Translation: Now let's take a look at AMD's updates at the 2026 Taipei Computer Show. AM5 slot support is extended until 2029. [New News] This is a further extension of the previous commitment, which covered until 2027, while also adding DDR5's EXPO ultra-low latency feature. The platform lifecycle is one of AMD's most underestimated competitive advantages: ensuring that users' motherboards can support future generations of CPUs, effectively reducing total cost of ownership, and firmly locking in the audiophile community and channel ecosystem. In the current situation where N1X and Snapdragon C are squeezing the client market from the Arm camp, and Intel is also recovering strongly, AMD needs this kind of user loyalty. This is a game of extremely low cost and high trust value, in sharp contrast to Intel's frequent replacement of slots. EPYC Venice（Zen 6）。 Previously published This processor, which has 256 cores, is based on a 2nm process, and has a performance improvement of about 70% compared to Turin, was initially released at the AMD Advancing AI event in June 2025. It was subsequently reiterated during the third quarter earnings call and CES in 2025, making it a reiteration rather than a new news at the Taipei Computer Show. But its strategic position remains at the core. The significance of Venice lies in its ability to extend AMD's biggest silent victory - the server CPU share has reached about 40% - into the era of increasingly prominent CPU value in inference; At the same time, it is also the CPU side of Helios, and the scheduled delivery of Venice is a prerequisite for AMD to establish credibility in the rack level market. The substantive update will be announced at the Advancing AI event in July. MI400 and Helios rack. Previously published Both have been released at the AMD Advancing AI in June 2025, including the 432GB HBM4 specification and the Helios launch target in the second half of 2026, so neither is new news at the Taipei Computer Show. Helios remains AMD's most important AI strategic layout to date, and it is also its first true challenge to Nvidia at the rack level system layer - which is where Nvidia has almost absolute say and earns the most profits. Retaining the full release until July for the Advancing AI event is a deliberate rhythm arrangement: AMD is able to have an independent stage that is not diluted by the hustle and bustle of the Taipei Computer Show. The unresolved issue remains the progress of ROCm software, but Nvidia's CUDA moat is still a gap that needs to be bridged. Ryzen 7 7700X3D and RX 9070 GRE are globally available. [New News] The specific SKU, price of $329, release date on July 16th, and Radeon global launch information are all new features of this event. These can be seen more as defensive measures rather than heavyweight highlights, but this is not inappropriate. At a time when the PC market is becoming increasingly crowded due to the influx of two new Arm forces and Intel's strong return, AMD is safeguarding its price competitiveness in the client market. A stable cost-effective team helps maintain OEM partnerships and shipment volumes, thereby providing financial support for the more strategically important data center roadmap. Ryzen AI Max local AI platform. Previously released; new OEM models added This APU platform equipped with 128GB unified memory was released at CES, so the platform itself is not new news at the Taipei Computer Show. What was added are more OEM system models. It remains the most differentiated story in AMD's local AI field and a direct response to Apple's path towards large memory laptops; From a strategic perspective, this is AMD's hedging strategy against the Arm NPU narrative. If local AI truly becomes the driving force for consumers' purchases, AMD can compete with its advantages in memory capacity and bandwidth, rather than handing over the local AI narrative to Qualcomm and Nvidia. Our Advancing AI analysis in July 2025 has laid a lot of groundwork for this. At that time, I highly praised Lisa Su's high "word for action" fulfillment rate, pointing out that the inference segment market (with a compound annual growth rate of about 80%) was AMD's more competitive battlefield, and reminding readers that when others had doubts about AMD's TAM numbers, we never wavered. In terms of EPYC Venice, we have written that AMD essentially only needs to maintain its upward and rightward growth trajectory - which is the correct choice to keep MI400 and Helios until July, rather than consuming them early at the Taipei Computer Show. I once characterized Helios as a truly challenging CPU+GPU+network integration project, as well as a practical path to pressure Nvidia, while retaining the cautious modifier 'there is still a lot of work to be done'. We can also introduce proprietary data: Signal65 testing shows that the inference throughput of MI355X on DeepSeek-R1 can reach up to 1.5 times that of B200.