Tech • IA • Crypto
Jensen Huang and Satya Nadella in Conversation at Microsoft Build
9/10TL;DR
Giants NVIDIA and Microsoft are accelerating the transformation of PCs and the cloud with AI systems capable of acting autonomously, from personal desktops to data centers.
KEY POINTS
The PC becomes a “personal AI”
The personal computer is evolving into an autonomous assistant capable of executing tasks without direct interaction. Thanks to new architectures, users can delegate coding, design, or remote modifications, with the PC launching tools and iterating on its own. This shift marks the transition from a simple tool to an active AI-driven agent.
RTX Spark and large-scale local AI
The RTX Spark system introduces power reaching one petaflop for AI computing, with 128 GB of memory enabling models with hundreds of billions of parameters to run. This brings cloud-level performance directly to local machines, making advanced AI accessible without remote infrastructure.
Intelligent agents at the core of usage
New so-called “agentic” models no longer just respond—they act. They automate complex tasks, interact with tools, and produce concrete results. This evolution explains the recent surge in usage, with strong growth in development and content generation activities.
An explosion in compute demand
Intensive use of agents is driving a massive increase in power needs. The compute required to train, fine-tune, and run these systems has risen sharply, while token production is becoming economically viable, further boosting demand.
Data centers redesigned for AI
Infrastructure is evolving toward integrated systems like Grace Blackwell and soon Vera Rubin, capable of handling massive workloads. With architectures like NVLink 72, an entire rack operates as a single computer. These systems enable major gains, reducing generation costs by up to 30x compared to the previous generation.
Energy efficiency and innovative cooling
New data centers adopt advanced technologies, including closed-loop liquid cooling that uses very little water. The goal is to combine extreme performance with reduced environmental impact while increasing compute density.
Stronger data security
The architecture includes full encryption of data, both in transit and in use. This “confidential computing” approach is becoming essential as agents handle sensitive information at scale.
Convergence between cloud and PC
The same models and architectures are now deployed from the cloud to individual machines. Environments like Windows and Azure directly integrate these technologies, creating continuity between personal use and large-scale infrastructure.
A fully accelerated software ecosystem
Development and data-processing tools such as SQL, Spark, and vector systems are optimized to run on GPUs. This acceleration is critical for agents, which require fast responses to iterate efficiently.
CONCLUSION
Agentic AI is reshaping both personal computing and cloud infrastructure, opening a new era where machines and software act autonomously at scale.