Live Updates From Computex 2026 🔴

During Qualcomm’s future-looking keynote, CEO Cristiano Amon offered even more embellishments about the inevitable arrival of next-gen connectivity, 6G. Qualcomm had been very consistent that updated data speeds will be necessary for devices that rely on cloud-based AI. When referring to smart glasses, Amon said “6G will make us walking cameras.” That’s an interesting thought, considering the significant privacy implications inherent in wearable devices like smart glasses.

6G devices are supposed to connect everything, so high speed connectivity of video sure does sound like the point is to turn the world into an inevitable panopticon. “The whole network is an AI network… it’s part of a very large data center,” Amon added. That doesn’t exactly make me very excited, to be honest. —Kyle Barr

Microsoft is now hooking up with Nvidia on its next Surface laptop. The Surface Laptop Ultra comes packed with the new N1X chip packed with the Blackwell GPU and the 20 CPU cores. Microsoft touts the promise of up to 128GB of unified memory and the efficient ARM-based architecture.

It’s meant to be a premium device with a mini LED display that claims up to 2,000 nits of peak HDR brightness. It will also feature a haptic trackpad and the plenty of ports. When’s it coming? Later this year, according to Microsoft. How much will it cost? Microsoft isn’t telling. —Kyle Barr

In its opening keynote, Computex tried to push the main theme of this year’s trade show, “AI Together.” Of all the AI generates images and videos permeating these recent keynotes, this odd rendition of Michelangelo’s Sistine Chapel ceiling feels the most telling. —Kyle Barr

Nvidia’s keynote is over, and already there’s a massive line forming to buy company merch. It’s unfortunate that the world’s most valuable company still feels the need to charge so much for branded clothing. A T-shirt is worth the equivalent of $30. That sweater is roughly $160. —Kyle Barr

The RTX Spark is Nvidia’s first new line of CPUs in a long, long time. And yes, they were built for running AI on-device. Nvidia CEO Jensen Huang made a ton of big claims about the ARM-based PCs these chips will power, such as that they’ll manage “100% of Nvidia’s software stack” and “every application that Windows has ever run.” That’s a stretch, especially for an ARM-based PC. Companies including Lenovo, Acer, Microsoft, Asus, MSI, and others are supposedly developing new RTX Spark systems. They may not completely reinvent PCs, as Huang was all too ready to claim, but it is still exciting to see more all-new PC chips on the scene. —Kyle Barr

Nvidia used to be a GPU company. After that, Huang called it a “technology company.” Now, it’s an “AI infrastructure company.” And what does that actually mean? Fulfilling orders for the massive buildout of AI data centers, or as Nvisia calls them, “AI factories.” It’s hard to feel enthused about all the billions of dollars being spent to create new data centers, as it seems the very essence of PC, the “personal computer,” is being erased for the sake of computation performed in the cloud. Oh, and these data centers require so much power and water that they’ll inevitably strain the grid. Huang himself said they need to “make the grid stronger.” Will that come before or after all these AI factories go live?’

After all this talk of AI buildout all over the world, Jensen ended his speech with a classic, “It’s OK to clap.”

And guess what? The company’s Vera Rubin AI training chips “are in full production.” Nvidia needs more factories if it wants to keep the gravy train rolling. —Kyle Barr

Listen, if you have new AI chip-designing software and you imagine you’ll hire even more chip designers, then good on you. But maybe you shouldn’t also acknowledge companies like Palantir, the surveillance and data company that works with governments and corporations and is led by Alex Karp, who tends to go full supervillain mode in public. —Kyle Barr

Nvidia CEO Jensen Huang seems adamant that what we were missing wasn’t a more powerful CPU, but rather a CPU that wasn’t built for us flesh bags. The Vera CPU, Nvidia’s latest processor, is scaled for AI training and explicitly built for AI agents. The CPU was designed to reduce latency and enable cores to communicate with each other using maximum bandwidth. The chip also uses LPDDR5X memory with a super-fast 1.2 TB/s bandwidth. He compared that number to similar AMD-made x86 CPUs and claimed it was nearly two times as fast at handling agentic tasks. —Kyle Barr

