Fabless vs IDM vs Foundry: Where Your Skills Fit

You have an offer from a company that sends its designs to a foundry, and a second one from a company that runs its own fabs. Both say "chip company" on the badge. Will your Tuesday look the same at either one? Probably not, and the difference shows up in the interview before it shows up in the job.

This post uses annual report filings and company pages from NVIDIA, Qualcomm, Texas Instruments, Micron, TSMC, and Intel, plus a small amount of anecdotal community material. The practice questions and the comparison table are ours, not reported from any employer.

What the companies say about themselves

NVIDIA's annual report (the FY2020 10-K) says it does not directly manufacture the semiconductors in its products and uses a fabless strategy. Suppliers handle wafer fabrication, assembly, test, and packaging, with TSMC and Samsung named as wafer suppliers. In its risk factors it says it does not own or operate a wafer fabrication facility and has no long-term commitment contracts with those foundries, which it lists as a source of supply, capacity, and yield risk.

Qualcomm's FY2025 10-K describes a fabless production model for its chip business, with exceptions for certain RF front-end modules and RF filter products made in internal facilities. It names TSMC, Samsung, and GlobalFoundries as primary foundry suppliers. It also describes two purchasing arrangements: foundries delivering fully assembled and tested parts, or Qualcomm buying die and contracting bump, probe, assembly, and most final test separately. That detail matters for you. Even at a fabless company, someone owns the supply chain work between the foundry and the finished part.

On the other side, TI's manufacturing page says it operates 15 sites including wafer fabs and assembly and test factories, that it develops and owns its process technologies, and that it plans for in-house operations to support more than 95 percent of its production by 2030. Micron's FY2025 10-K says it manufactures in its own facilities in several countries and uses subcontractors for some steps, such as certain assembly work.

TSMC describes itself as a pure-play foundry with an exclusive focus on manufacturing its customers' products. Its company profile reports making 12,682 products for 534 customers in 2025. Intel's foundry page describes process technology, advanced packaging, test, and wafer manufacturing offered to customers. Intel has long been described as an IDM, and how its product and foundry groups relate is something the page does not explain, so ask about it rather than assuming.

How does the day-to-day work change by company model?

These are the broad patterns. Titles and teams vary a lot, so read them as tendencies.

	Fabless	IDM	Foundry
Day-to-day work	Architecture, RTL, verification, physical design, post-silicon bring-up, plus foundry-facing work like PDK use and tapeout sign-off	Design plus process co-development, product and test engineering, yield, often closer to the fab	Process modules, integration, yield, equipment, design enablement, customer support
Interviews tend to probe	Design depth, tools and flow, debugging methodology, project ownership	Device and circuit fundamentals, how design choices meet process limits, manufacturing awareness	Process physics, data analysis, troubleshooting, statistics
Skills to show	Tapeout experience, timing, verification coverage, clear ownership	Cross-boundary thinking: a layout choice and its yield or reliability consequence	Hypothesis-driven problem solving, measurement discipline
Questions to ask	Which foundry and node? Who owns foundry issues? How much silicon bring-up is in this role?	How close are designers to the fab? Who owns yield learning?	Which module or customer problem does this role own? What is the shift or on-call pattern?

A fabless designer typically works with a process through its design kit and rules, and learns about manufacturing mostly through what the foundry tells them. An IDM engineer may sit near the people who change the process, which can be a real advantage if you like the physical side of the work. A foundry engineer's customer is often a design team at another company, which pushes the job toward consistency, throughput, and data.

What interviews tend to emphasize

For a design role at a fabless company, expect your preparation to center on the design topics in your area: RTL, timing, physical design, or verification. Foundry knowledge helps most when it explains a decision, for example why a particular constraint exists.

For process, yield, or equipment roles, our foundry process engineer guide covers the data problem approach: define the change, check the measurement, list competing explanations, and say what evidence would separate them. For memory, where the product and the process are tightly linked, see our memory design interview guide.

An IDM can sit between the two. Here is a practice prompt that is ours: a design passes simulation but shows a parametric shift on silicon. Walk through what you would check on the design side, and what you would ask the process side. A good answer respects both views instead of blaming one.

Can you move between fabless, IDM, and foundry roles?

The public record here is thin, so take this as anecdote. A search summary of Blind threads describes a fab process engineer who wanted to move toward design, with replies suggesting an internal transfer or a master's degree, and another poster who worked on technology development and found fabless postings asking for design-tool experience they lacked. The sample is tiny, self-selected, and possibly out of date, and we cannot verify any of it.

What it does fit is a reasonable pattern: moving between design and process is possible, but the tool and project experience on your resume is what a hiring manager reads. Roles that sit at the boundary, such as design enablement, device modeling, product engineering, and yield, may be the easier bridges. Our hybrid skills article and the later posts on process, device, and integration roles and yield versus failure analysis explore that boundary.

How to decide

A business model is not a job. A fabless company can have a team that works closely on process issues, and an IDM can have a design team that rarely meets the fab. Ask each team what you would own, who you would work with weekly, and how the work reaches silicon. Use the questions to ask interviewers and the offer comparison checklist to put the answers side by side, and watch the red flags guide for vague replies.

If you want to rehearse the explanation of where your skills fit, you can book a mock interview on MockVise.

Sources: NVIDIA FY2020 10-K, Qualcomm FY2025 10-K, TI worldwide manufacturing, Micron FY2025 10-K, TSMC company profile, Intel Foundry, Blind thread (search result summary). This is general preparation guidance, not an official process or reported company questions. Company descriptions reflect the cited documents and may have changed, and candidate reports are anecdotal and may be out of date. Confirm details with your recruiter. MockVise is not affiliated with or endorsed by NVIDIA, Qualcomm, Texas Instruments, Micron, TSMC, Intel, Samsung, or Blind.

FAQ

Q: What is the difference between fabless, IDM, and foundry companies? A fabless company does not directly manufacture its chips, as NVIDIA and Qualcomm describe. An IDM such as TI or Micron manufactures in its own facilities, and a pure-play foundry such as TSMC focuses exclusively on manufacturing its customers' products.

Q: Do fabless engineers need to know about foundries? Foundry knowledge helps most when it explains a decision, for example why a particular constraint exists. A fabless designer typically works with a process through its design kit and rules.

Q: What do foundry interviews emphasize? Foundry interviews tend to probe process physics, data analysis, troubleshooting, and statistics. A good approach is to define the change, check the measurement, list competing explanations, and say what evidence would separate them.

Q: Which roles make it easier to move between design and process? Roles at the boundary, such as design enablement, device modeling, product engineering, and yield, may be the easier bridges. The post notes the public evidence is thin and based on anecdote.