"AI Bull Market Narrative" Expands Again! Taiwan Semiconductor Partners with Sony to Bet on "AI Eye"

According to the Zhitong Finance APP, TSMC (TSM.US), known as the "king of chip foundry," and Sony Semiconductor Solutions Corporation, a semiconductor business under Japanese tech giant Sony (SONY.US), plan to establish a large joint venture to manufacture the next generation of high-performance image sensors. According to a joint statement, Sony will become one of the major shareholders and will set up production lines in its new large wafer factory located in Kumamoto Prefecture. Both parties hope that through the joint venture, they can combine Sony's expertise in sensor design with TSMC's exclusive advantages in advanced chip manufacturing processes to further enhance the performance of image sensors.

Sony stated that if the Japanese government provides support, the project could be further scaled up. In recent years, the company has been focused on expanding its image sensor business, extending its application areas from consumer electronics to the electric vehicle industry, and now sees significant potential in fully autonomous driving and emerging humanoid robotics.

Sony CEO Yoshida Hiroki told analysts during a conference call on Friday, "The joint venture with TSMC will be our first step towards a lightweight wafer factory. Previously, we insisted on completing all processes from R&D to manufacturing independently, but in the future, we hope to not only rely on our own capabilities but also to enhance advanced chip manufacturing capabilities by bringing in partners."

The latest collaboration between TSMC and Sony highlights TSMC's efforts to extend its advanced process chips and high-end semiconductor manufacturing capabilities from traditional logic chips, AI GPUs/AI ASIC foundry, and CoWoS, 3D advanced packaging, to the cutting-edge high-performance image sensors, which serve as the "physical AI entry layer."

According to NVIDIA CEO Jensen Huang's insights, "physical AI" emphasizes enabling robots/autonomous operating systems to perceive, reason, and complete a full set of actions in the real world, and an era where "physical AI" assists the evolution of human civilization is approaching. "Physical AI" emphasizes enabling robots/autonomous systems to perceive, reason, and act in the real world, and these three capabilities are the key tools to advance models from "only able to converse" to "able to work in the physical world."

From an industrial logic perspective, the long-term growth narrative of high-end manufacturing leaders such as Sony, Hyundai, and Tesla is shifting from electric vehicle manufacturing to "physical AI-level super platform companies." Once their expected growth engines transform into physical AI segments such as Robotaxi, humanoid robots, and fully autonomous driving platforms, the actual infrastructure bottlenecks will no longer primarily be batteries and vehicle components, but rather high-end logic chips, 2.5D/3D/3.5D advanced packaging, data center storage chips and power chains, large energy storage systems, and core supply chain security under geopolitical war situations.

Meanwhile, Japan's capacity layout is also a key part of TSMC's capacity expansion mainline. TSMC has previously established a manufacturing base in Kumamoto through the JASM project and formed closer connections with Japanese industrial chain forces such as Sony and Toyota; media reports in February of this year indicated that TSMC plans to mass-produce advanced process 3nm high-end chips in Kumamoto, a project believed to be related to meeting AI chip demand and strengthening Japan's semiconductor supply chain security At the same time, the Japanese government is also subsidizing Sony's Kumamoto image sensor factory, with the Minister of Industry explicitly mentioning that image sensors are crucial for future technologies such as autonomous driving and Physical AI.

TSMC and Sony Bet on "AI Eye"

TSMC and Sony plan to establish a next-generation image sensor joint venture in Kumamoto, Japan. On the surface, this appears to be a capacity cooperation in the CIS/image sensor field, but in essence, TSMC is extending its AI semiconductor landscape from "cloud computing chips" further into "physical world perception chips."

Sony's official statement indicates that both parties have signed a non-binding memorandum of understanding, planning to establish strategic cooperation around the development and manufacturing of next-generation image sensors; the joint venture will be majority and controlling owned by Sony and will establish development and production lines within Sony's new wafer fab in Kumamoto Prefecture.

