome has already occurred at Sejong.

Who could benefit, and who carries the risk?

Equipment and materials suppliers could benefit during the building phase, provided they actually receive orders. Chip and server customers could later benefit from additional qualified supply. Factory employees and specialist engineers may see opportunities if the expansion proceeds. None of these effects can be quantified from the headline investment amount alone, and this article makes no jobs estimate.

Shareholders fund exposure to execution and demand risk. A technologically useful product can still be an expensive investment if construction overruns, customer requirements change or competitors price aggressively. Customers also retain risk: extra nominal capacity is less valuable if it is late or has not passed their approval process. Capacity, qualified supply and profitable supply are different milestones.

What a new investor should watch next

Now, look for the company’s own detailed disclosures and subsequent capital-spending updates. Over the next one to three years, the useful checks are construction progress, customer qualification and whether the September 2028 production target holds. Beyond that, watch utilisation, margins and cash generation. A margin is the share of revenue left after the relevant costs; a rising sales figure alone does not establish an improving business.

The structural lesson is that AI investment reaches beyond processors into specialised manufacturing. The financial lesson is to follow the chain from announced spending to working capacity, then to paying customers and cash. This announcement gives a concrete project to monitor. It does not establish an attractive share price or predict an investment return.

Sources

• The Korea Times: Samsung Electro-Mechanics to invest $3.14 bil. for substrate plant in Sejong — 2026-09-28T18:58:00+09:00
• Newsis: 삼성전기, 세종에 4조2700억 투자…AI 서버용 반도체 기판 생산라인 증설 — 2026-09-28
• Samsung Electro-Mechanics: Package substrate technology at KPCA Show 2025 — 2025-09-04