Huang laid out some… interesting math. He claimed that if the salaries of millions of software engineers are worth $3 trillion, that’s easily overcome by “useful AI,” such as algorithms coding inside GitHub, which he proclaimed is generating $9 trillion in value for companies. Who knows where the Nvidia CEO is getting his figures? He went on to call the idea of AI replacing jobs “complete nonsense,” proclaiming that companies will still want to hire more software engineers to increase productivity. That may make sense in a bubble where companies can experience infinite growth, but as evidenced by recent polls of CEOs, execs indeed imagine AI will replace jobs within a year. —Kyle Barr

Here comes Jensen, black jacket and all (his closet must be very hard to parse at this point in his life). The theme for this year’s GTC Taiwan talk? Come on, you already know. It’s that “tokens” are a new currency we’re going to use for everything from VTOL planes to robotics. “Today, we’re going to talk about the whole ecosystem,” Huang said. —Kyle Barr

As has become consistent for each one of Nvidia CEO Jensen Huang’s big conference appearances, Nvidia’s GTC showcase here at Computex 2026 is already packed to the gills. That makes more sense here in Taiwan, where Huang is considered something of a hometown hero (the CEO was born in Taipei). It also helps that Nvidia is now the most profitable company in the world.

If anybody wants to cast bets for how long Nvidia takes to talk about its new chips, I’m starting the bid at 50 seashells we won’t hear a peep until two hours in. —Kyle Barr

Badge has been received. I’ve been in Taipei, Taiwan, for a little more than 24 hours and the jet lag is hitting hard. I had a whole sordid affair surrounding the necessity of business cards for this event that I’ll refrain from ranting about here. I’m now at Nvidia, ready for its big reveals (hopefully). —Kyle Barr

You already can’t afford DDR5 RAM. Unfortunately for you, the news I’m about to share may make you even hungrier for the latest memory sticks. Multiple companies known for PC components, including Corsair and G.Skill, have unveiled new CUDIMM DDR5 (which stands for “clocked unbuffered dual in-line memory”) modules that we expect to reach higher frequencies than other DDR5 modules. What’s even more interesting is Corsair’s 4R CUDIMM, which packs 256GB of capacity into a single stick. Corsair said these RAM sticks were tuned to Intel’s recent mid-range desktop gaming CPUs.

What’s even crazier is that Corsair plans to show off this RAM with a “Vortex water block” on top. This add-on also includes a screen to help you monitor PC metrics like temperatures and operating frequencies. —Kyle Barr

Remember Razer’s Project Ava, more commonly known as the “AI Waifu”? It was a hologram in a jar, powered by a generative AI that created its speech and movements. MSI has taken that concept up a notch with its MEG Vision X2 AI+. The concept is a giant PC case with a glass cylinder in the front meant to hold a “Holostage” featuring the company’s lovable dragon mascot, Lucky.

The PC is supposed to run MSI’s LuckyClaw, which seems to be a play on the open source AI agent OpenClaw. It’s unclear whether the hologram Lucky will be able to manage your PC environment for you, but I guess we’ll know more once Computex kicks off properly. I can’t wait to ask AI Lucky if he would prefer to live outside his glass cage. —Kyle Barr

A 15-hour flight, plus a packed week in Taipei, Taiwan, demands a heap of tech to keep me occupied. I’m taking several items that I’m in the middle of testing, including an HP Omnibook Ultra as well as a GoPro Mission 1 Pro action camera and accompanying attachments. For my own pleasure, I’m bringing along a Boox Palma 2 Pro and the Asus ROG Xbox Ally X (covered by DBrand’s Killswitch case), packed with a few games I’ve been meaning to finish, like Esoteric Ebb. I’m also carrying along a Genki Covert Dock 3 that the company sent me for testing, though I’m saving my impressions for later. I have my trusty Ricoh GRIIIX fixed-lens camera and a Lenovo travel dock for any extra photos. And I’ve started rereading Patrick O’Brian’s Aubrey and Maturin series, so that’s why fellow sailing fiction fans may spot Post Captain among my haul. Believe it or not, this is me traveling light. —Kyle Barr

