This article first appeared in The Edge Malaysia Weekly on May 25, 2026 – May 31, 2026
LAST October, I wrote about an artificial intelligence (AI) bubble in the stock market that was about to pop. By the time 2026 came around, the “bubble” narrative had faded, in part because key AI-related stocks were no longer rising at the giddy pace they were months earlier. At the end of March, US stock benchmark the S&P 500 was down over 7% from the start of the year. It is currently trading up nearly 10% for the year, if you include reinvested dividends. Indeed, the stock barometer has doubled since AI chatbot ChatGPT was introduced in late 2022. American stocks have outperformed markets in Southeast Asia, China, Europe, Latin America, the Middle East and Australasia year to date (YTD). Only the Japan, South Korea and Taiwan markets have done better during this period, though based on one-, five- and 10-year returns, the S&P 500 remains far and away the global leader.
For chip stocks, the fastest growing sector in the US, it is starting to feel like a melt-up. The VanEck Semiconductor ETF (SMH), which represents the broader chip sector, is up 52% this year and 560% since it bottomed in October 2022. The Philadelphia Semiconductor Index (SOX), the other chip barometer, is up 62% this year and 458% from its 2022 lows. Here’s how important the sector has become to the US economy: The chip ecosystem made up just 6% of the S&P 500 in April last year; today, it represents over 22% of the benchmark index.
Shares of microprocessor giant Intel, the premier chip stock at the height of the 2000 tech bubble, are up 205% this year. From its previous peak in June 2000, the stock had fallen a whopping 76% even as the SMH ETF rose nearly 1,180%. Shares of SanDisk, the memory card, USB flash drive and solid-state drive maker, are up 425% this year and a parabolic 4,826% since their lows in April last year. SanDisk was spun off from data storage firm Western Digital, whose shares are up 1,520% since last April. Memory chip maker Micron’s shares are up 743% over the past year. In 2011, it was a US$5 stock. Last week, Micron surpassed US$800, or a 160-fold increase, though it has dipped a bit since. Another big winner is AI chip maker Broadcom. Its shares are up 82% over the past year and 1,416% since late 2017 when I wrote The Edge’s cover story on how its Penang-born CEO Tan Hock Eng was shepherding it towards becoming a chip colossus.
If you are looking for other runaway chip winners, look no further than Seoul. South Korea’s chip-heavy benchmark Kospi Composite Index is up 85% this year and nearly 198% over the past year. Just two memory chip giants, Samsung Electronics and SK Hynix, make up about 48% of the Korean bourse’s total market capitalisation. SK Hynix shares are up 199% since January and 868% over the past 12 months. Samsung Electronics shares are up 147% YTD and 433% over the past year. With a market value of US$1.3 trillion (RM5.2 trillion), Samsung is now the world’s 11th largest listed firm. SK Hynix, with a market cap of US$917 billion, is the 14th largest listed firm on earth, just behind Warren Buffett’s Berkshire Hathaway. Shares of chip foundry giant Taiwan Semiconductor Manufacturing, or TSMC, are up 26% this year and 548% since October 2022. The world’s sixth largest firm makes up 47% of Taiwan’s total market cap. Meanwhile, shares of the UK’s ARM Holdings are up a whopping 123% over just the past two months.
What’s going on? For one thing, the AI world is changing fast. We are moving from chat-based AI, like ChatGPT or Google Gemini, to agentic AI. Last November, Austrian AI developer Peter Steinberger unveiled Clawdbot, an AI-based open-source agent that can execute an array of tasks through large language models using messaging platforms as the main user interface. In January this year, Clawdbot morphed into Moltbot, and soon after was renamed OpenClaw.
Here is what OpenClaw can do: It can automatically browse the web, summarise PDF documents, schedule calendar entries, conduct agentic shopping, and send or delete emails on a user’s behalf. Back in the days of chat-based AI, a journalist like me would just pepper the chatbot with a ton of queries and it would respond to them. The process was cumbersome and while it sometimes saved me a bit of time, it did not give me what I really wanted. The old Google Search was terrible, with a ton of annoying ads popping up all the time. ChatGPT, Gemini and Claude were a breath of fresh air. Eventually, I realised while the search ads may have been annoying, chatbots were not making me any more efficient in finding answers. And, oh, ChatGPT began rolling out ads in late February. This past week, Google announced it would merge its eponymous search with the free version of Gemini. Chatbots are suddenly starting to resemble ad-based search.
Enter agentic AI, the new AI systems capable of taking action on behalf of users without human intervention. It’s basically a software that acts autonomously to do tasks for you. Let’s say you want to find an apartment or a house to rent or buy. All you have to do is to spin up an agent and share your requirements, amount of built-up space — let’s say 2,000 sq ft — and the number of rooms your family requires, plus the neighbourhood you’d like to live in as well as your budget. The agent will then browse the web, filter the homes based on your criteria, and spit out a shortlist. If the listing provides an email address, it could send an email to the renter or seller of the home and ask when you could see it. It will then tell you what day the landlord or owner might be free to show you the place. Of course, you can tell the AI agent that you can only look at the home over the weekend, preferably in the afternoon. It will also make an appointment in your calendar and remind you on that day. No phone calls required, no time wasted surfing the web for hours or using that irritating search engine with ads popping up all the time.