Both parties hope to combine Sony's sensor design capabilities with TSMC's most advanced high-end manufacturing process technology and strongest manufacturing execution capabilities globally, significantly enhancing the performance of image sensors covering cutting-edge technology fields such as consumer electronics, autonomous driving, industrial vision, and even AI humanoid robots.

From the perspective of the semiconductor industry chain, future image sensors will no longer just be core components of smartphone cameras but will serve as the "visual perception forefront" in physical AI subfields such as autonomous driving, robotics, industrial vision, drones, security, AR/VR, and edge AI devices. For AI large models to understand the real world, they first need high-quality, low-noise, high dynamic range, low power consumption, and low latency visual inputs.

Therefore, the next-generation Sony CMOS/stacked image sensors are essentially the perceptual infrastructure of Physical AI. Sony Semiconductor Solutions has recently emphasized its global market share in image sensors, covering a wide range of fields including industrial, consumer cameras, automotive, robotics, and mobile applications.

The AI Investment Frenzy Extends Beyond GPUs and ASICs! "AI Bull Market Narrative" Expands Again

TSMC's US ADR (TSM.US) has recently hit record highs, with a year-to-date increase of 40%. The underlying logic is not only due to strong first-quarter earnings reports but also the increasing capital expenditures on AI computing power by North American tech giants, the scarcity of advanced chip manufacturing processes, and advanced packaging capacity—especially the extreme tightness in supply and demand for CoWoS/3D advanced packaging, along with the global supply chain reconfiguration driving valuation re-pricing. As the theme of AI computing power investment becomes the strongest engine of the global stock market bull trajectory, Asian chip giants have leaped to become the "main axis of the AI bull market narrative," with TSMC, Samsung, and SK Hynix benefiting from the nearly endless demand for AI computing power driven by the AI training/inference wave, with long-term orders and capacity locked in strong demand expectations until at least 2028.

For TSMC, the collaboration with Sony to establish a next-generation image sensor joint venture is not a significant event that can rewrite the main line of AI GPU/AI ASIC foundry business in terms of short-term revenue and profit scale, but it significantly enhances TSMC's non-cloud AI computing semiconductor foundry business landscape. In the past, the market's understanding of TSMC's AI growth mainly focused on the exponential growth trajectory of AI GPU/ASIC, CoWoS advanced packaging, and extensive high-performance computing business revenues driven by Nvidia, AMD, and Broadcom This collaboration with Sony indicates that Taiwan Semiconductor Manufacturing Company (TSMC) is also becoming more deeply involved in the physical AI hardware chain of "AI training/inference from the cloud to real-world perception." In other words, TSMC is not only the foundry base for AI large model computing power but is also becoming the manufacturing infrastructure for robotics, autonomous driving, and edge AI vision systems.

TSMC is enjoying the first layer of dividends from the global "AI computing power arms race," and the next-generation image sensors developed in collaboration with Sony signify that global demand for AI computing power is accelerating towards physical AI subfields such as AI humanoid robots, fully autonomous driving, industrial vision intelligence, AR/VR, and edge AI. This can be seen as a landmark signal that the AI investment boom is expanding from the "almost limitless training/inference computing power landscape" to "visual perception/physical AI."

From the perspective of industry chain investment and capital markets, this is a positive diffusion signal for the "global semiconductor super bull market driven by the AI investment frenzy": the semiconductor bull market dominated by the AI investment boom is expanding from GPUs/HBMs/advanced packaging to high-performance sensors, ISPs, edge AI computing, in-vehicle vision intelligence, and humanoid robot perception chips.

Several market research institutions predict that the image sensor market will continue to grow in the coming years, driven by applications such as autonomous driving/ADAS, AI vision systems, 3D imaging, security, and industrial robots and automation. Research and Markets expects the image sensor market to grow from less than $30 billion in 2026 to approximately $41 billion by 2030, with core growth drivers including autonomous driving, humanoid robots, AI vision, smart cities, 3D imaging, and medical imaging