Why This Field Matters

At least 22 professors and researchers have left elite American universities this year, or taken leave from them, to join OpenAI, Anthropic, Meta and Google DeepMind. Anthropic hired the chair of UC Berkeley’s electrical engineering and computer science department. At least three computer science professors joined Meta’s AI lab in late June. Losing a department chair does not just scramble a teaching schedule. Doctoral advising, curriculum revision and faculty search committees stall at the same time.

The student side has turned too. The 2025 CRA Taulbee Survey reported record doctoral degree production alongside the first clear contraction in new undergraduate majors and total enrollment. Output is at a high while intake thins. The postwar assumption that a degree leads to an office job is under review, and computing is no longer the exception that proves it still works.

The pull is not only frontier labs. In memory and foundry manufacturing the same competition runs company to company. Samsung’s union said more than 200 members moved to SK Hynix after April, an internal survey in June found 81.5 percent of foundry staff wanted to change employers within two years, and SK Hynix added 2,152 people in the first half of the year alone. Wherever a scarce skill gets repriced, teaching that skill gets repriced with it.

For anyone who teaches computing, both currents arrive at the same question. The decision worth making is not stay or leave. It is knowing which terrain pays for teaching, because academia, corporate enablement, bootcamps and joint appointments ask for different muscles.

Required Skills

The base is still algorithms and systems, but every one of these four routes now demands the willingness to throw material away. Models and toolchains turn over in months. Anyone teaching from three-year-old slides loses the room. Deciding what to discard requires separating the concept layer from the tool layer when the course is designed, not when it breaks.

• Running lab infrastructure. GPU allocation, cloud credits, sandboxed environments. Whoever sets these up and watches the bill is often solving the real bottleneck, not the lecture.
• Assessment that assumes the tools are on. Design tasks and checkpoints that reveal what a learner can do with a coding agent open next to them. Submitted artifacts alone no longer answer that.
• Verifiable industry time. Corporate education programs hire on shipping history more than on credential type. A record of having built and operated the thing being taught travels further than another certificate.
• Writing that outlives the session. In developer relations and internal enablement the deliverable is documentation and sample code, not a talk. Scale comes from writing that keeps being read.
• Program and credential design. Translating a curriculum into course credit, an internal job certification or a completion standard is what gets a budget attached to it.

Certifications matter less here than a track record. The strong candidates can say how many people finished a program they designed and what those people could do afterward.

Career Path

Staying in academia is not the closed door it appears to be, but the openings have moved. Teaching-track and professor-of-the-practice lines, industry-funded institutes, and community college programs are hiring while tenure-track searches in some departments sit unfilled after the senior person leaves. A department that just lost a chair to a frontier lab has a course load nobody is covering, and that gap is where a teaching-focused appointment gets created.

Corporate enablement is the fastest-growing employer of computing educators. Internal AI academies, engineering onboarding, and developer education inside product companies all need people who can turn a live system into a curriculum. Titles cluster around developer education, solutions architecture, and forward deployed engineering, where teaching and implementation sit in the same job. The learners are practicing engineers, so the work is less about explaining concepts and more about converting production problems into exercises.

Bootcamps and edtech have lost the product they were built on. When junior hiring contracts, a job-placement curriculum sells poorly to individuals. The contracts still signing are corporate reskilling programs that run on the client’s own data and internal tools, and those buy program designers rather than instructors.

Joint appointments may be the most durable option. MIT has long appointed senior industry practitioners as professors of the practice, and versions of that arrangement, from adjunct teaching to formal cross-appointments, let someone keep shipping while keeping a classroom. Choosing neither side outright and accumulating both records is the path least likely to close.

Any of the four can be tested inside one term. Rebuild the lab environment for a single course around the toolchain the industry actually uses, then write down why it changed and where students got stuck. That document works as a course proposal, as a corporate education application, and as a developer relations portfolio without being rewritten.