Inside Computex 2026: Nvidia, Intel and Qualcomm Reveal the Future of AI

[00:00:00] Charlie Campbell: Hi, my name is Charlie Campbell. I'm an editor at large at Time Magazine, and I'm here in Taipei at Computech, the largest IT trade fair in the world. Focus this year is on data centers and AI factories, not just the personal gadgets that might make your lives a little bit easier, but the broad infrastructure that powers the AI revolution. I've shown you ecosystem slides of the past, a computing ecosystem. This is an AI factory ecosystem. Another focus is on affordable AI. Intel is showcasing its Core 3 Wildcat Lake processors, and Qualcomm is revealing its Snapdragon C chips, both specifically targeted at the $300 to $500 laptop range, capable of running on modest on-device AI features, but with great battery life. Another major talking point is the Tai1 Plus AI, or Tai ecosystem, because AI at this scale requires millions of complex components from companies such as Intel, Nvidia, Qualcomm, and Taiwanese manufacturing giants such as TSMC and Mediotech to build out this ecosystem at a phenomenal scale. Jensen Huang, CEO of Nvidia, has just delivered his keynote at the sidelines of Computech. The headline reveal was Nvidia's first foray into PC technology with a partnership with Microsoft. Huang says Nvidia's 4 RTX laptops will reinvent the PC. These include a 1 petaflop super chip, a 4 RTX and CUDA ecosystem, [00:01:27] Speaker 2: and Windows native agents. This computer literally runs everything the world has ever created, plus it now [00:01:36] Charlie Campbell: runs agents. In fact, there were three Windows Nvidia machines unveiled, including a laptop, workstation, [00:01:42] Speaker 2: and desktop. Just like you have a home theater in your house, you have stereos in your house, you want to assist AI agent computers running in your house. And these, in time, it becomes more like C3PO to you. [00:01:54] Charlie Campbell: Huang also unveiled Nvidia Cosmos [00:01:56] Speaker 2: 3, an open frontier omnimodal for physical AI. Whenever you want to create a robot that involves physical world, you now have a companion, a Cosmos 3, that can understand and reason. [00:02:08] Charlie Campbell: Huang said we've now reached an era of agentic AI, where instead of apps and operating systems, we have agents and running within harnesses. [00:02:16] Speaker 2: Huang said this is the reason why this entire system obeys confidential computing. Vero Rubin is really a miracle. [00:02:24] Charlie Campbell: Huang also championed Nvidia's pivot from chips into AI infrastructure, talking up his DSX AI factory ecosystem. [00:02:33] Speaker 2: Huang said we are building a full stack system, helping customers build AI factories and deploying AI factories is incredibly important. And the reason for that is this. Compute is revenue now. [00:02:46] Charlie Campbell: Huang's CEO, Cristiano Aymon, gave Monday's keynote at Computex in Taipei and sat down with time exclusively to discuss its firm's pivot from edge to data centers. [00:02:55] Speaker 3: Huang said we went from a company focused on mobile, we built business across a number of industries, automotive, we went to the PC, we went to industrial, broadband. And now I think we go to this next phase of Qualcomm, which is as we expanded into the data center. Demand is massive, but it's important to understand that the data center is going to phases, right? So in the very beginning, it was all about training. It's about creation of models. Then, as the models get created and you go from creation of models to put models in production, and you're generating tokens, then people realize, I need a more dedicated solution to do inference. Now you're going to the next phase, which is growing so much, you're starting to disaggregate the data center. You have dedicated costs for different tasks. And the interesting thing about data center, it has this issue that we have seen in the mobile industry for a long time. This is the growth of compute that you need. This is the availability of energy. So you have to close the gap and that creates opportunity for companies that have high density of computes and efficient power consumption to enter the space. That's our bet. [00:04:04] Charlie Campbell: Today, downloading apps on our smartphones is second nature, but the rise of AI agents means the days of app stores and operating systems may soon be a thing of the past. [00:04:13] Speaker 3: I think it's important to understand that this transition is very significant. This construct of OSs and app store is for us as a user. It doesn't apply to agents. An agent understands human