This is not investment advice. The author has no position in any of the stocks mentioned. Wccftech.com has a disclosure and ethics policy.
The Taiwan Semiconductor Manufacturing Company (TSMC) is slated to benefit heavily from an uptick in demand for semiconductor products that power data science and artificial intelligence computing systems. 2024’s start has seen Intel, NVIDIA and AMD enter the year with new products designed specifically for A.I. use cases, and the tight nature of the semiconductor industry makes the coming couple of years some of the most interesting ones in its history, especially if today’s industry report from Taiwanese sources stands the test of time.
Tide In Semi Sector Turns To Advanced Machines & Utility Of Contract Manufacturing Model For Firms Such As TSMC
Today’s report comes from none other than the United Daily News (UDN) and it quotes industry sources in Taiwan to speculate that AMD’s artificial intelligence products rely on multiple implementations of TSMC’s 5-nanometer and 6-nanometer semiconductor manufacturing processes. These are integrated through TSMC’s packaging technologies, which have become one of the most highly sought after manufacturing processes in the world.
The UDN’s sources believe that a need to diversify and shift risk will make big ticket A.I. players such as Microsoft and Meta shift some of their computing hardware to AMD. Should this, along with other tailwinds, materialize, then the demand for AMD’s MI300 accelerators can touch 400,000 (40 萬顆) units this year and 600,000 (60萬顆) units in 2025.
Crucial to this million unit demand will be TSMC’s ability to scale up its manufacturing capacity. A tightness in chip packaging has been a concern in analyst circles as a headwind against A.I. ramp up for months now, and earnings conference call have seen crucial comments from TSMC and NVIDIA’s executives as well.
Along with AMD, NVIDIA will also play a major role in beefing up demand for TSMC’s chip nodes this year, according to the report. NVIDIA’s shares soared in 2023, defying a broader market tightness amidst optimism for the endurance of A.I. in enterprise and personal use.
While the report is quick to mention that TSMC has never commented on customer orders and its management has repeatedly stressed the need to be on the watch out for constraints such as global economic slowdowns or high inflation, it nevertheless adds that NVIDIA can ship up to a million units of its A.I. products in 2024. More than double AMD’s potential shipments, this builds up on earlier reports that have suggested that NVIDIA shipped half a million A.I. GPUs during Q3 2023.
Today’s report comes as the turn of the year marks Intel receiving the first high NA EUV chip manufacturing machines in the world. With 3-nanometer products already in mass production, chip manufacturers Intel and TSMC have to focus further down in their product technology pipeline. Since shrinking down circuits to thousands of times smaller than the width of a human hair naturally has its limitations for processes below 2-nanometer.
Further pushing the envelope requires advanced machines and on this front, Intel has already started to get its hands on the new equipment. Unconfirmed whispers in the industry also suggest that TSMC might not use high NA EUV tools until 2030, which provides quite an update from the firm’s official estimate in 2022 of buying the first such machines this year.