> ## Content Index
> Fetch the complete content index at: https://www.sotkn.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Fabless vs IDM: Who Owns the Fab, Who Books It, and Who Carries the Risk
- URL: https://www.sotkn.com/explained/fabless-vs-idm/
- Published: 2026-11-17T10:28:14.000Z
- Updated: 2026-11-17T10:28:14.000Z
- Description: A fabless company designs chips and books someone else’s factory. An IDM designs chips and owns the factory. Nvidia keeps 75 cents of each revenue dollar as gross profit; Intel, which runs its own fabs, keeps 43. But the IDM wins when its fabs are full and it never has to queue.
- Author: Sean
- Tags: Versus, #desk-explained

EXPLAINED · Versus · Last reviewed: 17 Nov 2026 · Next review: after Intel’s, Texas Instruments’ and Samsung’s fourth-quarter 2026 results, or Nvidia’s next quarterly filing

Fabless and IDM are the two main ways to run a chip company. A fabless company designs chips but owns no fab and pays a foundry to make them. An IDM, or integrated device manufacturer, designs chips and makes them in fabs it owns.

**In short.** Fabless and IDM differ on one question: who owns the factory. A fabless company such as Nvidia books a foundry’s fab and pays per wafer. An IDM such as Intel or Texas Instruments owns its fabs and pays for them up front. The fabless model keeps more of each dollar: Nvidia’s gross margin was 75.0% in its latest quarter, against 43.0% at Intel. But the IDM model pays off when its fabs are full and paid for: Texas Instruments reached 64.2%. Our read: both models pay for empty capacity. The fabless company pays through purchase commitments, which at Nvidia reached US$279 billion. The IDM pays through fabs that cost money whether or not they run. The stuck step is filling the fab, whoever owns it.

Why an investor cares: the two models put the cost of a chip factory in different places. In a fabless company it sits in purchase commitments and the foundry’s price per wafer. In an IDM it sits on the balance sheet, as fabs that have to be filled. Knowing which model a company runs tells you where to look when demand changes. This page builds on our [fabless explainer](https://www.sotkn.com/explained/fabless-explained/) and sits in our [Semiconductors](https://www.sotkn.com/stack/semiconductors/) layer.

## Fabless vs IDM: the short answer

Both kinds of company design chips and sell them under their own name. The difference is the factory. The fabless company rents it, one wafer at a time, from a foundry. The IDM owns it. The table sets out what follows from that one choice.

|                                    | Fabless                                                                     | IDM (integrated device manufacturer)                                            |
| ---------------------------------- | --------------------------------------------------------------------------- | ------------------------------------------------------------------------------- |
| **Designs the chip?**              | Yes                                                                         | Yes                                                                             |
| **Owns the fab?**                  | No, books a foundry                                                         | Yes                                                                             |
| **How it pays for manufacturing**  | Per wafer, as it buys; commits to volumes ahead                             | Up front, in clean rooms and tools; then wages, power and materials to run them |
| **Latest gross margin**            | Nvidia 75.0%                                                                | Texas Instruments 64.2%; Intel 43.0%                                            |
| **Factories on the balance sheet** | Small                                                                       | Large: Intel carries US$105.3bn of net property, plant and equipment            |
| **Main risk**                      | Committing to supply before demand is known; waiting in the foundry’s queue | Fabs that run below capacity; spending on a process that falls behind           |
| **Where it fits best**             | Leading-edge logic: GPUs, custom AI chips, phone chips                      | Memory, analog and power chips, where the process is part of the product        |
| **Examples**                       | Nvidia, Broadcom, AMD, Qualcomm, MediaTek                                   | Intel, Texas Instruments, Micron, SK hynix; Samsung is an IDM and a foundry     |

Margins are GAAP gross margin for each company’s latest reported quarter: Nvidia to 26 July 2026, Intel to 26 September 2026, Texas Instruments to 30 September 2026\. Not a rating of companies or investments.

## What is a fabless company?

A fabless company designs chips and owns the customer, but owns no fab. It sends finished designs to a foundry, which prints them on silicon wafers for a fee, and it buys memory, packaging and testing from other specialists. Nvidia, Broadcom, AMD, Qualcomm and MediaTek all work this way. Our [fabless page](https://www.sotkn.com/explained/fabless-explained/) walks through the five steps from design to shipped chip.

