When Lisa Su met Foxconn in Taipei on 6 October, with a meeting with TSMC also due, AMD’s chief executive was preparing to increase output of central processing units and graphics processing units in 2027 to meet AI demand. The plan is best read as a response to a market in which personal AI agents are creating demand for the server CPUs that organise their work, alongside the accelerators more commonly associated with AI computing.
AMD crossed $1 trillion in market value for the first time on 21 September, when its shares rose about 10% to $613.92. It became the fourth US chipmaker to reach the threshold, after Nvidia, Broadcom and Micron Technology.
The milestone caps a sharp change in AMD’s position under Su, who took over on 8 October 2014 when the company’s market capitalisation was just over $2 billion. In nearly 12 years, AMD’s value increased by about 500 times to $1 trillion.
The immediate commercial opportunity lies in AI systems that need both kinds of chips. AMD said in its first-quarter earnings report that inference and Agentic AI were increasing demand for CPUs and accelerators at the same time. Its data-centre business generated $6.7 billion in the second quarter of 2026, up 107% year on year and accounting for about 58% of quarterly revenue.
AMD had laid groundwork with large customers before personal agents arrived. In October 2025, OpenAI signed a multi-year agreement to deploy 6 gigawatts of AMD GPUs, with the first 1 gigawatt scheduled to begin in the second half of 2026.
Meta followed in February 2026 with a long-term deployment agreement for up to 6 gigawatts. Its first batch was to use custom GPUs based on the MI450 architecture, paired with EPYC processors and Helios.
AMD launched its Helios rack-level system in July 2026. The system contains 72 Instinct MI455X processors and 18 EPYC Venice processors, with Pensando networking and ROCm integrated into the system.
The newer agent products illustrate why the CPU side matters. Meta’s Muse and OpenAI’s Dots run on AMD’s EPYC-branded CPUs. Muse uses two EPYC cores, while Dots runs on nine cores through an AMD EPYC 9V74 processor.
Muse launched in September 2026 and reached 5 million downloads in 22 days. Sensor Tower said it held first place on the US App Store’s free iPhone chart for 12 consecutive days.
The chips powering the two agents cost below $3,000 on the secondary market, compared with Nvidia GPUs that can top $30,000 apiece. Morgan Stanley estimates that Muse could account for 20% of AMD’s 2026 chip sales.
AMD holds about 46% of the x86 CPU market, according to Mercury Research data. Its shares extended a five-session winning streak after the $1 trillion threshold, gaining roughly 24% over that period. The stock has gained 32% over the past month as of October 2026 and surged by roughly 187% so far this year.
The company’s forecasts show the scale of the opportunity it sees. In July 2026, AMD put the total CPU market at $220 billion in 2030 sales, up from an earlier estimate of $60 billion. That month, it raised its estimate of the potential market it could address in 2030 to about $2 trillion.
Su’s planned supply increase will determine how much of that demand AMD can serve. The company says it will sharply increase chip supply in 2027 as AI demand rises.
Comments (0)