Acer is debuting a whole bunch of gaming-related products at Computex 2026. It’s dropping refreshes of the usual set of Predator and Nitro gaming laptops, including a massive Predator Helios 18, as well as an unexpected streaming-only handheld. The Acer Nitro Blaze Link runs on Linux, not Windows, and is designed for device-to-device streaming via open-source apps like Sunshine and Moonlight.

Acer isn’t offering any cost information for that or any of its other mobile PCs coming later this year. That makes sense, considering memory prices are in such disarray that even major OEMs aren’t willing to offer a clear idea about prices even a few months out. —Kyle Barr

If you thought that Qualcomm’s Snapdragon X chip series was going to be the company’s only computer chip, think again. Its next ARM-based CPU built for laptops is the Snapdragon C. In what’s perhaps the most on-the-nose branding ever devised, the “C” stands for “compute.” We don’t know much about this chip, except that it will appear in budget laptops designed to compete with Apple’s $600 MacBook Neo. —Kyle Barr

Acer offered a sneak peek at its first gaming handheld, which it promises will actually come to the U.S., unlike its still-MIA Nitro Blaze 7. The Acer Predator Atlas 8 is one of the first devices confirmed to sport the Intel Arc G3 and Intel Arc G3 Extreme chips inside, meaning it will pit itself against similar AMD-based devices like the Asus ROG Xbox Ally X. We don’t know much it will cost, but we should hopefully see it on the Computex show floor. —Kyle Barr

Intel’s Arc G3 chips are Team Blue’s attempt to knock AMD from its handheld monopoly. Though it’s built on the same 18A process as the company’s top-end Panther Lake CPU, the Arc G3 chip features half as many P (performance) and E (efficiency) cores. Instead, it’s relying on the Arc B370 and B390 GPUs to potentially deliver high-end graphics performance with—hopefully—solid battery life. We already know of two handhelds that plan to make use of the Arc G3. We should learn more during Computex proper. —Kyle Barr

Nvidia CEO Jensen Huang has been crawling around Taiwan for the last week in preparation for his company’s GTC Taipei showcase that starts June 1. So far, Huang’s been yammering on and on about hyperscalers this and data centers that, but there’s a reason why you should pay attention next month.

There are enough rumors going around to suggest Nvidia could finally reveal its first laptop CPU in well over a decade. Multiple leaks suggest the N1 and N1X chips will be based on ARM microarchitecture, similar to Qualcomm’s Snapdragon X and Snapdragon X2 chips. However, the exciting part is how Nvidia could apply its GPU know-how to an APU (accelerated processing unit) that could compete with AMD’s latest Strix Halo and Gorgon Halo processors.

It all seems promising, though the ongoing leaks suggest that Team Green had to delay its chips for more than a year to squash a mountain of bugs before release. Hopefully, Nvidia will have those issues ironed out before we see the N1X in action. —Kyle Barr

Let the APU (accelerated processing unit) battle commence. AMD’s new Ryzen AI Max+ Pro 495 chip seems beefy since it’s packing 16 Zen 5 CPU cores and a 40 CU (compute units) GPU. Despite that, the new chip doesn’t seem like a major upgrade from the last-gen Strix Halo lineup. That’s mostly because it’s using the same RDNA 3.5 GPU microarchitecture on a slightly updated Radeon 8065S graphics chip.

As if you couldn’t get enough “Halo,” AMD is also producing its own mini PC using both last- and current-gen Halo chips called—of course—Ryzen AI Halo. As you can guess by the title, it’s built for AI. Compared to Nvidia’s DGX Spark “AI supercomputer,” AMD’s version is running on good ol’ x86 and supports Windows. By comparison, Nvidia’s little AI box runs on ARM and only supports a customized Linux backend. —Kyle Barr