The model is now the bigger half of the chip business by revenue. TrendForce estimates that the 10 largest fabless companies took in more than US$141 billion in the second quarter of 2026, about 2.6 times the roughly US$53 billion taken by the 10 largest foundries, by our arithmetic. Nvidia alone accounted for 64% of the fabless total.

## What is an IDM?

An integrated device manufacturer, or IDM, designs its chips and makes them in fabs it owns. It runs the clean rooms, buys the tools, employs the process engineers and decides what runs on which line. This was how the whole chip industry worked before foundries existed, and it is still how most memory, analog and power chips are made.

Intel is the best-known IDM in processors. Texas Instruments is one in analog chips, the parts that manage power and signals in nearly every device. Micron, SK hynix and Samsung are IDMs in memory; our [who makes DRAM](https://www.sotkn.com/explained/who-makes-dram/) and [who makes NAND](https://www.sotkn.com/explained/who-makes-nand/) pages cover them. Samsung is also a foundry, so it sits in both camps.

![Two ways to get a chip made: both design the chip; fabless books a foundry; IDM builds and fills a fab; both buy memory, packaging and test; fabless carries commitments, IDM carries the factory](https://storage.ghost.io/c/cd/10/cd10e222-131f-4ba4-b4a9-ca4214e8c12d/content/images/2026/11/fabless-vs-idm-flow.png)

How the two models get a chip made, and what each carries. Second Order.

## Which model keeps more of each dollar?

Gross margin is the simplest test: the share of revenue left after the cost of making what was sold. It captures the factory question directly, because an IDM’s cost of goods includes running and writing down its own fabs, while a fabless company’s includes the price the foundry charges.

Nvidia’s gross margin was 75.0% in the quarter to 26 July 2026, so the cost of making its products was about a quarter of its US$96.2 billion revenue. Intel reported a 43.0% gross margin for the quarter to 26 September 2026, 32 percentage points lower, by our arithmetic. Texas Instruments, an IDM in a different market, reported 64.2% for the quarter to 30 September 2026.

![Bar chart of gross margin in the latest quarter: Nvidia 75.0%, Texas Instruments 64.2%, Intel 43.0%](https://storage.ghost.io/c/cd/10/cd10e222-131f-4ba4-b4a9-ca4214e8c12d/content/images/2026/11/fabless-vs-idm-margin-chart.png)

Gross margin, latest reported quarter. Company releases; quarters differ. Chart by Second Order.

The gap is not only about the model. Nvidia sells the most sought-after chips in the world, and its prices reflect that. But the comparison shows the trade: the fabless company pays the foundry a margin on every wafer, yet still keeps more, because it never carries an empty fab. An IDM keeps the manufacturing margin for itself only when its fabs are full and its process is competitive.

## Intel: what the IDM model costs while it builds

Intel splits its results in a way that shows the IDM model from the inside. Intel Products designs and sells the chips. Intel Foundry runs the fabs and sells wafers, mostly to Intel Products, at prices Intel says are meant to approximate the market.

| Intel, third quarter 2026                             | Revenue, US$m | Operating income (loss), US$m | Margin  |
| ----------------------------------------------------- | ------------- | ----------------------------- | ------- |
| **Intel Products** (the chips it designs and sells)   | 15,415        | 5,324                         | 34.5%   |
| **Intel Foundry** (the fabs that make them)           | 6,697         | (1,853)                       | (27.7%) |
| **Whole company**, after other items and eliminations | 16,513        | 2,522                         | 15.3%   |

Intel segment results for the quarter ended 26 September 2026, from its Form 10-Q. Foundry revenue is mostly sales to Intel Products, so the two segments add up to more than the whole company. Margins are our arithmetic.

The design side earned an operating margin of 34.5% in the third quarter, by our arithmetic. The factory side lost US$1.85 billion on US$6.7 billion of revenue: it spent about US$1.28 for every dollar it took in. That loss cancelled about 35% of the profit the product side made.

The loss is narrowing. It was US$2.32 billion a year earlier, so it improved by US$468 million. Intel says higher yields, faster cycle times and fuller fabs are bringing wafer costs down, which is the IDM model working as intended: fill the fab and each wafer gets cheaper.

The factory is also mostly serving one customer. Intel Foundry booked US$411 million from outside customers in the quarter, up from US$32 million a year earlier. That is still only about 6% of its revenue; the other 94% came from Intel itself, by our arithmetic. An IDM that wants to become a foundry has to win outside work to fill its fabs, and that takes years.

