Today’s most important event is NVIDIA GTC Conference, which is basically an AI version of A Brief History of Humankind.

SmartContractAuditor · 2026-03-27T07:36:18+00:00

At the NVIDIA GTC conference, Jensen Huang introduced the new generation Vera Rubin chip, significantly reducing AI computing costs. Groq integrated improvements to enhance AI Agent efficiency, and the NemoClaw open-source platform was also released, allowing enterprises to deploy AI employees. Soon, a new architecture called Feynman is expected to be launched. NVIDIA is entering the laptop CPU market, demonstrating its significant progress in AI and chip technology.

SmartContractAuditor

2026-03-27 07:36:18

Abstract generation in progress

Today’s most important event is the NVIDIA GTC conference, it’s practically an AI version of a brief history of humankind.

Jensen Huang hasn’t even taken the stage yet, and the amount of leaked information is enough to write a book.

Wang Wang has organized three major highlights, come on, friends, follow me.

The cost of AI computing power has been directly slashed by 90%.

The previous generation Blackwell was already impressive, right? The new generation chip Vera Rubin is about to be announced for mass production.

What’s so great about Vera Rubin? Simply put, it’s two words: cheap.

For the same AI model, the number of chips is reduced to a quarter, and the inference computing cost drops by 90%. Ninety percent, folks. The three major cloud providers, AWS, Microsoft, and Google, are the first to get on board.

Groq, which was acquired for $20 billion last year, is submitting its homework today.

Previously, Jensen Huang mentioned in the earnings call that Groq would be integrated into the NVIDIA ecosystem as an expanded architecture, just like when they acquired Mellanox to enhance network capabilities.

Groq’s LPU is in the same data center as NVIDIA’s GPU; the GPU understands the problem, while the LPU is responsible for quickly spitting out answers.

The division of labor between the two types of chips directly reduces latency in Agent scenarios.

AI Agents do the work for people; a single task might cycle through models dozens of times, with each round burning inference computing power, and users are waiting, so a slight delay can ruin the experience.

Inference is done in two steps: first, understanding your question, and then spitting out the answer word by word.

GPUs excel at the first step, but for the speed and stability of the second step, Groq’s LPU is stronger.

Is $20 billion expensive?

Just think about it, every company will run hundreds of Agents, with each Agent calling models thousands of times a day.

NVIDIA’s version of OpenClaw is launching, called NemoClaw.

It’s an open-source platform that companies can install to deploy AI employees to run processes, handle data, and manage projects. It is said that they are already in talks with Salesforce and Adobe.

The interesting part is that NemoClaw doesn’t require you to use NVIDIA’s chips. Just consider this logic. Selling chips only earns you money on the hardware layer; setting the rules allows you to make money on the entire chain. Jensen Huang has this figured out clearly.

Jensen Huang said he will showcase “a chip the world has never seen before.”

It’s highly likely that the next-next-generation architecture Feynman will make its first appearance, set for mass production in 2028, using TSMC’s most advanced 1.6nm process.

Additionally, there’s a lesser-known piece of info that I find quite interesting.

NVIDIA has released laptop processors, two models, mainly targeting gaming. Those selling graphics cards are coming to snatch the CPU market too.

Wang Wang, I feel like Jensen Huang is destined to become a great figure in the future.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.

1 Likes