A financial journalist like me might ask the AI agent to get me data on a company I am writing about from the Securities and Exchange Commission, or SEC’s, website, from the CEO’s earnings call last quarter, or from CNBC’s website, and compare that data with other similar firms’. That requires 10 to 100 times more tokens to be generated. What is this token, you might ask. Think of it as the basic unit of data in AI computing. Agentic AI systems produce a lot of tokens. Growth in token utilisation has gone through the roof. At end-January, or before OpenClaw emerged, token growth was about 20% over the previous two months. In the following two months, growth was over 120%. Since then, growth has further accelerated. “Demand has gone parabolic,” thanks to the rise of agentic AI, Nvidia’s CEO Jensen Huang said during the chip behemoth’s earnings call on May 20. “Tokens are now profitable, so model makers are in a race to produce more. In the AI era, compute capacity is revenue and profits,” he said.
What’s happening here is that the AI agent is replacing a human being and working round the clock, searching for data, comparing and analysing it, and then giving me the data so I can then explain to you what’s going on at a particular tech company or sector. I have written this column for The Edge for many years now. Sometimes, getting the data from the SEC, quotes from CEOs’ earnings calls or comparing one company against another can take not hours, but a day or two. AI agents can work 24 hours a day, or 48 hours non-stop, and come back to me the next day, telling me what they found. They can also send emails to my contacts and analysts who I think might be helpful, and even help set up appointments for Zoom chats. If necessary, I can send the AI agent to get me more data, or information from more companies, 10-K or 10-Q filings on the SEC website, all the Substacks that I subscribe to, scan, read and summarise all the analyst reports that I get, so I have more time to write or craft a better story that puts things in a better perspective for readers.
To make an affective AI agent, we need an orchestration engine. We require something that organises all the tasks that an AI agent has to undertake. Nvidia’s graphic processing units, or GPUs, are excellent at training AI. They are good for repetitive tasks. AI training using GPUs helped create great chatbots like ChatGPT. But here’s the thing: Chatbots are so 2022. ChatGPT was a nice little novelty four years ago. Today, we have AI technology and software that can power an AI agent to do things for us.
That’s where CPUs, or central processing units, made by Intel, Advanced Micro Devices or AMD, or ARM Holdings which designs CPUs, come in. In the chatbot era from end-2022 until recently, GPUs were hot. We needed 12 GPUs and maybe one CPU. That went down to eight GPUs to one CPU. The new agentic AI era needs one CPU for every GPU. And how fast will the growth be? On May 20, Nvidia forecast that its revenues will grow an annualised 95% in the second quarter ending July 2026. Imagine CPU demand for agentic AI workloads growing seven times faster. That’s why CPU-related stocks such as AMD, Intel and ARM Holdings have been surging lately. Nvidia is joining the CPU party too. It recently flagged US$20 billion in revenue from sales of CPUs this year, adding to its core GPU revenues.
Here’s another way to look at what’s going on: There are about 1.4 billion knowledge workers around the world who spend the day staring at a computer screen, reading and responding to emails, sharing workloads and working on spreadsheets and an array of different tools. The advent of agentic AI means agents will begin handling some of that work. Instead of trying to figure out how many of those knowledge workers will be fired, focus first on how much more productive every knowledge worker who uses these AI agents becomes. While some knowledge workers will lose their jobs over time, there will be far more new jobs created for them. Want a coffee break? Just tell the AI agent to send a note to your supervisor: “Away from the desk for 10 minutes to fetch a Starbucks Latte.”
All that, in turn, will push AI usage to new heights. That means a lot more compute, a lot more GPUs and CPUs, and memory chips, particularly high bandwidth memory. Here’s how the new agentic AI era is different from the old chatbot era: A chatbot session is quick, GPU-heavy and low-context. An agentic session is just the opposite. It runs for hours, sometimes a day or two, constantly does things, and pulls on CPUs to orchestrate actions, as well as high bandwidth memory, for more context.
Goldman Sachs forecasts consumer and enterprise agents will push monthly token consumption to 24 times current global capacity by 2030. We are still in the very early stages of agentic AI. Over the next two to three years, we will start to see commercial enterprises deploying AI agents on a mass scale. Goldman analysts believe token economics will turn positive this year. Any operator will be able to deploy an agent whose output value exceeds its compute cost. That will force companies around the world to use AI agents to remain competitive. That’s when demand for chips will reaccelerate. Not just GPUs — yes, they will grow too, albeit at a much slower pace than they have in the recent past — but mostly CPUs, as well as high bandwidth memory that will power the race to build AI agents for all of us. It also means the babble about the imminent popping of the AI bubble will only grow louder.
Assif Shameen is a technology and business writer based in North America
Save by subscribing to us for your print and/or digital copy.
P/S: The Edge is also available on Apple’s App Store and Android’s Google Play.
Copyright © 1999-2026 The Edge Communications Sdn. Bhd. 199301012242 (266980-X). All rights reserved
Leave a Reply