The balance sheet shows the weight. Intel carried US$105.3 billion of net property, plant and equipment at the end of September, about 2.3 times its revenue for the first nine months of 2026\. It spent US$10.2 billion on new plant and equipment in those nine months, and its commitments for future capital spending rose to US$19.6 billion from US$12.8 billion at the end of 2025\. Our [capex vs depreciation](https://www.sotkn.com/explained/capex-vs-depreciation/) page explains why that spending hits profit years after it is paid.

## Texas Instruments: what the IDM model pays when fabs are full

Texas Instruments shows the other side. It makes analog and embedded chips in its own fabs, and over the past few years it spent heavily to move production onto 300-millimetre wafers, which hold more than twice as many chips as the older 200-millimetre size. It closed its remaining 150-millimetre production last year.

Now that spending has slowed, the payoff shows. Third-quarter revenue rose 28% from a year earlier to US$6.06 billion. Gross margin rose 6.8 percentage points to 64.2%, and operating margin reached 47.5%. Capital spending fell to US$463 million in the quarter, from US$1.20 billion a year earlier, and to US$2.58 billion over 12 months, 46% less than a year before, by our arithmetic. Free cash flow over 12 months was US$8.4 billion, about 3.5 times the year before.

The company itself credits “the benefit of 300mm production”. The lesson for the versus: the IDM model pays when the fabs are built, full and making chips whose process is part of the product. Analog chips do not need the newest node, so a fab can earn for decades.

## Samsung: an IDM that is also a foundry

Samsung runs both models under one roof. Its memory business is a classic IDM: it designs DRAM and NAND and makes them in its own fabs. Its Foundry business makes chips for others, and its System LSI business designs processors and image sensors, some of which Samsung Foundry makes.

The memory side dominates. Samsung’s chip division, Device Solutions, posted KRW 147.1 trillion of revenue and KRW 106.5 trillion of operating profit in the third quarter, a 72% operating margin, by our arithmetic, on record memory prices. Samsung expects to spend about KRW 84.4 trillion on capital projects in 2026, 91% of it in that division.

TrendForce notes that Samsung Foundry’s orders include base dies for high-bandwidth memory. That is the hybrid at work: the foundry arm makes the logic layer that sits under the memory stack the IDM arm sells. Samsung says its new fab in Taylor, Texas, should start operating in 2027.

## Who carries the risk?

Neither model escapes the cost of capacity that is not used. They carry it in different places.

- **The fabless company carries it in commitments.** To be sure of wafers, memory and packaging, it promises to buy them ahead. Nvidia’s commitments rose from US$119 billion to US$279 billion in one quarter, mainly for memory: 2.3 times the previous level and 2.9 times its quarterly revenue, by our arithmetic. If demand falls, those promises still have to be met. Our [purchase commitment](https://www.sotkn.com/explained/purchase-commitment-explained/) page explains how these work.
- **The IDM carries it in the fab.** A fab costs about the same to own whether it runs full or half empty. When volumes drop, the cost of each wafer rises and gross margin falls. Intel’s guidance for the fourth quarter assumes a GAAP gross margin of 37.1%, down from 43.0%.
- **The IDM also carries the process risk.** If its manufacturing process falls behind, it cannot simply switch suppliers; it has to spend more to catch up. A fabless company can, in principle, move a future design to another foundry, though redesign and new masks make that slow.
- **The fabless company carries the queue risk.** When a foundry is full, the fabless company waits. TrendForce says TSMC’s 5/4nm and 3nm capacity was fully booked in the second quarter. An IDM with spare room in its own fab does not queue.

## Which model fits which chip?

The split follows the chip, not fashion.

- **Leading-edge logic goes fabless.** AI GPUs, custom accelerators and phone processors need the newest process, and a leading-edge fab costs tens of billions of dollars. Only a foundry serving many customers can keep one full, so most designers book TSMC. Our [who makes AI chips](https://www.sotkn.com/explained/who-makes-ai-chips/) page covers the designers.
- **Memory stays IDM.** DRAM and NAND are commodity-like products whose cost depends on the process itself. The memory makers design the chip and the process together, so they keep both.
- **Analog and power stay mostly IDM.** These chips use older processes tuned for each product, and the fabs last for decades. Texas Instruments is the clearest case.
- **Many sit in between, as fab-lite.** Companies such as NXP keep some fabs and send the rest of their output to foundries, or share the cost of a new fab with a foundry partner.

## Who gets paid?

Our read: the answer depends on which model is growing.

