RoboBoom
VERTEX· June 3, 2026· 24:10

LIVE | NXP CEO Kurt Sievers Delivers Keynote at Computex 2026 in Taiwan | VERTEX

NXP CEO Kurt Sievers presents a Computex 2026 keynote in Taiwan on physical AI, robot world models, edge hardware and trust, with Raphael joining him to outline NXP’s approach.

In this talk

  1. 0:01

    Teaching robots to understand physics

    Sievers distinguishes fluent movement from understanding what actions will do. People build physical intuition through experience, but letting robots learn freely in the real world could be unsafe. He uses a cough syrup bottle to show why recognizing an object is not enough: robots also need to account for its contents, inertia and friction.

  2. 2:53

    Bringing world models to edge devices

    Sievers says teaching robots by demonstration can be slow and costly. World models could help them predict outcomes without first experiencing every action, while vision-language-action models connect perception, instructions and movement. NXP’s eIQ toolkit imports, quantizes, prunes and compiles models to fit edge processors with limited power, memory and processing capacity.

  3. 6:27

    Building trust into physical AI

    Sievers argues that useful machines must also be trustworthy from deployment. He describes risks from changing environments, malicious attacks, design flaws, model limits and hardware wear. Raphael outlines NXP’s approach: contain failures through isolation and redundancy, protect hardware and credentials, verify behavior through assurance programs such as SafeAssure, and adapt devices with ongoing updates.

  4. 13:12

    Finding valuable deployments today

    Raphael says the practical question is where physical AI can create value safely today. He cites factory deployments reporting 40% productivity improvements over conventional automation and NXP’s work with Boston Dynamics on robots cooperating with people on factory floors. Sievers points to a reported 610% rise in diagnostic and lab robot sales in 2025 and NXP’s engagement with GE HealthCare on anesthesia systems.

  5. 16:54

    Scaling with the right architecture

    Sievers compares elite robotics to Lionel Messi’s performance, emphasizing the engineered systems behind visible skill. He says high-performing machines need low latency, low power use and strong trust, with a neural architecture that places intelligence appropriately for real-world conditions.

Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.

More talks