CGL 集团,中国领先的高管寻猎与领导力咨询集团
By George Guo, Senior Vice President, CGL Group
Hiring inside a chip company is rarely a sourcing problem. The candidates exist. The failure is putting a person on the wrong gate.
A chip goes through six of them before it reaches volume: whether the money holds, what you are building, whether it can be manufactured, whether it is correct, whether it can be sold into the target market, and whether the organisation can keep up. Each gate needs a different person. Move someone one gate sideways and they usually cannot do the job, even with a strong CV.
I have seen companies with a complete-looking team stall at the fourth gate: the design was right, the tape-out failed. I have also seen all six gates pass and the last one collapse, with half the core team gone the week before volume production.
What follows is one role per gate, and only the part that matters.
Gate One: CFO. The bet, its size, and how long it stays open
A CTO is hard to find because the pool is small. A CFO is hard to find because the wrong ones look right. The people are not scarce; the experience that transfers is.
The financial model of a chip company is structurally different. Tape-out is a one-off, heavy-asset bet: the money is committed, the outcome arrives at a single point in time, and the risk cannot be spread. Inventory cuts both ways, an asset when supply is short, an impairment when prices fall. Capital expenditure follows equipment lead times, so cash leaves before demand is confirmed. Government grants and industry funds are part of the revenue structure, and how they are recognised against R&D spend changes the reported numbers.
An audit background does not automatically cover this. Neither does an investment banking background. One question separates them: after a failed tape-out, how long until the money comes back.
There are three common sources. Semiconductor finance people understand tape-out, inventory and grants, but often lack the capital markets narrative and struggle at fundraising. Banking and private equity people can run a financing or an acquisition, but lack manufacturing cost intuition. Big Four audit people are solid on controls and compliance, and tend to turn finance into after-the-fact bookkeeping.
Use three questions. How do you calculate the payback period after a failed tape-out. Have you owned a build and stock decision, and was it driven by orders, lead times, or a customer's verbal commitment. In a year when grants step down, how do you resequence R&D spend. Anyone who stalls on the first question has probably never owned a chip budget.
If the company is fabless and does not carry tape-out risk itself, the financial complexity drops a level and a finance director is enough.
Gate Two: Product Line Lead. Defining what gets built
This role does not own schedule. It owns the definition: which specification, which price band, which customer segment, when to launch, and when to kill.
What makes it unusual is the design-in cycle. In consumer electronics, contact to volume runs six to twelve months. In automotive, it runs in years. You hold a judgement for three years with no revenue and stay calm when the customer changes requirements in year two. That single condition eliminates a large share of candidates from fast-cycle industries.
Three sources, three gaps. An FAE moving over understands real customer pain and has the technical base, but lacks pricing and portfolio thinking. A hardware lead from a system house knows how the box gets specified, but underestimates supply-side constraints. A product manager from a large international vendor has the method but not the local resources or speed.
Ask how many design-ins they have run and how long each took from contact to volume. Anyone who cannot describe the full cycle usually saw only one phase. Ask which product line they killed and why; people who only add and never cut will dilute resources until nothing is funded. Ask what they did when a customer changed requirements, and listen for whether they separated a parameter change from an application change. Ask what the switching cost of their product was. If they cannot say, their pricing was a guess.
If the portfolio has one main line and the specification is set jointly by the founder and the technical lead, you do not need this role. You need a senior product manager who knows the customer.
Gate Three: Process and Integration. Yield, not feasibility
In the early phase of import substitution, customers ask whether you have the part at all. Product people are worth more then, because defining the device and winning design-ins is the whole game. In the middle phase, customers ask how good it is and how cheap it is. That is when process people matter: on the same line, five points of yield is a visible cost advantage, and it is what keeps you alive in a price war.
So it is not a choice between the two profiles. It is a question of phase, and most segments in China have already moved from whether to how good. The tighter gap today is on the process side.
Process expertise accumulates on an experience curve and does not transfer across nodes. A person needs one complete cycle on a node, from ramp to volume. Ask which nodes they worked and which ones they took all the way through; partial coverage does not count. Ask where yield started, where it ended, and which step was the bottleneck. Only people who can name the bottleneck have actually moved a yield number. Ask about a line excursion during volume production and how they localised it. That is the core skill, and it matters more than theory.