- **When fabless grows, the foundry gets paid per wafer.** Every fabless dollar of cost is a foundry dollar of revenue. With TSMC taking 72.5% of foundry revenue in the second quarter, by TrendForce’s estimate, most of that flows to one company. Our [what TSMC does](https://www.sotkn.com/stack/semiconductors/cowos/what-does-tsmc-do/) page covers it.
- **When IDMs build, tool makers get paid up front.** An IDM buys its own tools rather than paying a foundry to buy them, so its capital budget goes straight to equipment suppliers. Intel’s rising capital commitments are orders for those suppliers. Our [who makes wafer fab equipment](https://www.sotkn.com/explained/who-makes-wafer-fab-equipment/) page covers them.
- **Memory makers get paid by both.** Fabless designers buy high-bandwidth memory from IDMs; Nvidia says its new commitments are mainly for memory.
- **The owner of a full fab gets paid most.** Whether that is a foundry or an IDM, the fab that runs near capacity on a process customers need earns the best return. The one that runs half empty pays for it.

## Where is the stuck step?

Our read: filling the fab, whoever owns it. A fabless company is stuck when it cannot get enough wafers or packaging slots booked; at the leading edge that means TSMC’s queue. An IDM is stuck when it cannot fill the fabs it has built, as Intel Foundry’s 94% reliance on Intel itself shows, or when it cannot build new ones fast enough to meet demand, as the memory makers are finding.

The clock is set by construction and qualification. A new fab takes years to build and fill, and a new customer takes months to qualify a process. Our [semiconductor yield](https://www.sotkn.com/explained/semiconductor-yield-explained/) page explains why the early wafers on any new line cost the most.

## Fabless vs IDM in Singapore

Singapore shows the middle path. On 28 September 2026, VisionPower Semiconductor Manufacturing Company, or VSMC, opened its first 12-inch fab in Tampines. VSMC is a joint venture of Vanguard International Semiconductor, a Taiwanese foundry, and NXP, a Dutch chipmaker that runs fabs of its own.

For NXP, the venture is a way to secure manufacturing without carrying a whole new fab alone, the fab-lite approach in practice. The fab has entered risk production, with volume production due in the first quarter of 2027, and is expected to reach about 44,000 wafers a month and 1,600 jobs by 2029, its operator says. That is about 528,000 wafers a year at full capacity, by our arithmetic.

Our [Singapore Markets door](https://www.sotkn.com/markets/singapore/) follows the local names, and our [Taiwan door](https://www.sotkn.com/markets/taiwan/) tracks the foundries.

## What would change the view?

- **Intel wins large outside foundry customers.** If outside revenue grows from about 6% of Intel Foundry’s sales to a large share, the IDM becomes a second leading-edge foundry, and fabless companies gain a choice.
- **A large fabless company builds its own fab.** That would blur the line from the other side. Watch for any designer announcing its own manufacturing.
- **Demand turns.** Then the fabless company’s commitments and the IDM’s empty fabs both start to cost money. Which side hurts more depends on how much each has committed.
- **Memory prices fall.** The memory IDMs’ margins today rest on record prices. A fall would show how much of the IDM payoff was price, not model.

The clocks: Nvidia’s next quarterly filing and its commitments table; Intel’s, Texas Instruments’ and Samsung’s fourth-quarter 2026 results in early 2027, with Intel’s gross margin against its 37.1% guidance and its outside foundry revenue; TrendForce’s third-quarter fabless and foundry tables; Samsung’s Taylor fab start in 2027; and VSMC’s volume production in the first quarter of 2027.

## Common misconceptions

- **People assume IDMs are old-fashioned.** Actually the most profitable chip division in the latest quarter, Samsung’s memory-led Device Solutions arm, is an IDM, and Texas Instruments earns a 64.2% gross margin from its own fabs.
- **People assume fabless companies carry no factory risk.** Actually they carry it through commitments. Nvidia’s reached US$279 billion in July 2026.
- **People assume owning fabs means cheaper chips.** Actually only full fabs are cheap. Intel Foundry spent about US$1.28 for every dollar of revenue in the third quarter.
- **People assume Intel is now a foundry like TSMC.** Actually about 94% of Intel Foundry’s third-quarter revenue came from Intel itself, by our arithmetic. It is still mainly an IDM’s factory arm.
- **People assume the choice is all-or-nothing.** Actually many chipmakers are fab-lite: they keep some fabs and buy the rest, as NXP does through its Singapore joint venture.

