I am a Computer Science graduate and work in Software and DevOps Engineering and I don’t personally thing CS is dying like the market believes…. but hiring is much more competitive until companies the ROI and product quality on AI is not as high as human engineers. The better way to say it is this: computer science is not dead, but the easy version of the computer science career path is. The “learn to code, get a six-figure job, work remote, and job-hop every two years” era has been violently corrected by layoffs, AI tooling, interest rates, outsourcing, and a flood of new graduates who all heard the same advice for the last decade.
At the same time, another engineering market is doing something very different. Electrical engineering, computer engineering, semiconductor manufacturing, chip design, embedded systems, power systems, controls, verification, and hardware-adjacent software are becoming more valuable because the world is realizing something obvious: software does not run without hardware. AI does not run without GPUs. Data centers do not run without power. National security does not run without chips. And every company talking about AI infrastructure eventually runs into the same bottleneck: someone has to design, manufacture, test, power, cool, and maintain the physical systems underneath it all.
That is where the discrepancy is becoming hard to ignore.
The Bureau of Labor Statistics still projects software developers, QA analysts, and testers to grow 15% from 2024 to 2034, with about 129,200 openings per year. So no, software is not “dead” in any rational economic sense. But that number hides the pain at the entry level: the job market can grow overall while still being brutal for new grads competing against experienced engineers, offshore teams, AI-assisted productivity, and companies that no longer want to train juniors the way they did during the zero-interest-rate hiring boom.
Meanwhile, the semiconductor industry has a much more concrete workforce problem. The Semiconductor Industry Association projects the U.S. semiconductor workforce will grow by nearly 115,000 jobs by 2030, but roughly 67,000 of those roles could go unfilled at current degree-completion rates. That gap is not just software. It breaks down into technicians, engineers, and computer scientists, with an estimated 27,300 engineering jobs and 13,400 computer science jobs at risk of going unfilled.
That is the difference. In software, the conversation is often, “How do I stand out from 1,000 applicants?” In semiconductors, power, embedded systems, and hardware, the conversation is increasingly, “Where do we even find enough qualified people?”
TSMC is a perfect example. Its career materials explicitly call out electrical engineering, physics, materials, mechanical, and automation engineering as the kinds of backgrounds it wants. That tells you something about where the bottleneck is. These are not jobs you can fake with a weekend bootcamp or a few LeetCode problems. Semiconductor manufacturing needs people who understand devices, process control, yield, cleanrooms, automation, reliability, power delivery, instrumentation, and the messy reality of building physical systems at scale.
Micron is in the same category. As a memory manufacturer, it needs engineers across electrical, computer, materials, process, product, firmware, validation, and manufacturing disciplines. The demand is tied not only to consumer electronics but also to AI, data centers, automotive systems, defense, and high-performance computing. In other words, the demand is not just “we need more apps.” It is “we need more infrastructure for the entire digital economy.” Micron’s own careers page reflects this broad technical hiring need across its global operations.
This is why I think electrical engineering and computer engineering are becoming more strategically valuable than people realize. CS became popular because software scaled beautifully. One engineer could write code that reached millions of users. That is still true. But the problem is that software talent also scaled. Universities pumped out more CS grads. Bootcamps sold the dream. Self-taught developers entered the market. Remote work made the applicant pool global. AI tools made average developers faster. The barrier to entry dropped.
EE and CE did not experience that same kind of flattening. You cannot fully virtualize a lab. You cannot vibe-code your way through signal integrity, semiconductor physics, FPGA timing, power electronics, RF, PCB layout, embedded firmware debugging, or manufacturing yield problems. AI can assist with these things, but it cannot easily replace the judgment that comes from understanding real-world constraints.
That is the part people miss when they say, “AI is going to replace engineers.” AI is much better at generating text and code than it is at owning consequences. It can write a React component. It can scaffold a Python script. It can help debug a Kubernetes manifest. But when a fab tool goes down, a power rail is unstable, a board fails compliance testing, a chip has thermal issues, or a production line loses yield, someone still has to understand the system deeply enough to make a real engineering decision.
Computer engineering sits in a particularly interesting middle ground. CE students usually touch both worlds: low-level software and hardware. They understand programming, but they also understand architecture, digital logic, embedded systems, microcontrollers, operating systems, and sometimes VLSI or FPGA design. That makes them harder to commoditize than a generic software candidate whose resume is just “React, Node, Python, AWS.” The closer you are to the machine, the less replaceable you become.
This does not mean everyone should abandon CS and run to EE. That would be an overcorrection. CS is still one of the most powerful degrees you can have if you use it correctly. The issue is that the market no longer rewards “I can code” by itself. That skill is becoming table stakes. The CS grads who will still win are the ones who can pair software with something harder: infrastructure, security, distributed systems, AI systems engineering, robotics, embedded systems, cloud cost optimization, data engineering, hardware acceleration, product intuition, or domain expertise.
In other words, CS is not dead. Generic CS is dead.
The same thing happened to IT. At one point, knowing how to image a laptop or reset a password was enough to get into the field. Then the field matured. Now the valuable people are the ones who understand cloud, automation, networking, identity, security, observability, compliance, and cost. Software engineering is going through that same maturation cycle. The easy layer is getting automated. The valuable layer is moving up and down the stack at the same time: closer to business outcomes on one side and closer to hardware/infrastructure on the other.