The product-side test is different. How many products they defined actually shipped in volume. What the opportunity they missed looked like in hindsight. How they tell a real customer requirement from a stated one.
Gate Four: Verification and DFT. Whether the first tape-out works
These are routinely merged into one role and they are not the same job. Verification proves the design is functionally correct, using UVM, formal methods and coverage closure. DFT makes the manufactured die testable, using scan chains, built-in self-test and automatic test pattern generation. One owns correctness, the other owns screenability.
One person can carry both at small scale. Once the team passes a certain size, or once several projects run at the same time, the priorities of the two keep colliding.
Candidates from communications and consumer electronics have two real gaps. The first is depth on timing closure: judgement formed in a wide-margin environment has to be rebuilt for the very narrow margins of advanced nodes. The second is a reliability verification system, because automotive qualification requires proving not only that the design works, but that it still works under extreme conditions.
Ask how many tape-outs they have been through and how many were first-pass successes. First-pass rate is the hardest single indicator. Ask what coverage number counts as enough and what the number is based on; people who give a figure without a rationale have usually not been burned by an escape. Ask for the latest bug they found late and how they found it. Ask how they handle conflict between verification and design teams, because that is a daily event in this role.
Whether to split the two roles depends not on headcount but on two facts: how many projects run concurrently, and whether automotive-grade work requires both sides to be near-perfect. A company with one product family and no more than two tape-outs a year can combine them.
Gate Five: Functional Safety. The gate that decides whether you sell into automotive
The common mistake here is hiring by industry. The shared language of this role is the standard, not the sector. A functional safety lead in automotive electronics and one in industrial control or robotics speak the same language: hazard analysis, safety integrity levels, failure modes. The candidate pool is therefore larger than it looks, because people can move laterally; and smaller than it looks, because very few have led a complete certification.
The deliverable on paper is four documents: safety goals, safety level decomposition, the safety case, and responses to the assessor. The difficulty is not writing them. It is that any one of them can be sent back by a department over which this person has no formal authority. The role needs coordination power, not writing skill.
Three sources. OEM functional safety departments understand the vehicle context and how customers audit, but lack device-level implementation detail. Tier one suppliers have complete projects and mature systems, but a narrow cross-industry view. Industrial control and robotics bring strong standards discipline and need to learn automotive process and culture.
A few questions sort the levels. How many complete certifications have you led, and at what safety integrity level; leading and participating are different claims. When one block of the safety goal decomposition had no owner, what did you do. Which corrective action from the assessor was hardest to close, and why. How do you decide a design is safe enough; only people who can answer that know where enough is. The real difficulty in functional safety is organisational, not technical.
If the company does not sell into regulated markets such as automotive, industrial safety or medical, the role is unnecessary. Many companies only discover they need it one quarter before a customer audit, which is too late to hire.
Gate Six: HRVP. Why the tape-out cycle has to be understood
An HR leader who does not understand the tape-out cycle produces hiring plans that are worthless in practice.
Talent demand inside a chip company moves with the node. Architecture and front-end design need architects and design engineers. Verification needs a full verification and DFT group standing up. Tape-out puts maximum pressure on layout and back-end. Before volume, test, process and reliability must all be in place. Miss the node and the hires arrive without work; hire too early and you carry people with nothing to do. That sense of rhythm is the real difference between a chip HRVP and a generalist.
Two other things have to be understood. One is project-based organisation: a project runs from tens to hundreds of people, teams form across functions and disband at the end, which is not how a functional structure works. The other is the dual technical ladder: many chip engineers do not want to manage, and the technical ladder has to carry equivalent compensation and voice. Companies that cannot build it keep losing senior engineers.
I would ask how much of the product roadmap they can describe; if they cannot say roughly where the next chip sits, they do not yet know the business. Ask how they sequence a year of hiring, by departmental requests or by project nodes. Ask how the technical and management ladders are structured and how conflicts between them are resolved. Ask whether they have lost a core team and what the reason turned out to be. In this industry, core team losses are rarely about money.
Below a hundred people and one project a year, the role can sit with the COO or the founder. Created too early, a dedicated HRVP degenerates into a recruiting manager, because there is no organisation design work to do yet.
What each gate actually tests
Gate one, the CFO: how long until a failed tape-out is repaid. Sources are semiconductor finance, banking, and Big Four audit. The pool is narrow.