## FAQ

**What is the difference between fabless and IDM?**  
A fabless company designs chips and pays a foundry to make them. An IDM, or integrated device manufacturer, designs chips and makes them in its own fabs. Nvidia and Broadcom are fabless; Intel, Texas Instruments, Micron and SK hynix are IDMs.

**Is Intel fabless or an IDM?**  
Intel is an IDM. It designs its processors and makes most of them in its own fabs, run by its Intel Foundry business. In the third quarter of 2026, about 94% of Intel Foundry’s revenue came from Intel itself, by our arithmetic. Intel also buys some wafers from outside foundries.

**Is Nvidia fabless or an IDM?**  
Nvidia is fabless. It designs its GPUs and networking chips and relies on foundries, chiefly TSMC, to make them. It buys memory, packaging and assembly from other companies too.

**Which is more profitable, fabless or IDM?**  
It depends on how full the fabs are. Nvidia’s gross margin was 75.0% in its latest quarter. Texas Instruments, an IDM with full, paid-off factories, reached 64.2%. Intel, an IDM paying for new factories, reported 43.0%. The fabless model wins on margin; a full, mature IDM fab can come close.

**Why would a company choose to be an IDM?**  
To control supply and cost. An IDM does not wait in a foundry’s queue and keeps the manufacturing margin itself. That works best for chips whose making is part of the product, such as memory, analog and power chips.

**What is fab-lite?**  
Fab-lite is the middle ground: a company keeps some fabs of its own and sends the rest of its output to foundries. Many chipmakers now work this way. The Singapore fab opened in September 2026 by VSMC, a joint venture of the foundry VIS and the chipmaker NXP, is one example of how such companies share fab costs.

**Is Samsung fabless or an IDM?**  
Samsung is an IDM and a foundry at once. It makes its own memory and processors in its own fabs, and its Foundry business also makes chips for other companies, including base dies for high-bandwidth memory.

**Is this advice to buy or sell any company?**  
No. This page compares two business models and who they pay. It does not rate companies as investments or give buy or sell calls. Not investment advice. Do your own research.

Related: [fabless explained](https://www.sotkn.com/explained/fabless-explained/); [who makes AI chips](https://www.sotkn.com/explained/who-makes-ai-chips/); [silicon wafers](https://www.sotkn.com/explained/silicon-wafer-explained/); [EUV vs DUV](https://www.sotkn.com/explained/euv-vs-duv/); [take-or-pay vs purchase commitment](https://www.sotkn.com/explained/take-or-pay-vs-purchase-commitment/); [capex vs opex](https://www.sotkn.com/explained/capex-vs-opex/).

## Stack doors

Layer: [Semiconductors](https://www.sotkn.com/stack/semiconductors/). Players: [TSMC](https://www.sotkn.com/stack/semiconductors/cowos/what-does-tsmc-do/) and [Samsung](https://www.sotkn.com/stack/semiconductors/hbm/what-does-samsung-do/). Markets: [Korea](https://www.sotkn.com/markets/korea/), [Taiwan](https://www.sotkn.com/markets/taiwan/) and [Singapore](https://www.sotkn.com/markets/singapore/).

[Get the Wire](https://www.sotkn.com/#/portal/signup) — short daily notes on where AI computing gets built and who pays for it.

Not investment advice. Do your own research.

**How we check this**

Nvidia’s revenue, gross margin and commitments come from its CFO commentary for the quarter to 26 July 2026\. Intel’s revenue, margins and guidance come from its results release of 29 October 2026; its segment results, external foundry revenue, property, plant and equipment, capital spending and capital commitments come from its Form 10-Q for the quarter to 26 September 2026\. Texas Instruments’ figures come from its results release of 21 October 2026, and Samsung’s from its release of 29 October 2026\. Fabless and foundry revenue and shares come from TrendForce’s second-quarter 2026 tables of 9 and 11 September 2026; these are a research firm’s calendar-quarter estimates, not company-reported figures. The Singapore fab figures come from the VSMC release published by Singapore’s Economic Development Board on 28 September 2026\. Companies report in different quarters and, for Samsung, a different currency, so we compare margins and ratios rather than adding totals. Margins, shares, ratios and changes are our arithmetic, checked in code and labelled as ours on the page. Who gets paid and the stuck step are our read. The featured image is an illustration, not a photograph of a specific plant. This page compares two business models and who they pay. It does not rate companies as investments or give buy or sell calls. Not investment advice. Do your own research.

