Why Aerospace and Deep Tech Is the Next Big Thing After AI

Every engineering student I talk to has the same question, phrased slightly differently: did I miss it?

They watched the 2023–2025 AI hiring wave from a dorm room. They watched compensation at frontier labs detach from reality. And now they're wondering whether the interesting decade already happened without them.

It didn't. But the center of gravity moved, and it moved somewhere most students aren't looking.

The short version: artificial intelligence didn't slow down. It ran into the physical world. Everything the next phase of AI needs — electricity, transformers, cooling, satellites, chips, launch capacity, factories — has to be designed, built, and flown by people who work in atoms rather than tokens. That work is aerospace and deep tech. And the people who do it are in shorter supply than at any point in modern industrial history.

Here is the data.

AI didn't hit a ceiling. It hit a supply chain.

The four largest hyperscalers — Amazon, Google, Meta, and Microsoft — guided to somewhere between $635 billion and $700 billion in combined capital expenditure for 2026, up from roughly $387 billion in 2025. Around $240 billion of that goes to physical infrastructure: land, buildings, cooling, and above all power.

That money is not converting to compute as fast as anyone planned, because the constraint stopped being chips.

Microsoft has disclosed roughly $80 billion in Azure orders it cannot fulfill because the power isn't there.
Gartner projects 40% of AI data centers will be power-constrained by 2027.
Grid interconnection approval in major US markets now runs 24–36 months.
Large power transformers have 3–5 year lead times. Medium-voltage switchgear is effectively sold out through 2028 in many channels.
Analysts tracking the 2026 build pipeline found only about one in three announced megawatts had broken ground.

Read that list again as a job board rather than a risk report. Every line is a hiring signal for power systems engineers, thermal engineers, manufacturing engineers, and people who can build hard physical things on a schedule.

This is the structural point students miss. Software scaled by writing more software. AI now scales by building more world. The bottleneck moved from the model layer to the physical layer, and the physical layer is desperately understaffed.

The capital already moved. Quietly.

While AI absorbed the headlines, defense and space technology went through the fastest re-rating of any industrial sector in decades.

Anduril is the cleanest illustration. Valued at $30.5 billion in mid-2025, it raised $5 billion at a $61 billion valuation in May 2026 — a double in eleven months — and by late July 2026 was reported to be in talks at roughly $100 billion. That's a company founded in 2017 that builds hardware.

SpaceX went public in June 2026 at a valuation north of a trillion dollars, making it one of the most valuable companies on earth. Its Starlink division alone booked around $11.4 billion in 2025 revenue — about 61% of company revenue — and passed 10.3 million subscribers across 160 markets by March 2026, up from 4.6 million at the end of 2024.

Sector-wide, defense tech venture funding set an all-time record in 2025 and 2026 is running ahead of it. Estimates range from roughly $29 billion to $49 billion for 2025 depending on how tightly you define "defense tech" — but every source agrees on the direction and the slope. Shield AI raised $2 billion. Saronic raised $1.75 billion for autonomous naval vessels. Mach Industries raised $300 million.

And the underlying market forecast is not a startup pitch deck. The World Economic Forum and McKinsey put the space economy at $630 billion in 2023, growing to $1.8 trillion by 2035 — roughly the size of today's global semiconductor industry.

The reason all of this became possible: launch costs collapsed

If you want one number that explains the last fifteen years of aerospace, it's cost per kilogram to orbit.

The Space Shuttle put mass in orbit at roughly $54,000 per kilogram in today's dollars. A Falcon 9 rideshare now does it for under $3,000 per kilogram — a decline of roughly 90% in fifteen years, driven by reusable first stages that fly, land, and fly again. Starship, if full reusability works as designed, targets a further order of magnitude below that.

Cost curves like that don't just make existing businesses cheaper. They make entire categories of business exist for the first time. Earth observation, in-space manufacturing, satellite broadband, and orbital logistics were all economically absurd at $54,000/kg and are merely difficult at $3,000/kg.

This is the same dynamic that made cloud computing possible when server costs fell — except it's happening in hardware, and the talent pool that understands it is a rounding error next to the software industry's.

The talent math is genuinely absurd — in your favor

This is the part worth internalizing if you're choosing where to spend the next decade.

The US aerospace and defense sector employs roughly 2.23 million people. And it is running out of them:

29% of the A&D workforce is 55 or older and eligible to retire.
An AIA/McKinsey 2025 study found 76% of companies struggling to hire engineering talent, and 56% short on skilled trades.
Attrition in aerospace manufacturing runs around 15% — more than double any other sector.
More than 70,000 security-cleared positions sat unfilled across defense as of 2023, before the current spending acceleration.

Meanwhile the Bureau of Labor Statistics projects aerospace engineering employment growing 6% from 2024 to 2034 — faster than average — with about 4,500 openings a year and a median wage of $134,830.

