NVIDIA CEO Jensen Huang keynote live news blog from GTC Taipei 2026

Well. That is it. Other than the extremely cringe and lame AI-gen "music" video at the end I think it was an incredible keynote! I can't wait to see what these RTX Spark laptops can achieve! 🔥

And that's a wrap! Thanks for joining us. Stay tuned to the main site for more in-depth breakdowns on the stories and for all things Computex 2026.

My stance on AI-generated music is... "no," and this reinforces that.

I think one more slide and a couple bullet points wrapping up the keynote would have sufficed 😂

Oh god what is this....

Wrapping it up.... Vera Rubin is in mass production. Vera CPUs deliver next-gen performance for agents, RTX Spark is a new Windows on Arm chip with RTX gaming capabilities and powerful local AI, and DGX Stations for Windows are bringing Grace Blackwell supercomputers to desks.

Vera Rubin in a nut shell - It's specifically designed to run agents, while Grace Blackwell was designed to power inference

Cars with reasoning, robotics. NVIDIA are building and laying the foundation for the future. That said, we’d still like to see some GeForce RTX action.

NVIDIA has unveiled "the world's first reasoning autonomous vehicle"

Interestingly, Tesla isn't listed as a partner....

I wonder how Elon feels about NVIDIA now building Robotaxis....

NVIDIA’s move into the PC market puts it in direct competition with Qualcomm, AMD, and Intel. Stuff like this doesn’t happen overnight, and now we know why it put in a bid to try to acquire Arm a few years ago. Partnering with MediaTek, TSMC, and Microsoft is huge for Windows, AI, creative, gaming, and more. And this is just the beginning.

Cosmos is the gateway to turning AI physical in the form of robotics

RTX Spark and DGX Station roadmap. NVIDIA is serious about entering the PC market with laptops and PCs for the new era of agentic AI. Blackwell this year, Rubin and Feynman on the way!

The real question is - what type of computer isn't NVIDIA working on?!?! 😂

What's that on the left? DGX Stations for Windows, desk-based supercomputers with up to 748GB of memory.

DGX Spark can run 1 trillion parameter models...

An MSI RTX Spark Mini PC. Impressive/we want one now, please.

Adobe Premiere and Photoshop are being updated and optimized specifically for RTX Spark Windows laptops and PCs, with support for powerful agents to accelerate creative workflows.

For gamers, you're looking at 1440p at 100+ FPS with DLSS 4.5 while playing the latest AAA games with ray tracing. "It runs everything the world has ever created, plus AI agents."

Jensen seen holding two RTX Spark laptops running Forza Horizon 6 and 007 First Light

"The dawn of a new personal computing revolution."

From agents to creative workloads and even gaming with DLSS 4.5, RTX Spark laptops from ASUS, Dell, HP, MSI, and more are coming later this year. There’s even going to be a custom Microsoft Surface.

What’s RTX Spark? It’s an all-in-one Windows on Arm chip that features a new 20-core NVIDIA Grace GPU developed with MediaTek, as well as an RTX Blackwell GPU with 6,144 CUDA Cores. That puts its GPU specs in the realm of a GeForce RTX 5070 Laptop GPU. But with cutting-edge chip-to-chip interconnect and up to 128GB of memory, you’ve got a portable device with a 1 petaflop of AI performance. This is the “new era of PC” that NVIDIA and Microsoft have been teasing ahead of Computex 2026.

Jensen has said 40 years of traditional PCs is now at an end. NVIDIA and Microsoft are reinventing the PC

Here we go..... something for Personal Computers!

"AI agents are going to create the largest opportunity for our partners and friends" said Huang

And it's already working on Nemotron 4... because of course.

NVIDIA has officially announced Nemotron 3 Ultra model

Chip Design Super Agents designing chips for agents, it's chip-ception! But really, it makes sense and is a testament to how far we've come as NVIDIA partners with Cadence to reinvent chip design. "From weeks to hours," Jensen says. "We have thousands" of human chip designers, and it's going to "hire" exponentially more agents.

NVIDIA's harness OpenShell is officially coming to Windows, Red Hat, Canonical, and more!

NVIDIA has been building the tools and ecosystem for agents for over a decade, and now it's combined everything from models to tools into a complete NVIDIA Agent Toolkit that's secure and open-source in OpenShell.

NVIDIA has partnered with the New York Stock Exchange

NVIDIA Vera kind of feels like the beginning of a big cloud forming over traditional x86 architecture that powers AMD and Intel CPUs. Double the bandwidth per core is the sort of stat that's a game-changer for data center and cloud environments.

