Advanced Micro Devices (AMD +0.81%) plans to invest more than $10 billion in Taiwan. But this does not mean that Taiwan Semiconductor Manufacturing (TSM +0.71%) will be the only beneficiary.
Instead, the money will go across Taiwan’s broader semiconductor ecosystem, including advanced packaging, chip substrates (the base materials used in advanced chip packaging), and manufacturing capacity for complete artificial intelligence (AI) systems. These investments are expected to run through 2029 and help its partners scale production of next-generation products such as its Helios AI racks.
Hence, CEO Lisa Su is investing now to ensure AMD can manufacture enough hardware if its rapidly growing AI demand translates 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-nanometer process technology. The company also uses TSMC’s SoIC-X and CoWoS-L advanced packaging technologies for some of its AI and data center chips.
The chipmaker is also expanding its supplier ecosystem beyond TSMC. The company is developing next-generation Elevated Fanout Bridge (EFB) chip packaging with ASE Technology and Siliconware Precision Industries. The company has completed testing of a panel-based version of its EFB packaging technology with Powertech Technology. AMD is also working with Taiwanese substrate suppliers and manufacturers that will help produce Helios AI systems at high volume.
The extra capacity could be critical. In July 2026, TSMC CEO C.C. Wei claimed that tight advanced-packaging capacity was limiting customers’ growth. AMD could therefore win AI customers but still miss sales if it cannot package and assemble enough chips.
Need to expand manufacturing capacity
Data Center revenue reached increased 107% year-over-year to $6.7 billion in the second quarter. This business accounted for about 58% of AMD’s total revenue. Management expects Data Center revenue to grow at a compound annual growth rate (CAGR) of above 60% over the next three to five years, including CAGR of more than 80% for data center AI.
AMD will need a much larger supply chain if it comes close to those growth targets. A Helios AI rack contains 72 Instinct MI455X GPUs and 18 Venice CPUs. Meanwhile, OpenAI, Meta Platforms, and Anthropic have announced AMD deployments that could total as much as 14 gigawatts. However, those deployments will occur over several years.
Today’s Change
(0.81%) $3.80
Current Price
$473.25
Key Data Points
Market Cap
Day’s Range
$462.11 – $477.36
52wk Range
$149.22 – $584.73
Volume
14.3M
Avg Vol
28.8M
Gross Margin
50.37%
Being fabless no longer means being capital-light
AMD purchased only $1.2 billion of property and equipment during the first half of 2026. Yet, the company exited the second quarter with $30.3 billion of broader unconditional commitments, primarily covering wafers, substrates, components, cloud capacity, software, and technology licenses. AMD also recorded a roughly $1 billion increase in prepaid expenses and other assets, mainly due to advance payments under supply agreements in the first half of 2026.
Hence, while AMD does not operate chip manufacturing factories, it still needs to commit significant capital to secure supply.
The added manufacturing capacity will matter only if it leads to profitable AI growth. AMD’s non-GAAP operating margin was 27% in the second quarter, significantly lower than the management’s target of more than 35% over the next three to five years. CEO Lisa Su is spending billions to make sure AMD can produce enough AI hardware if customer demand grows as expected.