intention. The agent has the ability. It's not confined by the structure of OSs and app store. The agent can go to the cloud, do things for you, can go to your devices. And the whole concept of an app changed. Two years ago, I provided this example. I provided an example of a banking app. Some developer created the app and thinking, what are the most users of the bank will like the app this way, this menu, this color. Here's going to display information. And the typical banking app will tell you, here's a checking account, here's your balance, here's your transactions. Let's say now the agent has your credentials. You can go to the agent and say, how much do I have in my checking account? Oh, that's how much you have. What's the last three transactions? Those are the last three transactions. So, the whole concept of an app is no longer applicable. So, I think what you're going to see more and more is, besides us, agents use those computer machines, OSs and app stores are a thing of the past. This is changing, and it's changing very fast. [00:05:21] Charlie Campbell: One of today's biggest tech stories has been the resurgence of Intel, the legacy American chipmaker whose share prices soared 450% over the past year. Silicon, the foundation of modern technology. And Intel had plenty of big reveals here at Computex in Taipei, covering gaming chips, physical AI, as well as data centers, showcasing its latest Xeon 6 Plus server processor based on its 1.8 node revolutionary new foundry process. [00:05:47] Speaker 4: So, for many of our customers, they're looking to refresh their data centers and the density of cores that we have in Xeon 6 Plus allows them to really free up space in their data center, power that they can then use for, you know, additional compute that they're going to need elsewhere. Like some of our partners, like Ericsson, they're seeing the ability to go from nine existing servers down to a single Xeon 6 Plus server. [00:06:10] Speaker 5: These same fundamentals can deliver far beyond the PC ecosystem. The demand for our processors at the edge has been booming. [00:06:22] Speaker 6: I'm Daan Kroh, I'm a research engineer at Intel, working on the physical AI framework, and what we have here today is the demo showcase of the full AI workflow. So, we have first the data collection part, so we do imitation learning, so we first show the AI model a lot of the tasks performed by a human. We then trade it on a discrete GPU and afterwards we can use our new Intel Core Ultra Series 3 to make it ready for the edge. [00:06:44] Speaker 5: Derived from the Core Ultra Series 3 and the Arc G3 is a tuned, high performance GPU specifically for handheld gaming. [00:06:53] Speaker 7: We just launched our latest Intel Arc G3 Extreme processor. This is the most performance we ever had in this form factor. So you get great performance for gaming, good battery life, and portability in this form factor. [00:07:09] Charlie Campbell: As AI transforms every industry with an exploding demand for compute, but also memory, in data centers, and at the edge. One person knows more than anyone about this trend is KS Pua, the founder of Taiwan-based memory fund, Phison, who 25 years ago actually invented the USB memory stick. Time caught up with KS at Computex in Taipei. [00:07:28] Speaker 8: Many plus years ago, Phison only with the USB memory card is consumer, then we go to the smartphone, PC, right? Still consumer. But today, AI, they got two parts. One is model training. Model training mainly using the GPU and VRAM, nothing to do with storage. But after model training, mature, the cloud guy, they need to get return, right? Return only go through inference. So when go inference, only two elements we consider. One is the power supply, one is the storage, because the only output is data. Data unit storage. So demand will be incredibly high. One SSD equal to 8 hard drives. 245 terabytes. We won an award. When you use AI, the bottleneck is not compute power. It's a memory space not enough constraint. So DRAM, small capacity, limited production, and price too high. So our invention is using the flash memory as a complementary of a DRAM. Extend the device memory space. Then we can use a limiter of a DRAM. The whole device price affordable. We are limiter of a flash memory. Then one can start to compute AI inference and the fine tuning. Right now, at this moment, the mature side is data center. It's a cloud AI, mature. But AI PC, go to the edge, right? So within 24 months, two years, when the human start to use NVIDIA or Intel or AMD CPU, GPU for edge, this will create the other boom in the storage. In the storage.