The AI boom has already rattled the global memory market, with surging demand for high bandwidth memory (HBM) absorbing DRAM capacity and squeezing supply for other uses. Now, that same AI-driven reshaping of the semiconductor industry appears to be spreading into smartphone chips, as MediaTek shifts attention and resources away from its mobile SoC business and into fast-growing AI-focused custom silicon.
A new report from Taiwan’s CTEE says MediaTek has reassigned some resources, including staff, from its mobile chip division toward so-called “blue ocean” categories like AI ASICs and automotive silicon. In industry terms, a blue ocean market is one with abundant opportunities and far less direct competition than mature, crowded segments such as smartphone processors. The takeaway is simple: MediaTek believes the biggest growth in chips over the next several years won’t come from battling for smartphone market share, but from building specialized AI hardware for data centers and enterprise customers.
That strategic pivot is already visible in MediaTek’s expanding ties with Google. MediaTek recently contributed to Google’s TPU v7 “Ironwood” project by designing input/output modules that help the processor communicate efficiently with connected components. This is notable because Google has often taken a tightly controlled approach to TPU development, typically leaning on established partners for collaboration on major design elements.
MediaTek now aims to deepen that relationship as Google’s next-generation TPU ramps toward mass production. Volume production is expected to begin in Q3 2026, with Google reportedly targeting manufacturing of around 5 million units in 2027 and 7 million units in 2028. To support those volumes, both Broadcom and MediaTek are said to be increasing wafer starts, reflecting how serious the scale of these AI accelerators could become.
Complexity is rising too. With Google’s TPU moving to TSMC’s advanced 3nm process, design and integration demands intensify. MediaTek’s response, according to the report, is to carve out a dedicated team for ASIC ambitions—an effort that requires pulling resources from elsewhere, including the mobile SoC group.
A major piece of MediaTek’s AI ASIC strategy is its in-house SerDes technology (serializer/deserializer). SerDes is critical in modern AI compute because it enables high-speed, efficient data movement between processors, memory, and peripherals—often becoming a decisive factor in overall accelerator performance and scalability in data center environments.
MediaTek’s current 112Gb/s SerDes DSP uses a PAM-4 receiver architecture and is said to deliver more than 52dB of loss compensation at a 4nm process, while maintaining low signal attenuation and strong resistance to interference. These characteristics matter in data centers and advanced packaging designs, where signal integrity and throughput can make or break real-world performance. MediaTek is also working on a next-generation 224Gb/s SerDes DSP, signaling continued investment in even higher bandwidth solutions.
Financially, MediaTek appears confident that this shift will pay off quickly. The company reportedly expects around $1 billion in revenue from AI ASICs in 2026, rising to “several billion dollars” in 2027. Beyond Google, MediaTek is also said to be pursuing a broader custom silicon relationship with Meta, which could further accelerate its ambitions in bespoke AI hardware if it materializes.
All of this suggests MediaTek increasingly views AI as a long-term structural transformation of its growth engine. But it also raises a key question for smartphone enthusiasts and Android device makers: what happens to Dimensity?
For now, MediaTek’s Dimensity chips remain highly competitive, and the company is still positioned to deliver strong smartphone performance. The broader mobile chip landscape is also moving forward aggressively, with both Qualcomm and MediaTek shifting upcoming flagship designs to TSMC’s N2P node, aiming for higher clock speeds with improved efficiency characteristics. Even so, if MediaTek continues to deprioritize its mobile SoC division in favor of AI ASICs and automotive silicon, it could become harder to sustain rapid innovation cycles in the fiercely competitive smartphone processor market.
That risk is amplified by the fact that the very top of the mobile SoC hierarchy has long been dominated by Apple’s A-series chips and Qualcomm’s Snapdragon lineup. Dimensity has made major gains, but it hasn’t consistently been the uncontested leader in premium mobile silicon. If MediaTek’s top engineering focus migrates toward data center-class AI hardware, the long-term competitiveness of future Dimensity generations could face new pressures—especially as rivals continue to invest heavily in smartphone performance, efficiency, and modem integration.
In short, MediaTek’s move highlights a broader reality in semiconductors: AI is no longer just a feature—it’s reshaping priorities, talent allocation, and capacity planning across the entire chip industry. And the ripple effects may soon be felt not only in data centers, but also in the smartphones people carry every day.
