Muhammad Zuhair
Data-Center Researcher at SemiAnalysis
MediaTek has ample ASIC opportunity as Google's TPU ramp accelerates and Broadcom's pricing becomes a problem for Google. Per Mizuho, TPU shipments could top 35M in 2028, about 8x 2026, while Broadcom's CoWoS plus N3 package has become expensive. MediaTek's potential challenges are 224G SerDes IP integration and EMIB-T packaging yields, though checks by @sssjeffpu put EMIB-T above 95%, so yield should not be a major problem now. Intel and Broadcom also have ASIC opportunities. x.comÂ
MediaTek is eroding Broadcom's ASIC moat as Google diversifies its TPUs toward it. Google moved to MediaTek for TPU v8, and a JPMorgan note says the TPU v9 volume shift to MediaTek will be significant, leaving Broadcom only an LPU-like chip in the v8 series. MediaTek's edge likely comes from its closeness to the Taiwanese ecosystem and its key role for Nvidia with GB10 and the RTX Spark. How it plays out remains to be seen. x.comÂ
Broadcom's integrated optics, electronics and packaging could lower custom-chip system costs because separate suppliers defend their margins. At similar valuations, it is preferred to Nvidia and has more room to beat expectations. Google TPU v8 shipments are estimated at 4.0M in 2026 and 9.1M in 2027, assuming Broadcom supplies the remainder after MediaTek's allocation. Anthropic financing adds value only if chip profit exceeds likely support payments and capacity costs. x.com x.comÂ
Google TPU adoption could broaden the AI hardware opportunity beyond Nvidia, with demand reaching custom ASICs, HBM, packaging, optics and testing. Per The Information, OpenAI reportedly plans to buy Google's TPUs and is hiring compiler engineers, but no purchase is confirmed. Broadcom, SK hynix, Samsung, Micron, TSMC, Coherent, Lumentum, FormFactor and Teradyne could benefit. x.comÂ
MediaTek's upside from designing Google's TPU looks limited because inference is the bigger market and the biggest players have their own chips. They will push their own silicon as much as possible and have no incentive to try TPUs; TPU had an edge two years ago, but Google did not act in time. That counters @KairosPraxis's case that the market underestimates MediaTek's TPUv9 ASIC work. x.comÂ
Google's growing TPU business needs HBM, adding demand for Micron, SK hynix and the memory basket. Google Cloud rose 103% to $30.7B, with backlog reportedly $561B; UBS models more than double TPU shipments in 2027. Micron should be up double digits on this, not down 5%. Samsung has LTAs with 10 major customers and is negotiating with 10 more, targeting 60-70% of capacity under contract. x.com x.comÂ
Intel's EMIB-T packaging could reaccelerate now that the root causes of substrate yields around 45-50% have been identified. Momentum for EMIB-T is rising for Google's TPU v10, supporting the mid- to long-term positive view despite the v9 timing slip. x.comÂ
Broadcom and MediaTek risk being bypassed after Google reportedly told TSMC it wants to become a major direct customer, per Tech Taiwan. That points to Google's ultimate goal of an Apple-style COT model, though both partners internally still see the shift as a distant threat. x.com x.comÂ
MediaTek should heavily outperform over the next two years. It is preferred over Qualcomm, which has photonics upside from acquisitions such as Alphawave's optical networking IP but too much exposure to China. x.com x.comÂ
Broader adoption of custom AI silicon could benefit Broadcom and Marvell. OpenAI reportedly plans to expand its use of Google's TPUs beyond Nvidia GPUs, and Google TPUs are coming to Oracle Cloud in 2027. Both are potential tailwinds for custom-chip suppliers. x.comÂ
Broadcom and MediaTek risk losing chip ownership and margin as hyperscalers pull custom silicon in-house for platform control and margin. @jukan05 relays a rumor that Google will move to customer-owned tooling from TPU v9, ordering wafers directly from TSMC like AWS's Annapurna and reducing Broadcom and MediaTek to IP and design-service roles at lower per-chip margins. Even if not now, it will happen eventually, because the end state is all about optimization. x.comÂ
Google's MediaTek-built TPU v9 should lift Semtech's content per chip. Per @StockSavvyShay, the Triggerfish variant adds SRAM, HBM4E memory and a simulation die, so data must travel farther and faster in the rack, pushing electrical links to their limit. MediaTek's SerDes is leaner than Broadcom's, so links need extra help, which Semtech's signal conditioners and redrivers supply; Google reportedly already relies on them. More Semtech parts attach with every accelerator Google ships. x.comÂ
