CES 2026: Jensen Huang and the "Physical AI" Revolution - AI Compute Leasing Unlocks a Trillion-Dollar New Market

The 2026 International Consumer Electronics Show (CES), themed "Defining the Physical Boundaries of AI," concluded in Las Vegas under the global tech spotlight. This edition transcended its role as a consumer electronics trendsetter, becoming a critical inflection point for artificial intelligence's leap from the virtual world into physical reality. The keynote speech by NVIDIA's founder and CEO, Jensen Huang, was undoubtedly the "crown jewel" of this gathering. He systematically outlined the vision for "Physical AI" and revealed that a new industrial infrastructure, centered on AI computing power and compute leasing, is rapidly taking shape.

01 The Show Landscape: A Panoramic View of the Physical AI Ecosystem, from Chips to Robots

CES 2026 moved decisively beyond mere conceptual showcases, presenting a complete, pragmatic, and highly integrated Physical AI ecosystem. The show's core focused on the closed loop of "Embodied Intelligence - Edge Computing - Cloud Brain."

• Scale & Focus: Over 4,000 exhibitors demonstrated full-stack innovation, from foundational AI computing power chips and AI compute leasing services, to mid-layer smart vehicles and AI PCs, and up to cutting-edge biomimetic robots. Among them, the contingent of Chinese tech companies stood out particularly in application-layer fields like intelligent cockpits and robotics.

• Paradigm Shift: Differing from the past pursuit of intelligence in single devices, the consensus at this year's show was that the capability boundary of any physically intelligent device is no longer limited solely by its own processor but depends more on the scale and speed of the cloud-based AI computing power it can access. This provides the logical foundation for explosive growth in the compute leasing market.

02 Jensen Huang's "Physical AI" Revolution: Redefining Computing Architecture

Huang's speech, titled "The Era of Physical AI," set the technical tone for the entire event. He posited that the next stage of Artificial General Intelligence (AGI) is to create agents that can understand and safely intervene in the physical world.

• Core Philosophy: "Physical AI" must master the "common sense" and causal logic of the physical world, not just process data patterns. This demands unprecedented new types of AI computing power capable of both intensive centralized training and supporting massively concurrent real-time inference and simulation.

• Strategic Announcements: He announced that NVIDIA's focus would shift from "providing computing power for data centers" to "providing intelligence for the physical world." This means its chips, software, and ecosystem will comprehensively serve robotics, autonomous driving, and industrial automation, driving demand for specialized, large-scale AI computing power in these fields.

03 Open Ecosystem & Next-Gen Hardware Platform: Rubin Reshapes Compute Economics

To support the "Physical AI" vision, Huang unveiled a game-changing new generation of technology platforms and ecosystem strategies.

• Vera Rubin Platform: The early unveiling of the Rubin platform, featuring a six-chip module design (CPU+GPU+3rd-gen NVLink), was a key highlight. Its breakthrough lies in dramatically improving the comprehensive energy efficiency ratio for running large Physical AI models through architectural innovation, directly reducing the total compute cost from R&D to deployment.

• Open Source & Open Access: NVIDIA announced the open-sourcing of a series of foundational models for robotics and AI agents, along with opening access to its digital twin simulation platform. This move aims to lower the barrier to entry, attracting millions of developers into the Physical AI field. This vast pool of developers and startups constitutes the primary target customer base for cloud-based AI compute leasing services, enabling them to access world-class AI computing power for innovation via leasing, without the need to build expensive private compute centers.

04 Compute Power Strategies of Major Players: Competition Intensifies, Leasing Becomes a Mainstream Option

In response to NVIDIA's strong position, other semiconductor giants showcased their differentiated AI computing power strategies at CES 2026, further validating the diversity of market demand.

• AMD: Introduced a full-stack AI acceleration portfolio from the cloud (Instinct MI455) to the edge (Ryzen AI 400), emphasizing its balance of high-performance computing and energy efficiency, providing the market with another reliable source for AI computing power.

• Intel: Focused on local inference capabilities for AI PCs. Its Panther Lake processor aims to offload more intelligent workloads from the cloud to the device. However, this does not diminish cloud computing demand; instead, it promotes a hybrid "cloud-edge collaborative" architecture, making enterprises' demand for elastic compute leasing more nuanced and complex.

Industry analysis indicates that the fierce competition among giants is driving down the unit cost of AI computing power while simultaneously increasing the technical and financial risks of building private compute centers. Consequently, adopting third-party compute leasing services is evolving from an option to a mainstream strategic choice for both large enterprises and SMEs.

05 Mass Adoption of AI Endpoint Devices: The Ultimate "Compute Guzzlers"

CES 2026 featured the first large-scale demonstration of practical AI endpoint devices, which represent the ultimate "compute guzzlers" for consuming AI computing power.

• Ubiquitous AI PCs: Comprehensive product line launches from Lenovo, Dell, HP, and others all featured built-in dedicated AI engines capable of running personal agents locally. These agents require continuous collaboration with cloud-based large models, generating persistent inference computing power calls.

• Dawn of the Robotics Revolution: Robots, from home companions to industrial handlers, showed remarkable progress towards practicality. The skill training, environmental simulation, and continuous learning for each robot consume massive amounts of cloud-based AI computing power. A significant portion of a robotics company's operational costs will transform into bills paid to compute leasing providers.

06 Future Trends & Compute Infrastructure Evolution: Hybrid Leasing and Green Computing

Based on the presentations at CES 2026, the trends and infrastructure evolution directions for the AI industry over the next two years are clear.

• Trend One: Hybrid Compute Leasing Models Become Enterprise Standard. Enterprises will dynamically distribute workloads based on data sensitivity, task latency requirements, and cost considerations. Core data training might use private AI computing power, while large-scale simulation testing and short-term peak demands will inevitably be met through public cloud compute leasing. A global, liquid market for trading AI computing power will emerge.

• Trend Two: Compute Infrastructure Competition Shifts from Pure Performance to "Efficiency" and "Sustainability". As the demand for AI computing power grows exponentially, energy consumption becomes an unavoidable bottleneck and a core cost factor. The future competitiveness of compute leasing providers will depend not only on the amount of PetaFLOPS offered but increasingly on delivering "optimal performance per watt" of green computing power. Liquid cooling, renewable energy power, and similar features will become essential elements of computing infrastructure.

Conclusion

CES 2026 marks the official entry of artificial intelligence into a new era of "embodiment." The vision of "Physical AI" championed by Jensen Huang is driving a transformation more fundamental than the internet revolution—the reshaping of global computing infrastructure.

In this process, AI computing power becomes the core production resource, akin to electricity in the industrial age. Compute leasing services act as the critical hub for the efficient and democratic allocation of this resource, similar to a national power grid. In the future, the AI competitiveness of a nation or enterprise will depend not only on its algorithm scientists but increasingly on its capability and efficiency in acquiring, orchestrating, and utilizing large-scale AI computing power. The grand opening act for a new trillion-dollar market, with AI compute leasing at its core, has already begun.