Vera Rubin is the first CPU to use LPDDR5X memory - 1.2 TB/s

Jensen presented an interesting perspective on the Vera CPU. Traditionally, CPUs are built for humans, with a billion PC users out there. However, with Vera, it's purpose-built for agents, countless agents. And this is why the specs here are, well, groundbreaking - with CPU, GPU, networking, storage, and interconnect speeds built for modern AI factories with so many agents we're not sure if this means tens of billions... or a trillion?

In the age of AI agents the CPU is now the bottleneck to the GPU.

Vera Rubin is the first PCIe Gen6 with 1.4 TB/s.... Holy 🤤

1.2 TB/s of LPDDR5X ECC memory - an insane number. And yes, this is the same memory technology found in most laptops and gaming handhelds you see today.

Storage and memory are the most expensive part of these systems because of the, well, need for speed. It's as simple as that, and it's why NVIDIA has built a brand-new CPU architecture for the era of AI agents.

Huang says its vital NVIDIA makes the CPUs as low latency as possible as AI agents think in "nanoseconds". NVIDIA says its "already one of the largest CPU companies in the world"

NVIDIA's Vera CPU is built for agentic AI. "The most advanced CPUs that have ever been built," Jensen says. And it all comes down to speed, and making sure that there's the lowest possible latency for agents. NVIDIA has already sold millions.

NVIDIA is now the largest networking company in the world, thanks to Grace Blackwell and Vera Rubin. Wow.

Vera Rubin = no cables, no hoses, no fans. Now there is a PCB in the middle, and the reliability and resilience of Vera Rubin is "off the charts"

NVIDIA's Jensen Huang can be seen standing in front of Vera Rubin, the new era of agentic compute is here!

Jensen says Vera Rubin is an Agent system and that AI Agents are the latest breakthrough in computer science.

NVIDIA has reinvented computing for the age of AI...

40 Petabytes of bandwidth. Vera Rubin is something else... Now we wait until we can get some version of this at home!

NVIDIA Vera Rubin NVL72 systems feature the new NVIDIA Vera CPU, NVIDIA Groq 3 LPX, NVIDIA Vera BlueField-4 STX storage, and NVIDIA spectrum-6 SPX Ethernet racks, all integrated into a single system. More than 18,000 components on one board, over a million in a single rack. It's an incredible feat of engineering.

"Profitable AI is here". Jensen has confirmed Vera Rubin is in full production

"Profitable AI is here."

"Our tokens per watt is utterly world class" - Jensen Huang

"Compute is revenue now." "Compute is profit.".... is this the new "the more you buy..."?

"NVIDIA ultimately isn't building a GPU," Jensen says, it's now in the AI factory business. The sheer scope, complexity, and power are mind-boggling. Apparently, it has to work from the get-go. And it was built with NVIDIA Omniverse. Yes, that tool that we've been hearing about for years, a system that can create a digital twin of our planet for weather simulation, is now being utilized to build AI factories.

AI factories are the largest infrastructure build out in human history....

Vera Rubin is "the most ambitious enveador in the history of our company." Apparently, everyone within NVIDIA has worked on it to deliver the most complex system ever designed. "Our customers don't want to buy computers, they want to build AI factories."

Vera Rubin showcased at GTC Taipei by NVIDIA CEO Jensen Huang!

Vera Rubin time!

Jensen says to think of AI Agents as workers in a workshop who can reason, plan, and execute. "Today they are simple users of tools," and tomorrow they're going to be experts.

Jensen says that 40-years of traditional PC interaction is coming to an end. The new era is AI agents

From robotics to scientific research and everything else... "a round of applause for math," Jensen Huang

Agentic AI is indeed the future of generative AI, with Jensen giving a nice little shout-out to Claude... It's no longer asking a question and getting an answer; it's giving AI agents access to a full suite of tools to deliver results. Jensen says, "This is how computers are going to work in the future."

Jensen has outlined the new computing model i.e Agents + LLMs

Exponential growth for AI, and it just keeps accelerating.

NVIDIA CEO Jensen Huang has said the annual GDP of Taiwan is expected to grow by 10%!

Jensen paying homage to all of the tech companies in Taiwan... there's a lot! And we'll be seeing maybe 5% of them at the show!

T minus 60 seconds!

The live show has finished. The countdown is on now till we start.

We’re still here. 18 minutes till this thing kicks off!

Jensen is due to hit the stage in 47 minutes. Once he starts, the updates will come thick and fast from our team. Updates will be a little slow till then. Leave the tab open. Enjoy! 🙂

This is our very first live blog. Everything seems to be running smoothly so far. Turn on sound to get a ping when we post an update.

There’s some sort of talk show being played in Chinese, discussing all things NVIDIA and AI.

We are inside now seated in the press area. An hour to go before Jensen hits the stage and it’s probably already 80% full.

Cute gift for press at check in! Jensen keyring.

The line was scary but it moved quite fast. Taiwan efficiency!

We made it to the Taipei Music Center along with the 1000s of other people!