Compare the competitive dynamics. A software engineering internship at a well-known AI lab might see tens of thousands of applicants for dozens of seats. A propulsion internship at a Series B launch company might see a few hundred. Same caliber of work. Wildly different odds.

The scarcity isn't because the work is unappealing. It's because the pipeline is small, the graduating cohort is small, and an entire generation of talent spent fifteen years being told the only ambitious career was software.

What "deep tech" actually means for a first job

"Deep tech" gets used loosely. Concretely, these are the hiring clusters worth knowing:

Launch and propulsion — SpaceX, Rocket Lab, Firefly, Relativity, Stoke Space. Engines, structures, avionics, GNC, test.

Satellites and space infrastructure — Planet, Varda, K2, Muon Space. Earth observation, in-space manufacturing, orbital servicing.

Defense autonomy — Anduril, Shield AI, Saronic, Castelion, Mach. Autonomy stacks, sensor fusion, munitions, unmanned systems.

Advanced manufacturing — Hadrian, Divergent, Nominal. The unglamorous, enormous problem of making precision parts fast enough to matter.

Energy and power — fusion, advanced fission, grid hardware, storage. Directly downstream of the AI power crunch described above.

Two things unify them: the product is physical, and the company cannot ship without engineers who can build. → Browse open aerospace and deep tech internships

The honest caveats, because nobody else will tell you

A thesis you can't argue against isn't a thesis. Four real drawbacks:

Citizenship restrictions are severe. A large share of aerospace and defense roles require US citizenship under ITAR or clearance eligibility. If you're an international student, this narrows the field substantially — not to zero, but you need to read postings carefully and target commercial-only work. → Read our guide to aerospace roles open to international students

Starting pay trails frontier AI. A $134,830 median is excellent by any normal standard and roughly half what a top AI lab pays a new grad. If maximizing near-term compensation is the goal, this is the wrong sector.

Hardware timelines are brutal. You may spend two years on a component that flies once. Software engineers ship weekly. Some people find the pace clarifying; others find it maddening. Know which you are before you commit.

Schedules slip. Starship slipped three times in 2026 alone. Programs get cancelled, funding gets reallocated, and physics doesn't negotiate. Optimism about the sector is not the same as optimism about any given company.

How to actually get in

The timing is unforgiving and almost nobody tells students this in time.

Aerospace and defense recruit earlier than any other engineering sector. Summer 2027 postings open August through October 2026 — roughly ten months ahead of a June start. SpaceX and Lockheed review on a rolling basis, which means applying within the first two weeks of a posting materially changes your odds. Fall engineering career fairs cluster in mid-to-late September. NASA's Summer 2027 deadline is February 26, 2027.

Three things that matter more than your GPA:

Join a build team. AIAA chapter, rocketry club, CubeSat program, Design/Build/Fly. Hiring managers in this sector read project work before coursework, because it's the only evidence you can actually make hardware.
Apply in the first fortnight. Rolling review is the single most exploitable fact about aerospace recruiting.
Track deadlines centrally. They're scattered across dozens of career pages and they close early. → See the Summer 2027 aerospace internship deadline tracker

The bottom line

AI is not over, and nothing here argues that it is. The argument is narrower and, I think, harder to dispute: the returns on AI now depend on physical infrastructure that doesn't exist yet, and the people who can build it are scarcer than the people who can train models.

Capital has already repriced the sector. Launch costs have already collapsed. A third of the workforce is heading for retirement. The demand is documented and the pipeline is thin.

That combination — real demand, thin supply, early in the curve — is what people mean when they say they wish they'd been in software in 2010.

Frequently asked questions

Is aerospace really the next big thing after AI? More precisely: aerospace and deep tech are where AI's next phase gets physically realized. Hyperscalers are spending $635–700 billion in 2026, roughly $240 billion of it on physical infrastructure, and power availability — not chips — is now the binding constraint. That shifts demand toward engineers who work in hardware.

Does aerospace engineering pay well? The BLS median for aerospace engineers is $134,830, with the top decile above $205,850. That trails frontier AI labs and beats nearly everything else.

Is it too late to switch into aerospace or deep tech? No. 29% of the aerospace and defense workforce is 55 or older, 76% of companies report difficulty hiring engineers, and attrition runs around 15% — more than double any other sector. The pipeline is the constraint, which favors anyone entering it.

When should I apply for aerospace internships? August through October for the following summer. Aerospace and defense recruit earlier than any other engineering discipline, and most large employers review applications on a rolling basis.

Can international students get aerospace internships in the US? Some, but the field is narrowed considerably by ITAR and clearance requirements. Focus on commercial-only work and read each posting's citizenship language carefully.

What counts as deep tech? Companies whose product depends on solving a hard science or engineering problem in the physical world — launch, satellites, defense autonomy, advanced manufacturing, energy, semiconductors, robotics — as distinct from software businesses built on existing technology.

That's exactly why we launched www.internships.space - the #1 job board for internships in aerospace.