The real value of AI lies in optimizing physical processes, not generating content.
En Computex 2026, el vicepresidente ejecutivo de Qualcomm Nakul Duggal planteó una distinción que podría reorientar la industria tecnológica: la inteligencia artificial más transformadora no es la que crea contenido, sino la que optimiza el mundo físico desde dentro de las propias máquinas. Su argumento sitúa el verdadero valor de la IA no en los servidores remotos, sino en el borde de la red, donde los sistemas industriales toman decisiones en tiempo real sin depender de la nube. Es un giro que desplaza el centro de gravedad del debate tecnológico desde la generación hacia la operación, desde lo virtual hacia lo tangible.
- La industria tecnológica lleva años midiendo el progreso de la IA por la potencia bruta de sus modelos generativos, pero ese relato empieza a mostrar sus límites frente a las necesidades reales del sector industrial.
- Duggal lanzó en Computex 2026 una tesis directa: los sistemas que detectan defectos en una línea de fabricación o guían robots sin supervisión humana generan más valor económico que los que redactan textos o producen imágenes.
- La latencia casi nula, la privacidad de los datos y la autonomía operativa son las ventajas concretas que el procesamiento en el borde ofrece frente a la dependencia de centros de datos centralizados.
- Escalar esta visión exige ecosistemas abiertos y una colaboración estrecha entre fabricantes de chips, desarrolladores de software e integradores de sistemas, lo que convierte la interoperabilidad en el verdadero cuello de botella.
- El mercado parece responder: en Computex 2026, la conversación giró hacia infraestructuras completas de IA que conectan el mundo digital con el físico en manufactura, logística y robótica.
Nakul Duggal, vicepresidente ejecutivo de Qualcomm responsable de las divisiones de automoción, IoT industrial y robótica, subió al escenario de Computex 2026 con un argumento que desafía el relato dominante sobre la inteligencia artificial. La IA que realmente importa, sostuvo, no es la que genera ensayos o imágenes, sino la que hace funcionar mejor a las fábricas, mover con más inteligencia a los robots y decidir más rápido a los sistemas industriales, sin depender de servidores lejanos.
Su ponencia, centrada en lo que la industria denomina IA física, defiende que el próximo gran ciclo de expansión de la IA no vendrá de escalar la computación en la nube, sino de llevar la inteligencia al borde de la red: directamente a los sensores del suelo de fábrica, a los brazos robóticos, a los dispositivos conectados. Las ventajas son precisas: latencia casi nula, datos que permanecen locales mejorando la privacidad, y sistemas capaces de reaccionar en tiempo real sin esperar una respuesta del centro de datos. Una línea de producción que detecta un defecto y se detiene sola. Un robot que sortea un obstáculo sin intervención humana.
Duggal subrayó que esta transición exige algo más que mejores chips. Requiere plataformas abiertas que permitan a los desarrolladores escribir una vez y desplegar en múltiples dispositivos, y una colaboración real entre fabricantes de silicio, creadores de modelos y las empresas que operan fábricas y redes logísticas. El contexto de Computex 2026 reflejó ese cambio de orientación: la potencia de cómputo dejó de ser el titular, y el debate se desplazó hacia los sistemas completos de IA que conectan el mundo digital con el físico.
Para Qualcomm, la apuesta estratégica es evidente en sectores como la robótica, el IoT industrial y los sistemas embebidos, donde la IA en el borde no es un complemento sino una necesidad. La distinción que Duggal trazó —entre IA que genera y IA que optimiza— puede parecer sutil, pero señala un desplazamiento profundo en la forma en que la industria concibe dónde se crea valor y dónde llegará el próximo ciclo de crecimiento.
Nakul Duggal stood before an audience of hardware makers, software engineers, and supply chain executives at Computex 2026 with a simple but pointed argument: the artificial intelligence that matters most is not the kind that writes essays or generates images, but the kind that makes factories run better, robots move smarter, and industrial systems decide faster without waiting for instructions from distant servers.
Duggal, Qualcomm's executive vice president overseeing automotive, industrial, embedded IoT, and robotics divisions, framed his vision around what the industry calls physical AI—machine learning deployed directly on the devices and machines that generate data, rather than shipped off to centralized data centers. His talk, titled "Growth with Industrial and Physical AI at the Network Edge," laid out a thesis that the next wave of AI expansion will not come from scaling up cloud computing, but from pushing intelligence outward, closer to where work actually happens.
The practical benefits he outlined were concrete. When inference and learning happen at the edge—on a factory floor sensor, inside a robot arm, embedded in a connected device—latency drops to near zero. Data stays local, which means privacy improves and companies keep sensitive operational information off the internet. Industrial systems can make real-time decisions without waiting for a round trip to the cloud. A manufacturing line can detect a defect and stop itself. A robot can navigate an obstacle without human oversight. These are not flashy applications, but they are the ones that move products, save money, and reshape how industries operate.
Duggal emphasized that this shift requires more than just better chips. It demands open ecosystems—platforms that can scale from low-power edge devices all the way up to complex industrial systems, allowing developers to write once and deploy across a range of hardware. It demands collaboration across the entire technology supply chain: chipmakers building the silicon, software developers writing the models, systems integrators connecting everything together, and the companies actually running factories and logistics networks putting it all to work.
The broader context at Computex 2026 reflected this reorientation. The conversation has moved away from raw computing power as the headline metric. Instead, analysts and industry leaders were discussing complete AI systems—the infrastructure, platforms, and applications that bridge the digital and physical worlds. Manufacturing, logistics, mobility, robotics: these sectors are being reshaped by AI that operates not in the cloud but in the machines themselves.
Qualcomm's strategic interest is clear. The company sees sustained growth ahead in robotics, industrial IoT, and embedded systems—areas where edge AI is not a nice-to-have but essential. The real value of artificial intelligence, Duggal argued, lies not in what it can generate or simulate, but in what it can optimize and improve in the physical world where actual work gets done. That distinction may seem subtle, but it marks a fundamental shift in how the industry thinks about where AI creates value and where the next generation of growth will come from.
Citações Notáveis
The true value of AI does not reside solely in content generation or conversational assistants, but in its capacity to optimize physical processes and improve operational efficiency in real-world environments.— Nakul Duggal, Qualcomm
A Conversa do Hearth Outra perspectiva sobre a história
When Duggal talks about physical AI at the edge, is he really just saying that processing should happen locally instead of in the cloud?
That's part of it, but it's more about a complete rethinking of where intelligence lives. It's not just moving computation—it's about systems that can think and act in real time without waiting for permission from a distant server.
Why does that matter so much for a factory or a robot?
Because the world doesn't wait. A defect on a production line, an obstacle in a robot's path, a logistics decision that needs to happen in seconds—these can't afford the round-trip delay of sending data to the cloud and waiting for an answer. Edge AI means the machine itself becomes intelligent.
But doesn't that require the chip itself to be smarter?
Yes, but Duggal's point is that it's not just about the chip. You need the whole stack—the hardware, the software frameworks, the platforms that let developers build once and deploy everywhere. That's the ecosystem piece.
So Qualcomm is positioning itself as the backbone of that ecosystem?
Exactly. They're saying: we'll provide the silicon and the platform infrastructure that lets manufacturers, roboticists, and logistics companies build AI systems that actually work in the real world, not just in demos.
And the privacy angle—that's real, or is that just marketing?
It's real. If your factory's operational data stays on your machines instead of flowing to the cloud, you control it. That matters enormously for competitive advantage and regulatory compliance.