AI Is Changing Entry-Level Work in Chip Design, but Demand for Electrical Engineers Remains High
- Brian Bailey
- 994 reads
AI tools are absorbing some junior tasks in chip design, raising concerns about training while creating new roles for early-career engineers.

AI is changing the work assigned to new engineers in electronic design automation (EDA) and chip design, but the available evidence doesn’t point to a broad decline in demand for electrical engineering graduates. The sharper change is in what employers expect those graduates to do: routine tasks are increasingly handled by AI tools, while new hires are expected to take on higher-level problem solving sooner.
That shift creates a training problem. Junior work has long given graduates a way to learn practical engineering, with experienced colleagues guiding them through real designs. If companies automate those tasks and senior staff have less time to mentor, fewer graduates may gain the experience needed to become the next generation of experts. At the same time, early-career engineers may be well placed to help colleagues adopt AI tools.

Clockwise from top left: Southampton University’s Davidmann; Breker’s Kelf; Synopsys’ Andersen; ChipAgents’ Kexun; llmda’s Gupta; and Cadence’s Alpert.
Different pressures on software and chip design
Hiring concerns among computer science graduates don’t necessarily describe the market for chip designers. Nagesh Gupta, CEO of llmda.ai, said his experience points to data science graduates facing more difficulty than computer science graduates. He has been seeking computer science graduates from strong universities and hasn’t found many available candidates.
Chuck Alpert, a fellow at Cadence, sees a reason for caution in software hiring: tools such as Cursor, Claude Code and GitHub Copilot can increase the productivity of software developers. Companies paying for tools from Claude and OpenAI may decide they need to hire fewer engineers, or to slow hiring while they assess how much of the work can be automated.
Thomas Andersen, vice president of AI and machine learning at Synopsys, draws a distinction between people who build EDA software and engineers who use those tools to design chips. Software teams are testing how AI can shorten development cycles and help them deliver releases faster. Chip designers, by contrast, are specialists in hardware design rather than conventional software programmers. Andersen said the industry has faced a shortage of those specialists for some time. AI may help them work more efficiently and produce more designs, but it doesn’t remove the need for human expertise.
That difference matters when interpreting reports of graduate hiring difficulties. Simon Davidmann, an AI and EDA researcher at Southampton University, cited figures for the UK graduate market: recruitment among leading employers fell 24% from 2022 to 2026. He said the decline reflects several pressures, including economic uncertainty, reduced budgets, market corrections and changes in hiring after the pandemic. AI has entered a job market that was already shifting, making its specific effect hard to isolate. Davidmann said 75% of graduate engineers find employment within 15 months of graduation, a result he described as typical.
These observations don’t establish a single cause for changes in graduate employment. They do suggest that outcomes vary by field and that broad measures of computer science hiring shouldn’t be treated as a direct proxy for demand in electrical engineering or chip design.
When automation removes the first rung
Dave Kelf, CEO of Breker Verification Systems, said a lack of practical experience can make it harder for graduates to find work. New hires may need substantial time working on real designs before they can contribute independently. If AI takes over tasks that once gave interns and junior engineers that experience, graduates could find it harder to get internships and arrive at their first job with less hands-on preparation.
There’s also a business pressure behind that change. If a company can assign AI tools to work previously given to junior staff, it may be difficult to justify hiring for those roles when competitors are using automation to control costs. The immediate productivity gain can therefore weaken the pathway through which graduates learn the practical side of engineering.
Davidmann sees this as a potential long-term skills gap. Engineering requires people to learn and apply skills that AI can’t supply on their behalf. Without enough graduates entering the profession and building experience, today’s experts may have fewer successors to take over their work. Andersen agreed that automating entry-level tasks doesn’t answer who will replace experienced engineers later.
Mentoring is another part of the problem. Gupta described a recent computer science graduate in his family who found work and reported that most of her graduating class had also secured jobs. Her observation was that new hires were finding it harder to get time with experienced colleagues. Senior engineers increasingly use AI for their own work, leaving less space for the traditional arrangement in which an experienced mentor brings a graduate up to speed. The employer she joined had begun a focus group to work out how to address that gap.
