AMD's $10B Taiwan Bet: Securing AI Supply

Advanced Micro Devices (AMD) has announced plans to invest more than $10 billion in Taiwan, a move that extends beyond its key partner TSMC. The investment will flow into Taiwan's broader semiconductor ecosystem, including advanced packaging, chip substrates, and manufacturing capacity for complete AI systems like its Helios AI racks. These investments are expected to run through 2029.

CEO Lisa Su is making this move to ensure AMD can produce enough hardware to meet rapidly growing AI demand. The company's data center revenue surged 107% year-over-year to $6.7 billion in the second quarter, now accounting for about 58% of total revenue. Management projects data center revenue to grow at a compound annual growth rate (CAGR) of over 60% for the next three to five years.

AMD's next bottleneck may not be the GPU itself but the packaging and assembly. The company uses TSMC's advanced SoIC-X and CoWoS-L packaging for some AI chips, and is also developing new packaging technologies with ASE Technology and Siliconware Precision Industries. It has completed testing of a panel-based version with Powertech Technology.

This investment is critical because TSMC's CEO noted in July 2026 that tight advanced-packaging capacity was limiting customers' growth. AMD could win AI customers but still miss sales if it cannot package and assemble enough chips.

Being fabless no longer means being capital-light. AMD purchased only $1.2 billion of property and equipment in the first half of 2026, but exited the second quarter with $30.3 billion in broader unconditional commitments, covering wafers, substrates, components, and more. The company also recorded a roughly $1 billion increase in prepaid expenses due to advance payments under supply agreements.

This investment is a strategic move to secure AMD's supply chain for the AI boom. With data center revenue growing at triple digits and expectations of over 80% CAGR for data center AI, AMD needs massive manufacturing capacity. A single Helios AI rack contains 72 Instinct MI455X GPUs and 18 Venice CPUs, and customers like OpenAI, Meta, and Anthropic have announced deployments totaling up to 14 gigawatts.

By investing in Taiwan's ecosystem, AMD aims to avoid the capacity constraints that have plagued the industry. This could give AMD a competitive edge over rivals like Nvidia, which also faces packaging bottlenecks. If AMD can scale production efficiently, it could capture more AI market share.

However, the investment comes with risks. AMD's non-GAAP operating margin was 27% in the second quarter, below its target of over 35%. The billions spent on capacity will only pay off if AI demand materializes as expected. If growth slows, AMD could be left with excess capacity and lower margins.

For investors, this signals AMD's confidence in its AI growth trajectory. The company is willing to commit significant capital to ensure it can meet future demand, which is a positive long-term signal. But it also means higher capital expenditures and potential margin pressure in the near term.

The success of this investment will depend on execution. AMD must integrate new packaging technologies and manage its supply chain effectively. If successful, it could solidify AMD's position as a major AI chip supplier and drive significant revenue growth.