KnowledgeSketches|← All EpisodesLenny’s PodcastResearch papers
From Lenny's Podcast
Knowledge Sketch

Why We're at the Beginning of
the AI Hardware Boom

Caitlin Kalinowski
Hardware engineer · ex OpenAI, Meta, Apple
MAY 17 2026
The Shift

The Next Frontier Is
the Physical World

VRARPHYSICAL AIONE TECH LINEAGE
"What you can do behind a keyboard with AI is going to saturate. When that happens, the next frontier is the physical world."
  • She sees a "dawning realization" in the AI labs: what you can do behind a keyboard with AI will saturate. Nobody knows when
  • When it does, the next frontier is robotics, manufacturing, industrialization
  • VR work figured out SLAM and depth sensing. Robots now use the same technology to move through space
  • Labs, big tech, and startups are all realizing this at the same time
Why It's Hard

Hardware Compiles Four or Five Times

SOFTWARECOMPILE DAILYHARDWARE4-5 TIMES, TOTALNO UPDATES AFTER RELEASE
4-5
compiles, total
0.5%
the last half percent to solve
3-5mo
per build, give or take
  • Software people can compile every day, every hour. Hardware gets four or five builds, total
  • Once you release for mass production, you are done. You can't ship updates to fix it
  • Part variance is high: the smallest of one part meets the largest of the next, across millions
  • So you stay conservative, run reliability checks in line, and solve the last half percent
Caitlin on the core constraint"In hardware, we only get to 'compile our code' four or five times." Lenny asked if that was per year. Her answer: "Total."
Playbook

Four Rules for Building Hardware

  • Be clear on goals: write them down as KPIs and change them as little as possible
  • Design the hardest part first: start at the pinch points where it could fail, not the part you already know
  • Iterate on what people touch: on a computer the trackpad, then maybe the keyboard, need way more iteration than the rest
  • Do it now: in hardware you do not actually have more time
Goals as KPIs

Set out to make something at $300, then say halfway it has to be $150, and you have almost burned a lot of that early time. Know your goals upfront, hit them, and you know you can ship.

The back of the cabinet

At Apple every design decision, even on the inside of the device, is considered. She says it forces the teams to ask what really matters for each part.

Supply Chain

One Missing Part Stops Everything

  • Each layer, from raw magnets to actuators to robots, has mostly been outsourced over the last 25 years to countries like China, Japan and Korea
  • Actuators are foundational. If you can't buy them, you don't get to make robots, and the US does not have great actuator companies yet
  • She wants the US to reindustrialize and "reteach ourselves how to make things at scale"
  • Her advice to startups: pre-buy memory if you can afford it, to ride out price spikes. She thinks prices will probably double, and does not know how soon
Caitlin on fragility"You can't build anything if you have one component missing." If you can't get your chip or your RAM, you redesign the entire guts of the product. She calls that a catastrophic redesign.
Contrarian

AI Hardware Myths

✗A humanoid robot can do everythingINSTEAD →✓ She is not sure one generalist shape works. Different jobs need different robots. The most advanced factory lines already run with few people, so they need more dedicated robots, not humanoids. She suspects humanoids will do some long-tail jobs.
✗Ask customers what they wantINSTEAD →✓ For something new, customers don't know what they want because they haven't seen it. Ask a hundred people about a touchscreen phone and they will say a keyboard on the screen.
✗AI can already design real hardwareINSTEAD →✓ Not yet. She says AI is not quite there on real CAD, and LLMs and video models don't understand friction, weight, or contact. She thinks it may take new model types.
✗Build every part customINSTEAD →✓ Use off the shelf whenever you can, especially in prototyping, where the goal is to show it can work at all. The final product is often highly custom, to hit a size or weight.
Based on Caitlin Kalinowski's episode on Lenny's Podcast. All ideas on this page are from the episode.Read on Lenny's Newsletter
Every sketch here began as a transcript.

Send us yours and we’ll draw it. $1 a sketch, or $49 for all 11 formats.

Sketch mine →
𝕏︎ X / Twitterin LinkedIn📸 Instagram🔗 Copy link
0:00