The BLS projections actually support this more nuanced view. Software roles are still projected to grow faster than average, but so are electrical and electronics engineers, which are projected to grow 7% from 2024 to 2034, with about 17,500 openings per year. Computer hardware engineers are also projected to grow 7%, with about 4,700 openings per year. Those numbers are smaller than software in absolute terms, but the talent pools are also smaller, the skill requirements are more specialized, and the strategic importance is rising.
The semiconductor shortage conversation also proves that the United States cannot simply “software” its way out of every problem. The CHIPS and Science Act put tens of billions of dollars toward rebuilding domestic semiconductor research, development, and manufacturing capacity. But money alone does not create qualified engineers overnight. Fabs require people. Packaging facilities require people. Test engineering requires people. Process engineering requires people. Tool maintenance requires people. The workforce pipeline is now one of the biggest constraints.
That should be a wake-up call for students choosing between CS, EE, and CE.
If you want the broadest possible job market, CS still gives you that. If you want to work in software, cloud, AI, cybersecurity, data, or startups, CS is still a great path. But you cannot treat the degree like a golden ticket anymore. You need proof that you can build, ship, operate, and reason through systems better than someone using the same AI tools as you.
If you want a harder but potentially more defensible path, EE and CE deserve more attention. They sit closer to the physical world, and the physical world is where AI has a much harder time pretending. Chip makers like TSMC, Micron, Intel, Samsung, GlobalFoundries, and Texas Instruments need people who can work across physics, manufacturing, automation, embedded software, power, reliability, and systems. That work is less glamorous than building an app, but it may become more stable and more strategically important over the next decade.
My take is simple: the “CS is dead” narrative is lazy. What is dead is the belief that every CS major is automatically entitled to a great software job just because they completed the degree. The market is forcing CS grads to specialize, build deeper systems knowledge, and prove they can create value beyond writing boilerplate code.
EE and CE are not magic shields either. They are difficult degrees. The work can be less remote-friendly. The industries can be slower-moving. Hardware has longer timelines, more regulation, more capital expense, and less instant gratification. But those same factors are why the roles are harder to flood and harder to automate away.
The future probably belongs to the hybrids: software engineers who understand infrastructure and hardware constraints, computer engineers who can write production-grade software, electrical engineers who understand automation and AI, and CS grads who are not afraid to leave the comfort of pure web development.
So no, being a computer science major is not dead.
But being a generic computer science major might be. And if the last decade belonged to software eating the world, the next decade may belong to the people who understand what software runs on.
Postmortem: What We Got Wrong About CS, EE, and CE
The biggest mistake the industry made was treating Computer Science like an infinite money machine. For years, students were told that if they learned to code, got a CS degree, or built a few projects, the market would take care of the rest. That was mostly true during the tech boom, but it created a false sense of security. Companies overhired, universities overmarketed CS, bootcamps sold unrealistic outcomes, and students entered the field believing demand would always outpace supply.
Then the market corrected.
Interest rates went up. Venture capital got tighter. Big Tech cut headcount. Companies became less willing to train junior engineers. AI tools made existing engineers more productive, or at least gave executives a reason to believe they could slow hiring. Suddenly, the entry-level CS market became brutally competitive. The degree did not become useless, but the advantage became weaker.
The second mistake was assuming all engineering talent is interchangeable. Software engineering, electrical engineering, and computer engineering are connected, but they are not the same labor market. A company can hire more software developers from a global applicant pool. It can outsource app development. It can use AI to generate boilerplate code. But it cannot magically create experienced semiconductor process engineers, power engineers, verification engineers, embedded systems engineers, or hardware validation engineers overnight.
That is where the market is now exposing the gap.
We spent the last decade telling everyone to become software engineers while underestimating the physical infrastructure behind the software economy. AI needs chips. Chips need fabs. Fabs need electrical engineers, computer engineers, materials experts, technicians, manufacturing engineers, process engineers, and automation specialists. Data centers need power systems, cooling, networking, and hardware reliability. None of that goes away because a chatbot can write code.
The third mistake was confusing short-term hiring pain with long-term career death. CS is not dead. Software is not dead. DevOps is not dead. AI is not replacing every engineer. But the lower end of the software market is being squeezed. The people who only know how to build CRUD apps, copy tutorials, or rely on frameworks without understanding systems are going to have a harder time. The people who can design, debug, secure, scale, and operate real systems will still be valuable.
The lesson is not “don’t major in CS.” The lesson is “don’t be generic.”
For students, that means choosing a direction earlier. Pair CS with cloud, security, distributed systems, AI infrastructure, embedded systems, data engineering, or hardware. For EE and CE students, it means realizing they may be entering a market with less hype but stronger structural demand. For universities, it means stop selling every technical student the same software dream and start rebuilding serious pipelines into semiconductors, embedded systems, power, and manufacturing technology.
For companies, the lesson is even more obvious: you cannot complain about a shortage of hardware and semiconductor talent after spending years optimizing recruiting around software roles, LeetCode screens, and generic tech hiring pipelines. If TSMC, Micron, Intel, Samsung, GlobalFoundries, and the rest of the chip ecosystem need more qualified engineers, then industry has to invest in internships, apprenticeships, lab partnerships, technician pipelines, and real entry-level training.
The final takeaway is this: software ate the world, but hardware feeds software.
The companies that understand both sides will win. The engineers who understand both sides will be harder to replace. And the students who see the shift now will be in a much better position than the ones still chasing the 2021 version of the tech job market.