Gate two, the product line lead: how many design-ins, over how long. Sources are vendor FAE, system-house hardware, international product management. Very narrow.
Gate three, process and integration: yield from where to where, and where it stuck. Sources are fabs, IDMs and OSATs. Extremely narrow.
Gate four, verification and DFT: how many tape-outs, how many first-pass. Sources are communications, consumer and automotive electronics. Narrow.
Gate five, functional safety: how many complete certifications led, at what level. Sources are OEMs, tier ones, industrial control and robotics. Very narrow.
Gate six, HRVP: how the year of hiring is sequenced. Sources are semiconductor firms, manufacturing, and HR business partners from large technology companies. Narrow.
If you can only close one gap, close process and integration. If you can close two, add functional safety, because it is a hard gate on customer access.
How to write the JD for these six roles
A JD is a list, but writing one involves judgement. One line each: what to write, and which sentence to delete.
Q: How should the chip CFO job description be written?
A: Write three things: the size of the tape-out budget this person signs off, the capital structure they have managed, and the three-year story they will tell investors. Keep requirements verifiable: how many tape-out budget cycles they owned, and what payback period they assume after a failure. Delete "familiar with semiconductor finance", which anyone can write and which screens nobody.
Q: How should the product line lead JD be written?
A: Start with the revenue and gross margin base of the line, then state which part of the design-in process they own, then state whether they hold kill authority. Replace "rich product experience" with how many of their defined products shipped in volume.
Q: How should the process and integration JD be written?
A: Three things must appear: which process segment, which node, and the yield range to be moved. Missing any one of the three removes all screening power. Delete "proficient in semiconductor process flows" and write which nodes they took from ramp to volume.
Q: How should the verification and DFT JD be written?
A: State the number of tape-outs and the first-pass ratio, the coverage standard and its rationale, and whether this is one role or two. "Familiar with UVM and formal verification" is meaningless; tools are an entry ticket, not a differentiator.
Q: How should the functional safety JD be written?
A: Three things decide the size of the pool: which standard must be met, which safety integrity level, and whether the person led or participated. Delete "familiar with ISO 26262", because people who have led a full certification and people who have never read the standard both write that line.
Q: How should the chip company HRVP JD be written?
A: Do not write "recruiting experience in semiconductors". Write how many projects they followed through a complete tape-out cycle, whether they built a dual technical ladder, and the team size and attrition they managed. The first decides whether they understand nodes; the other two decide whether they keep people.
Frequently asked
Does a chip company need a CFO? Asset-light companies with no external funding plans can run with a finance director. If you carry tape-out risk and plan capital expenditure, you need a CFO.
Can a chip CFO come from the internet sector? Yes, but two tests apply: whether they accept that one failed tape-out is several quarters of profit, and whether they understand that inventory cuts both ways. Failing either, they will keep making the wrong calls.
Does a product line lead need a technical background? Deep enough to judge whether a customer will pay more for a given specification. Not able to draw circuits, but able to hold a conversation with engineering.
After import substitution, is the gap in process or product people? It depends on the phase. Early phase needs product people to define the device; middle phase needs process people to deliver yield and cost. Most segments are in the middle phase now.
Does a functional safety lead have to come from automotive? No. The shared language is the standard. Industrial control and robotics backgrounds transfer, with a gap in automotive process culture.
Which of the six is hardest, and how long does it take? Process and integration is the narrowest, because node experience must be complete. Product line leads come next, because the cycle demands patience. Search to start typically runs 45 to 60 days; process and functional safety run longer, and many only move on a long-standing relationship.
Contact
The first step in a search like this is placing the company on the six gates and identifying which one it is stuck on. The same title means something completely different during tape-out and during volume production. Fix the gate first, then the person.
For chip and semiconductor roles, CFO, product line lead, process and integration, verification and DFT, functional safety and HRVP, you can reach me directly. Include the company's current stage and the target role, and the reply will be faster.
Email: george.guo@wearecgl.com
LinkedIn: linkedin.com/in/gg6561771
George Guo is Senior Vice President at CGL Group. He is a startup mentor at Zhejiang University, a former partner at a venture fund, and a co-initiator of the AI Native Organization Research Institute. He holds an undergraduate degree from McGill University and a master's degree in computer science from the University of Waterloo.
Public profile: https://global.wearecgl.com/adviser/detail/243