"Double supplier" strategy formed? Reports suggest Alphabet Inc. Class C (GOOGL.US) partnering with MediaTek to develop the eighth generation TPU.

According to media reports citing informed sources, MediaTek may be involved in the design of the upcoming chips by Alphabet Inc. Class C (GOOGL.US). The report states that MediaTek may be responsible for I/O chips and backend design services (i.e. packaging). Alphabet Inc. Class C will be responsible for the design of the TPU 8t computing cores for training. Broadcom Inc. (AVGO.US) will be responsible for the design of the TPU 8i cores for inference. The report also mentions that Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR will use its 3-nanometer technology to manufacture the TPU 8t and provide its CoWoS advanced packaging technology. At the Alphabet Inc. Class C Cloud Next 2026 conference held last Wednesday in Las Vegas, Alphabet Inc. Class C Cloud unveiled two new products for the eighth generation Tensor Processing Units (TPUs) TPU 8t and TPU 8i. This marks the first time Alphabet Inc. Class C has split training and inference tasks into separate chips, signaling a significant shift in its AI hardware roadmap. Both chips are planned to be officially available later in 2026. Compared to the seventh generation Ironwood TPU released last November, the TPU 8t offers a 2.8x performance improvement at the same price, while the TPU 8i offers an 80% performance improvement; both chips have more than doubled their performance per watt, with TPU 8t achieving 124% and TPU 8i achieving 117%. The decision to split the eighth generation TPU into two parts is a direct response to the increasing diversification trend of AI workloads. Pre-training, post-training, and real-time inference tasks have significantly different computing characteristics: training tasks focus on ultimate throughput and scalability, while inference tasks are more sensitive to latency and concurrency. A single chip cannot efficiently optimize for both scenarios. Alphabet Inc. Class C stated in a technical blog that the design philosophy of the eighth generation TPU revolves around scalability, reliability, and efficiency. Both chips share the core genes of the Alphabet Inc. Class C AI software stack, but each has been individually optimized to address different bottlenecks. It is worth noting that this is not the first collaboration between Alphabet Inc. Class C and MediaTek. In December last year, reports emerged that with the surge in demand for Alphabet Inc. Class C's AI chip TPU, Alphabet Inc. Class C had doubled the order volume for the next generation "TPU v7e" chip entrusted to MediaTek to ensure a steady supply of next-generation products. This move indicates that Alphabet Inc. Class C is eager to reduce risks through a diversified supplier strategy, and MediaTek's position in Alphabet Inc. Class C's AI chip landscape is rapidly rising, expected to reap substantial profits from this additional order wave. Alphabet Inc. Class C's TPU, with its outstanding energy efficiency, not only meets the internal massive computational demands but also attracts the attention of many external customers. Faced with a surge in market interest, Alphabet Inc. Class C has decided to expand its supply chain system, no longer relying solely on Broadcom Inc., but introducing MediaTek as a new key partner to cope with the supply shortage. Reports indicate that Alphabet Inc. Class C's choice of a "dual-supplier" strategy (Broadcom Inc. + MediaTek) is backed by deep tactical considerations. The current core bottleneck restricting AI chip shipments lies in Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR's CoWoS advanced packaging capacity. Alphabet Inc. Class C, by dispersing orders, can leverage both MediaTek and Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR's long-standing close cooperation to more efficiently secure scarce CoWoS quotas and advanced process capacity.