This raises a practical question for design organizations: how can they capture productivity gains while still giving new engineers enough contact with real engineering problems and experienced staff? The source of the risk isn’t only the technology’s ability to perform a task. It’s also the way teams distribute the work and preserve opportunities to learn.
Employers want AI fluency and stronger judgment
Kexun Zhang, head of research at ChipAgents, described a mismatch between how graduates have been trained and what some employers now expect. Companies adopting AI want new hires to use those tools to solve higher-level problems, define what needs solving and exercise judgment about possible approaches. They may no longer expect graduates to spend as much time on the routine work traditionally assigned to interns or junior engineers.
Universities have had little time to adjust their programs to the pace of AI development. Zhang said many recent graduates received much of their education before AI became a routine part of engineering work, even as employers have begun to value people who can use it effectively. That puts this graduating cohort between two sets of expectations: the training they received and the work companies now want them to perform.
Davidmann sees education as needing two complementary modes. Students need a foundation in mathematics, algorithms and debugging so they understand what a system is doing and can identify when it fails. They also need professional preparation that includes using AI agents as part of engineering work. These modes serve different purposes: one builds the ability to reason about a design, while the other prepares students to work with tools they are likely to encounter in industry.
He described a university project developed with technology supplied by a large EDA company. Students worked on verification and coverage. They first set up the simulation, test generators and coverage process manually, then explored how AI could improve the work. By learning the workflow before applying AI, they could assess the tool in the context of a process they understood.
That approach also points to a challenge for educators. Universities are still working out how students should use AI in coursework and how to assess whether they have learned the underlying skills. Davidmann said instructors need to understand these tools and establish ways to teach both foundational knowledge and practical AI use. Companies, too, are still learning how staff should incorporate AI into engineering tasks.
Early-career engineers can teach experienced teams
AI adoption may also give junior engineers a role that runs in the opposite direction. Alpert said engineers hired within the previous two years are often among the most useful people to consult about AI. In an industry with many experienced specialists, younger colleagues can bring practical knowledge of tools and methods that senior engineers haven’t yet adopted.
Andersen said that willingness to learn new technology is one of the strengths younger employees can bring. He also sees a gap between what universities have taught recent graduates and what companies expect them to know about AI. Even so, their interest in new tools and capacity for self-directed learning can help them contribute to adoption across a team.
Kelf compared the present transition with earlier changes in EDA. When logic synthesis became part of design practice, some senior engineers accustomed to working at gate level were slower to adopt it. Newer engineers learned hardware description languages and helped bring synthesis into use. A similar pattern appeared in randomized verification: newer graduates were more familiar with object-oriented programming and test benches, while some experienced verification engineers had to learn those practices on the job.
AI could create another such handover, though Kelf cautioned against assuming that graduates are already experts. New engineers may become more comfortable with AI tools as their education changes, while experienced staff learn how to work with them through practice. Understanding where AI can produce unreliable results, including hallucinations, will be part of that learning.
More output may keep demand for designers high
The prospect of AI agents performing tasks once assigned to junior engineers raises the question of whether electrical engineers will also face reduced demand. Alpert’s view is that productivity gains will meet a growing workload. He pointed to increasing design complexity, including packaging and 3D-ICs, and the need to verify across more fields. He described the resulting workload as roughly 10 times more complex, while noting that the industry wants to produce more designs without adding equivalent resources or time.
For Alpert, that workload leaves room for AI to increase productivity without eliminating the need for design engineers. EDA has long focused on helping teams accomplish more, and agents add another way to close the gap between available engineering capacity and the work required. In his view, demand for designs is likely to remain high even if individual designers become much more productive.
Andersen likewise said demand for chip designers exceeds supply. He treats “junior” and “senior” as descriptions of the tasks people perform, rather than fixed stages that require a set number of years. If AI automates some junior tasks, engineers can move toward higher-level work earlier in their careers.
That progression depends on more than assigning agents to routine tasks. Graduates need a foundation that lets them judge AI output, practical exposure to real designs and access to experienced colleagues. Experienced engineers need time to mentor as well as opportunities to learn from newer staff. The engineering teams that manage both sides of that exchange will be better positioned to gain productivity while continuing to build expertise.