Why we’re at the beginning of the AI hardware boom | Caitlin Kalinowski (ex–OpenAI, Meta, Apple)
Caitlin Kalinowski, a veteran hardware leader from Apple, Meta, and OpenAI, discusses the future of AI hardware, robotics, and physical AI. She covers the challenges of hardware development, supply chain vulnerabilities, the rise of humanoid robots, and why she left OpenAI over governance concerns around a defense deal.
Summary
Caitlin Kalinowski, who led hardware teams at Apple (MacBook Pro, Air, Mac Pro), Meta (VR/AR including Quest and Orion), and OpenAI (robotics division), joins Lenny's podcast to discuss the emerging hardware and robotics boom driven by AI.
On VR and AR: Kalinowski argues that VR's apparent failure was actually a necessary technological step that developed foundational capabilities (SLAM positioning, depth sensing, spatial computing) now being applied to robotics and physical AI. She believes AR glasses are part of the future but notes current products like Meta's Orion are ahead of their time due to immature waveguide and micro-LED manufacturing.
On robotics and humanoids: She expresses cautious optimism about humanoid robots, noting that current prototypes still require humans to stay 3 feet away for safety reasons. She argues that humanoids aren't necessarily the right solution for all problems — dedicated manufacturing robots are more appropriate for repetitive industrial tasks. She emphasizes that scale (millions of units) remains the biggest challenge, compounded by complex supply chain dependencies on China and Asia for key components like actuators and magnets.
On supply chain and reindustrialization: Kalinowski argues strongly for reindustrializing the United States to achieve supply chain independence, particularly for military readiness. She highlights magnets, actuators, memory (DRAM), and silicon as critical bottleneck components. Memory prices are expected to spike dramatically (potentially 6x) driven by AI data center demand, threatening consumer hardware and robotics companies.
On military technology: She agrees with Palmer Luckey's view that investment should shift from aircraft carriers to drones, citing Ukraine as a real-world example of how AI and drone warfare is transforming military conflict. She personally avoids working on lethal technology but supports those who do.
On hardware development principles: Drawing from her Apple training, she outlines key hardware principles: define KPIs early and stick to them; start with the hardest/riskiest components first; invest disproportionate iteration on components users touch most; move immediately on known tasks because surprises are always around the corner; and use off-the-shelf components during prototyping but custom components when final design KPIs demand it.
On AI's impact on hardware engineering: She notes AI is beginning to assist with PCB routing and component selection but cannot yet do true parametric 3D CAD. She argues that world models — not just LLMs — will be needed for AI to truly assist mechanical engineering, as current models lack understanding of physics, friction, and spatial manipulation.
On robot design and human connection: She references researcher Leila Takayama's work on social robotics, noting that robots must signal intent before moving, acknowledge human presence, appear soft and non-threatening, and exhibit approachable nonverbal cues. She cites Pixar and Disney as world leaders in expressing emotion and intent through character design.
On leadership lessons: From Sam Altman, she learned to think at 100x or 10,000x scale. From Steve Jobs, she internalized an unwavering bar for excellence that motivated rather than demoralized. From Mark Zuckerberg, she observed exceptionally clean organizational decision-making — decisions pushed to the lowest possible level, clear objectives, and technically engaged leadership.
On leaving OpenAI: She departed over concerns about the governance process and speed of decision-making around the announced Department of Defense deal, choosing a 'third path' of public disagreement rather than either silent compliance or scorched-earth criticism.
On hiring: She looks for strong generalists who can transfer knowledge across domains, a mix of zero-to-one builders and scaling experts, mission-aligned team members, and crucially, AI-native young engineers (often in their early 20s) who approach problem-solving fundamentally differently and can teach older engineers new paradigms.
Key Insights
- Kalinowski argues that VR's apparent commercial failure was actually a productive step in a longer technological arc, with SLAM positioning, depth sensing, and spatial computing now foundational to robotics and autonomous systems.
- She claims that current humanoid robots still require humans to maintain a 3-foot safety distance, meaning they are advanced prototypes rather than consumer-ready products despite media hype.
- Kalinowski asserts that dedicated task-specific robots (e.g., a robot that screws 10 screws into a laptop case 10,000 times a week) are more practical than generalist humanoids for most industrial applications.
- She argues that memory (DRAM) prices are heading toward a 6x increase driven by AI data center demand, which is an existential supply chain threat for consumer hardware and robotics companies that are more price-sensitive than hyperscalers.
- Kalinowski contends that the U.S. needs to significantly reindustrialize — particularly around magnets, actuators, and raw material processing — to maintain military safety, noting that current allies may not remain allies in the future.
- She claims that AI and current LLMs cannot perform true parametric 3D CAD because they lack physical understanding of friction, weight, pressure, and contact — and that world models, not LLMs, will likely be required to transform hardware engineering.
- Kalinowski argues that hardware development requires a fundamentally different mindset than software because teams only get 4-5 major 'compile' iterations per year, making upfront goal-setting and risk-first design sequencing critical.
- She argues that the most important leadership lesson from Steve Jobs was his unwavering quality bar, which she found highly motivating rather than demoralizing for ambitious engineers.
- Kalinowski claims that Mark Zuckerberg ran Meta's hardware organization with exceptional clarity — pushing decisions to the lowest possible level and maintaining technically engaged leadership capable of engaging with 20-page engineering trade-off documents.
- She argues that Sam Altman's most distinctive leadership quality is pushing teams to think 100x or 10,000x bigger, which she found revealed blind spots in her own ambition scale.
- Kalinowski asserts that early-20s AI-native engineers approach problem-solving fundamentally differently from older digital natives, making them essential hires who can teach established engineers new AI-first paradigms.
- She argues that there will likely be more change in military technology than in consumer electronics over the next two years, citing Ukraine's drone warfare evolution as evidence of how rapidly AI is transforming conflict.
Topics
Full transcript available for MurmurCast members
Sign Up to Access