The semiconductor landscape is undergoing a fundamental shift, moving from a mobile-first volume game to an AI-driven value competition.
According to DIGITIMES analyst Luke Lin, Apple's recent unveiling of the US$599 MacBook Neo is a strategic attempt to expand its ecosystem, but it highlights a growing disparity in revenue contribution at TSMC.
While Apple remains a volume titan, Lin suggests that Nvidia's AI dominance and Google's rising profile are reshaping the foundry's top-tier customer hierarchy.
High volume, hollow gains
Luke Lin's analysis of the Mac Neo — priced aggressively at US$599 — reveals a calculated move to lower entry barriers by repurposing the A18 Pro chips originally designed for the iPhone 16 Pro.
While this leverages Apple's existing silicon investment, Lin points out that even an optimistic scenario of 10 million additional units would only equate to a fraction of annual iPhone sales. Apple is reportedly preparing an initial stock of 5 million units, a figure that Lin describes as "barely moving the needle" for TSMC's total wafer volume.
This conservative outlook is supported by broader institutional forecasts for 2026. Ming-Chi Kuo (TF Securities) and J.P. Morgan align with Lin's baseline, projecting annual shipments of approximately 4.5 to 5 million units. While IDC suggests a potential peak of 5.5 million units driven by the back-to-school season, the consensus remains that the Neo is a "gateway drug" for brand loyalty rather than a volume savior for the supply chain.
The CoWoS divide: Apple's packaging paradox
Lin identifies the "CoWoS Factor" as the primary reason Apple has lost its position as TSMC's largest revenue contributor. Despite Apple's solid 29.93% year-over-year growth, its reliance on mature and cost-effective InFO/PoP packaging limits its financial weight.
In contrast, Nvidia consumes massive amounts of CoWoS (Chip-on-Wafer-on-Substrate) advanced packaging for its AI chips.
Lin observes that while a single 12-inch CoWoS-L wafer carries a premium price tag exceeding US$10,000—a level comparable to advanced-node logic wafers—it remains more cost-effective than the US$20,000+ 3nm wafers used for the A18 Pro.
However, for AI GPU production, the sheer volume of CoWoS-L consumption now rivals or even surpasses that of foundry wafers, marking a significant shift in the semiconductor cost structure.
This technical divide is echoed by early product reviews. While YouTubers like Dave2D and Max Tech have praised the A18 Pro's "thermal headroom" and snappy single-core performance (Geekbench 6 score of 3,461), they also highlight the strategic hardware "ceilings" Apple has imposed — such as the lack of Thunderbolt and a fixed 8GB RAM.
These limitations ensure the Neo remains a budget-friendly volume play that does not require the expensive, high-margin advanced packaging that fuels Nvidia's financial dominance.
A Chromebook rival emerges
Beyond the foundry impact, the MacBook Neo is sending ripples through the PC supply chain. According to Henry Chang, a computing devices analyst at DIGITIMES, feedback from OEM and ODM partners suggests a nuanced competitive landscape.
"While the MacBook Neo is aggressive, its immediate impact on the mass Chromebook market remains limited," Chang notes. "The bulk of Chromebook volume resides in the education sector, where prices typically sit US$100 to US$200 below the Neo's US$599 entry point."
However, Chang identifies a more vulnerable target: Google's high-end Chromebook Plus initiative. "The Chromebook Plus is priced much closer to the MacBook Neo. In this premium-tier bracket, Apple's superior hardware-software integration and ecosystem stickiness give it a significant edge. While the education market is safe for now, the Neo is poised to be a major disruptor for Google's ambitions in the high-end ChromeOS segment over the next two years."
Google's silent ascent at TSMC
Perhaps the most striking insight from Luke Lin is the ascent of Google as TSMC's "hidden" third-largest customer.
While Google's direct orders for its Pixel smartphones are relatively small (approximately 10 million units), its AI infrastructure demand is colossal. Lin explains that Google routes its massive TPU (Tensor Processing Unit) and custom server CPU orders through IC design service firms like Broadcom, MediaTek, and Global Unichip (創意).
When these indirect orders are consolidated, Google emerges as a silent powerhouse that utilizes advanced manufacturing nodes and significant CoWoS capacity. Lin predicts that Google has the potential to challenge for the number two spot at TSMC within the next year.
This shift underscores a new reality in the semiconductor industry: revenue leadership is no longer just about the number of devices in consumers' hands, but about the silicon powering the global AI cloud — a race where Google and Nvidia are currently out-leveraging Apple's mass-market consumer strategy.
Article edited by Jerry Chen
