Why we’re at the beginning of the AI hardware boom

Caitlin Kalinowski is a hardware and robotics leader with prior roles at Apple, Meta (Oculus/Orion AR glasses), and OpenAI, where she led robotics. The conversation covers why AI is pushing attention toward the physical world, the difficulty of shipping hardware, supply chain and memory constraints, robotics safety, and lessons on building hardware teams and products.

01Key learnings

• Hardware can only be 'compiled' a handful of times, so teams must be more conservative, run reliability checks early, and solve yield problems before final production.
• Define clear, written goals (cost, weight, resolution) before starting and resist changing them, since hardware is far less adaptable to mid-course changes than software.
• Design the riskiest or hardest-to-physically-build parts first rather than starting with the components you already know how to design.
• Iterate most heavily on the parts customers touch most, since those interactions shape how the product feels and determine reliability.
• Pre-buy scarce components like memory when you can afford it; a single missing part can force a catastrophic redesign and months of delay.
• Use off-the-shelf parts for prototyping to prove a concept works, then move to custom components once the design is fixed and mass production is planned.
• When hiring for zero-to-one hardware, prioritize hybrid generalists, people who have scaled similar things, and AI-native younger engineers, aligned on mission and eager to learn.
• Dedicated, purpose-built robots will likely outnumber humanoids in real industrial use; humanoids remain advanced prototypes with significant safety and scaling challenges.

02Frameworks mentioned

Summary and takeaways written by PM Atlas; quotes are short excerpts. © the original author.