Zettascale (YC S24) Is Hiring Chip Engineers. Does Y Combinator Actually Fund Hardware Startups in 2026?

This week, Zettascale — a Summer 2024 YC company building energy-efficient AI chips — hit the Hacker News front page hiring founding FPGA and ASIC engineers. The company (formerly Exa Laboratories) claims its reconfigurable "XPU" architecture is up to 27.6x more efficient than an H100, and it's a four-person team in San Francisco backed by YC partner Tom Blomfield.

For founders sitting on a hardware or deep-tech idea, a company like this raises the obvious question: does YC actually fund anything that isn't a web app? If you're considering F26 with a chip, a robot, a sensor, or a biotech product, here's the real answer.

Does Y Combinator fund hardware startups?

Yes, and it has for over a decade. YC funded Cruise (autonomous vehicles, acquired by GM), Momentus (space propulsion), Solugen (chemical manufacturing), Helion-adjacent energy plays, and a long line of robotics and devices companies. Zettascale is simply the current proof point: a semiconductor startup, in YC, hiring hard-tech engineers in 2026.

But here's the nuance most applicants miss. YC doesn't fund hardware the way a deep-tech fund does. It isn't underwriting a ten-year fab build. It's betting that a small, technical team can show enough progress in three months to raise a strong seed round on Demo Day. The bar isn't "is this important?" It's "can these specific founders move unusually fast on something most people think is too hard?"

That reframes everything about how you should apply.

What does YC actually look for in a deep-tech application?

The pattern across hardware companies that get in looks like this:

Founders who can compress timelines

Zettascale's founders are Elias Almqvist, a self-taught engineer and Chalmers dropout, and Prithvi Raj, a Cambridge MEng who worked in scientific machine learning. That combination — someone who builds obsessively plus someone with deep domain training — is exactly what YC wants to see on a hard problem. The application has to make the partners believe you can do in months what a corporate R&D team would take years to attempt.

A wedge that fits in a 12-week batch

You can't tape out a chip in a batch. So winning hardware applications scope down to something demonstrable: a working prototype on an FPGA, a simulation with credible numbers, a signed pilot, a critical sub-component that proves the hard part works. Zettascale leaned on simulation-backed efficiency claims rather than a finished product. Your application should name the one thing you'll have working by Demo Day, not the full vision.

Numbers that make an investor lean in

"27.6x more efficient than an H100" is doing a lot of work. It's specific, it's falsifiable, and it maps directly to a buyer's pain (data-center energy cost). Deep-tech applications die when they're vague about the magnitude of the improvement. If you can't put a number on why your approach is 10x better, the partners will assume it's incremental.

How is applying with hardware different from applying with software?

Three differences matter most.

First, capital efficiency is a question you must pre-answer. YC and seed investors know hardware burns cash. A strong application shows you've thought about how to get to a meaningful milestone on a seed-sized check — contract manufacturing, off-the-shelf components, FPGAs before custom silicon, design partners who pre-pay.

Second, "why now" carries more weight. Hardware ideas often failed before because the enabling technology wasn't ready. The AI compute crunch is the "why now" for a company like Zettascale: GPU power draw keeps climbing while compute-per-area has stalled. Your application needs an equally sharp reason the window just opened.

Third, regulatory and manufacturing risk get scrutinized. If your product touches health, climate, aerospace, or chips, partners will probe how you de-risk the parts you can't control. Acknowledging this head-on reads as maturity, not weakness.

Is 2026 a good time to apply to YC with deep tech?

The ecosystem signals point yes. The same AI wave that's flooding YC with software agents is creating real demand underneath the stack: chips, energy, cooling, robotics, and manufacturing. A hardware company that's adjacent to AI infrastructure — like Zettascale — gets to ride AI tailwinds while standing apart from the crowd of nearly identical AI-wrapper applications.

That differentiation is worth a lot. When a large share of a batch's applications are AI software, a credible hardware team is memorable for the right reasons. The risk is the opposite one: applying with a deep-tech idea but a thin technical story that doesn't convince partners you can actually build it.

How do you pressure-test a deep-tech application before you submit?

The failure mode for technical founders is assuming the technology speaks for itself. It doesn't — the application has to translate hard engineering into a crisp founder-investor story: who you are, why you'll move fast, what works by Demo Day, and how big the improvement is.

That translation is hard to judge from the inside, which is the whole idea behind YC Roaster — getting your application read by founders who've actually been through the YC process, including the ones who got in with non-obvious, technical companies. A hardware founder who can explain a 27.6x claim in two sentences a generalist partner understands is far ahead of one burying it in jargon.

The bottom line

Zettascale is a live reminder that YC funds far more than SaaS. If you're building hardware or deep tech, don't talk yourself out of applying to F26 — but don't apply like a software team either. Show founders who compress timelines, a wedge that fits a 12-week batch, and a number big enough to make an investor lean in. Get that story right, and "too hard for YC" becomes exactly why they want you.