**Last reviewed:** 17 Nov 2026 · **Next review:** after Intel’s, Texas Instruments’ and Samsung’s fourth-quarter 2026 results, or Nvidia’s next quarterly filing

Spot an error? [Tell us](mailto:sean@thekopinotes.com) and we will correct it and note the change here.

[Our editorial standards →](https://www.sotkn.com/editorial-standards/)

## References

1. [TrendForce — Top 10 Global Fabless IC Design Houses Post 73% YoY Revenue Growth in 2Q26; AMD Breaks into Top Three (11 Sep 2026)](https://www.trendforce.com/presscenter/news/20260911-13231.html?ref=sotkn.com) — Second-quarter 2026 revenue of the 10 largest fabless companies: more than US$141.45bn combined; Nvidia about US$91.16bn, 64% of the top 10
2. [TrendForce — Global Foundry Revenue Approaches US$53.49 Billion in 2Q26 as SMIC Narrows Market Share Gap with Samsung (9 Sep 2026)](https://www.trendforce.com/presscenter/news/20260909-13225.html?ref=sotkn.com) — Top 10 foundries about US$53.49bn; TSMC about US$40.2bn and 72.5% share, 5/4nm and 3nm fully booked; Samsung Foundry orders include HBM base dies
3. [NVIDIA — CFO Commentary on Second Quarter Fiscal 2027 Results, SEC exhibit 99.2 (quarter ended 26 Jul 2026)](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm?ref=sotkn.com) — Revenue US$96.2bn, gross margin 75.0%; commitments up from US$119bn to US$279bn, mainly memory
4. [Intel — Intel Reports Third-Quarter 2026 Financial Results (29 Oct 2026)](https://www.intc.com/news-events/press-releases/detail/1786/intel-reports-third-quarter-2026-financial-results?ref=sotkn.com) — Revenue US$16.5bn, GAAP gross margin 43.0%, operating margin 15.3%; Intel Products US$15.4bn and Intel Foundry US$6.7bn; fourth-quarter guidance US$15.0–16.0bn, GAAP gross margin 37.1%
5. [Intel — Form 10-Q for the quarter ended 26 Sep 2026](https://www.intc.com/filings-reports/all-sec-filings/content/0000050863-26-000191/intc-20260926.htm?ref=sotkn.com) — Segment results (Intel Foundry revenue US$6,697m, operating loss US$1,853m); external foundry revenue US$411m; net property, plant and equipment US$105.3bn; capital commitments US$19.6bn
6. [Texas Instruments — TI reports third quarter 2026 financial results and shareholder returns (21 Oct 2026)](https://www.ti.com/en-us/about-ti/newsroom/news-releases/2026/2026-10-21-ti-reports-third-quarter-2026-financial-results-and-shareholder-returns?ref=sotkn.com) — Revenue US$6.06bn, gross profit US$3.89bn, operating profit US$2.87bn; capital expenditures US$463m in the quarter and US$2.58bn over 12 months; 12-month free cash flow US$8.4bn
7. [Samsung Electronics — Samsung Electronics Announces Third Quarter 2026 Results (29 Oct 2026)](https://news.samsung.com/global/samsung-electronics-announces-third-quarter-2026-results?ref=sotkn.com) — Device Solutions revenue KRW 147.1tn and operating profit KRW 106.5tn; Foundry gains on HBM base dies and 2nm orders; 2026 capital expenditure about KRW 84.4tn, KRW 76.8tn in Device Solutions; Taylor fab to start in 2027
8. [Singapore EDB — VSMC Celebrates the Grand Opening of Its First 300mm Fab in Singapore (28 Sep 2026)](https://www.edb.gov.sg/en/news-and-insights/vsmc-celebrates-the-grand-opening-of-its-first-300mm-fab-in-singapore?ref=sotkn.com) — VIS–NXP joint venture fab in Tampines; risk production, volume production 1Q27; about 44,000 12-inch wafers a month and 1,600 jobs by 2029

S

**[Sean](https://www.sotkn.com/author/sean/)**

Writes Second Order by TKN — a plain-English map of AI infrastructure and semiconductors for non-specialist investors. Focus: who gets paid, where money sticks, and which step is stuck.

Covers AI infrastructure, semiconductors, and second-order stack reads.

[Author page →](https://www.sotkn.com/author/sean/) · [Editorial standards →](https://www.sotkn.com/editorial-standards/)