Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/analysts-believe-that-apple-might-no-longer-be-tsmc%E2%80%99s-biggest-customer-this-year-even-though-it-contributes-25-2-percent-to-the-manufacturer%E2%80%99s-annual.21789/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Analysts Believe That Apple Might No Longer Be TSMC’s Biggest Customer This Year, Even Though It Contributes 25.2 Percent To The Manufacturer’s Annual

Daniel Nenni

Admin
Staff member

Citi: TSMC's profits will be strong until 2026, Huida (Nvidia) is expected to overtake Apple and become its largest customer​


Citi Securities analysts are optimistic that TSMC’s AI-related revenue is expected to grow significantly this year. Reuters
Citi Securities analysts are optimistic that TSMC’s AI-related revenue is expected to grow significantly this year. Reuters

Citi Securities analysts are optimistic that driven by NVIDIA, TSMC (2330)' s AI-related revenue is expected to grow significantly this year. NVIDIA may even surpass Apple and become TSMC's largest customer, and its contribution to TSMC's revenue will double. to 20%.

Supply chain data compiled by Bloomberg show that TSMC’s current largest customer is Apple, accounting for approximately 25.2% of revenue, while Huida accounts for approximately 10.1%.


Citigroup bullish on TSMC
Citigroup bullish on TSMC


In addition to Huida, Citi analysts pointed out that the strong growth momentum in demand for special application ICs (ASICs) designed for AI in the next two to three years will also further support TSMC's performance.

In addition, most AI chips will be upgraded to 3nm production starting from the end of 2025. As technology improvements bring higher average selling prices (ASPs), TSMC's profit growth will continue until 2026.

Citigroup predicts that TSMC will set an annual revenue growth target of 20% to 25% in 2025, with gross profit margins maintaining a high level of over 50%. Capital expenditures in 2025 are expected to be between US$35 billion and US$38 billion.

Although it is expected that "on-device AI" equipped with AI functions in PCs and smartphones will shine in 2025, experts believe that the recent financial reports released by companies such as Micron, HP and Dell have revealed the prospects of consumer device-side AI. In a cautious atmosphere, traditional PC and smartphone businesses are still weak. Cloud AI in data centers will still be the mainstream in 2025, and Huida will continue to benefit.

Tae Kim, author of "The Huida Way: Huang Jen-Hsun's Secret to Building a Chip Empire and Leading the AI Wave" believes that cloud AI in data centers may continue to grow to new heights in 2025 because the performance and benefits of cloud AI are much higher than those of devices on AI, and now every device can be connected to the Internet.

In addition to the hot trend of transforming from traditional central processing units (CPUs) to parallel systems relying on graphics processing units (GPUs), there are other drivers for AI growth in 2025, including the further development of AI agents that can complete multi-step tasks, and Expansion of multi-modal AI models that process images, videos, and audio, as well as advances in test-phase computing that allow for more thorough research, inference, and reflection on AI models.

Jintai said that all these developments require more Huida GPU computing power, showing that Huida will continue to reap the fruits of AI growth and benefit suppliers such as server manufacturers and memory chip manufacturers.

Although AMD's GPU chips can compete with Huida in terms of technology, they lack the software maturity of Huida's CUDA platform, and are also weak in the ability to provide complete end-to-end data center solutions, making it difficult to provide comprehensive and unified data center solutions. In order to optimize hardware, system and network technology for larger-scale GPU clusters, AMD may still need to work hard for a while to catch up with Huida's software advantage of more than ten years.

 
Back
Top