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

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

Jensen Huang hasn’t even gone on stage yet, but the amount of leaked information is enough to fill a book.

Wang Wan has summarized three main highlights, so let’s go, friends, follow me.

1. AI computing power costs are cut in half

The previous generation Blackwell was already impressive, right? Soon, the new generation chip Vera Rubin will be mass-produced.

What makes Vera Rubin so powerful? Simply put, two words: cheap.

Running the same AI model,
chip count is reduced to a quarter, inference computation costs drop by 90%.
A 90% reduction, friends.
AWS, Microsoft, and Google’s top cloud providers are all jumping on board first.

2. The Groq acquisition for $20 billion last year delivers results today

Previously, Jensen Huang said at the earnings call that Groq would be integrated as an extension architecture into the NVIDIA ecosystem, just like Mellanox was acquired to enhance networking capabilities.

Groq’s LPU and NVIDIA’s GPU are housed in the same data center; the GPU handles understanding problems, while the LPU is responsible for quickly spitting out answers.

The division of labor between the two chips, combined with agent scenarios, directly reduces latency.

AI agents do work for people; a task might require dozens of model adjustments back and forth, each round burning inference power, and users are waiting. If it’s slower, the experience crashes.

Inference involves two steps: first understanding your question, then outputting the answer word by word.

GPUs excel at the first step, but for the second step—spewing out words quickly and stably—Groq’s LPU is better.

Is 20 billion expensive?

Think about it—every company in the future will run hundreds of agents, each adjusting models thousands of times a day.

3. NVIDIA’s version of OpenClaw launches, called NemoClaw

It’s an open-source platform that enterprises can deploy to have AI employees handle workflows, process data, and manage projects.
It’s said to be in talks with Salesforce and Adobe.

What’s interesting is that NemoClaw doesn’t require you to use NVIDIA chips.
Think about this logic.
Selling chips only earns money from hardware; setting the rules allows you to earn from the entire chain. Jensen Huang has a clear grasp of this.

4. Jensen Huang says he will showcase a “chip never seen before”

Most likely, the next-generation architecture Feynman will make its debut, with mass production in 2028, using TSMC’s most advanced 1.6nm process.

There’s also an obscure rumor I find quite interesting.

NVIDIA has released laptop processors, two models, aimed at gaming.
The graphics card sellers are coming to compete for CPU market share.

Wang Wan feels that Jensen Huang might become a great figure in the future.