Key Points
- AMD is committing more than $10bn in Taiwan as it prepares its supply chain for a much larger AI business.
- It is targeting more than 60% annual revenue growth for the data‑center segment over the next 3 to 5 years.
- For AMD’s fabless model, advanced packaging and manufacturing capacity have become critical to AI growth.
Advanced Micro Devices (NASDAQ: AMD) plans to invest more than $10bn in Taiwan. The investment is not solely for Taiwan Semiconductor Manufacturing (NYSE: TSM); it will spread across Taiwan’s broader semiconductor ecosystem.
The funding will support advanced packaging, chip substrates, and manufacturing capacity for complete artificial‑intelligence (AI) systems. The initiative is expected to run through 2029, helping partners scale production of next‑generation products such as AMD’s Helios AI racks.
CEO Lisa Su is moving now to ensure AMD can produce enough hardware should rapid AI demand materialize into large‑scale deployments.
Image source: Getty Images
AMD’s next AI bottleneck may not be the GPU
AMD’s new Venice EPYC server CPU is already ramping production using TSMC’s advanced 2‑nm process. The company also leverages TSMC’s SoIC‑X and CoWoS‑L packaging for some AI and data‑center chips.
Beyond TSMC, AMD is expanding its supplier network. It is developing next‑generation Elevated Fanout Bridge (EFB) packaging with ASE Technology and Siliconware Precision Industries, and has completed testing of a panel‑based EFB version with Powertech Technology. Collaborations with Taiwanese substrate suppliers and manufacturers will aid high‑volume production of Helios AI systems.
Having enough packaging capacity is critical; TSMC CEO C.C. Wei warned in July 2026 that tight advanced‑packaging supply could limit customer growth. AMD could win AI customers yet miss sales if it cannot assemble enough chips.
Need to expand manufacturing capacity
Data‑center revenue jumped 107% year‑over‑year to $6.7bn in Q2, representing roughly 58% of AMD’s total revenue. Management expects the segment to grow at a compound annual rate above 60% over the next three to five years, with AI‑related data‑center revenue rising more than 80%.
To meet these targets, AMD requires a larger supply chain. A Helios AI rack incorporates 72 Instinct MI455X GPUs and 18 Venice CPUs. Major cloud providers—OpenAI, Meta Platforms, and Anthropic—have announced AMD deployments totaling up to 14 GW, though roll‑outs will span several years.
Being fabless no longer means being capital‑light
In the first half of 2026, AMD added only $1.2bn of property and equipment. Yet the company closed Q2 with $30.3bn of unconditional commitments, covering wafers, substrates, components, cloud capacity, software, and technology licenses. Prepaid expenses and other assets rose roughly $1bn, reflecting advance payments under supply agreements.
Although AMD does not own fabrication plants, securing supply now demands substantial capital. The added manufacturing capacity will matter only if it translates into profitable AI growth. AMD’s non‑GAAP operating margin was 27% in Q2, below the management’s target of over 35% for the next three to five years. CEO Lisa Su’s multi‑billion‑dollar investment aims to position AMD to meet anticipated AI hardware demand.
