AMD commits $10B to Taiwan's semiconductor chip packaging, manufacturing ecosystem

AMD (NASDAQ: AMD) said Thursday it plans to invest more than $10 billion into Taiwan’s chip-making industry and rolled out new processors designed to take on Nvidia (NASDAQ: NVDA) in artificial intelligence computing.

The semiconductor company said the money will go toward Taiwan’s chip production facilities and AI technology development. Most of the investment targets improvements in how chips get manufactured and put together, particularly for next-generation AI systems.

Taiwan remains central to global chip production thanks to Taiwan Semiconductor Manufacturing Co. (TPE: 2330), the world’s largest contract chipmaker. TSMC makes components for major tech companies like Nvidia and Apple (NASDAQ: AAPL)

AMD’s stock has doubled this year as businesses keep pouring money into AI infrastructure. The company is trying to narrow the gap with Nvidia, which reported strong quarterly earnings Wednesday.

“Working with strategic partners in Taiwan and globally, AMD is advancing leading-edge silicon, packaging and manufacturing technologies that enable higher performance, greater efficiency and faster deployment of AI systems,” the company said in its announcement.

The Taiwan commitment will focus on partnerships with local companies to improve chip packaging and manufacturing for future AI hardware. AMD said it’s working with Taiwan firms ASE and SPIL on technology that links chips together, which can improve performance and efficiency.

The upgrades are meant to support Helios, AMD’s AI server platform set to launch in the second half of 2026. Manufacturing partners for Helios include Sanmina, Wiwynn, Wistron, and Inventec.

AMD’s new processor packs 16 cores

AMD (NASDAQ: AMD) also announced its Ryzen AI Max+ Pro 495 chip, trying to get ahead of Nvidia’s rumored laptop processors. The chip has a Zen 5 CPU with 16 cores and 32 threads that can hit speeds of 5.2GHz. But it uses the same RDNA 3.5 graphics architecture from the older Ryzen AI Max+ 395.

The graphics chip, a Radeon 8065S, has 40 compute units and supports up to 160GB of video memory. AMD says this is the first x86 processor that can handle a 300 billion parameter AI model on its own.

Two other versions came out too: the Ryzen AI Max Pro 490 with 12 cores and 24 threads, and the Ryzen AI Max Pro 485 with eight cores and 16 threads. Both have smaller graphics chips with 32 compute units.

AMD’s earlier Strix Halo processors became popular for gaming more than AI work. The company wanted to release simpler versions for gaming devices, but manufacturers canceled handheld gaming systems because of memory supply problems.

$4,000 desktop aims to challenge Nvidia’s developer platform

AMD instead launched a small computer called AMD Ryzen AI Halo that’s six inches by six inches. The $4,000 system has the older Ryzen AI Max+ 395 chip, a 2TB solid-state drive, and 128GB of unified memory.

That puts AMD up against Nvidia’s $4,000 DGX Spark desktop, which uses ARM processors and Blackwell graphics. Nvidia’s system only runs Linux, while AMD’s works with both Linux and Windows, so it can handle regular computer tasks beyond AI development.

AMD claims its AI Max+ processor handles generative AI workloads four times better than Apple’s M4 Pro chip. Apple’s Mac mini has gotten attention lately for running AI software locally.

The Ryzen AI Halo with the 395 chip should be available for preorder in June. A version with the newer Ryzen AI Max+ Pro 495 is “coming soon.” Laptops from Asus, HP, and Lenovo with the new 495+ chip should show up in the third quarter this year.

That timing puts AMD in competition with Nvidia’s expected laptop processors, which will probably debut at the Computex conference in June, along with Intel’s Panther Lake chips.

Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.