Wiwynn2026-10-14 01:36:04Wiwynn share allotment lands as OCP showcase highlights diamond cooling and CPO integrationWiwynn, the AI server maker listed as 6669, had its share allotment officially credited on Oct. 14, while the company used the 2026 OCP Global Summit to present an end-to-end server architecture built around four core technologies. According to veteran media commentator Chang Yu-hsuan, the company moved beyond the standardized full-spec rack approach seen elsewhere and pushed deeper into chip-level thermal and power design to address three major AI server bottlenecks: optics, electricity, and heat. The technologies highlighted at the event were diamond-based microchannel liquid cooling, a double-sided cold plate design, Vertical Power Delivery, and full-system CPO, or co-packaged optics, connectivity. The report said the vertical power design shortens the power path and can cut power loss by as much as 80%, lowering total cost of ownership for data centers. The article also tied the product showcase to market attention around Wiwynn’s recent share allotment. After its Sept. 2 ex-rights event, the stock fell from the reference price of NT$2,615 to a late-September low of NT$2,075. FactSet’s latest survey cited in the report put consensus EPS at about NT$121 and the average target price at roughly NT$2,715.470
NVIDIA2026-10-12 15:08:52NVIDIA unveils open architecture updates with partners at OCP Global SummitNVIDIA announced a set of open architecture updates with partners at the OCP Global Summit in San Jose, according to Techub News, citing the NVIDIA Blog. The announcements covered several parts of AI infrastructure, including open design standards for next-generation AI factories, new open-source software for NVIDIA DSX AI Factory, networking work with Meta, a new module standard with Cisco, and transport support planned with Google. Specifically, NVIDIA said it worked with hyperscale data center operators to release open design standards for the next generation of AI factories. The company also introduced new open-source software components for NVIDIA DSX AI Factory. On networking, NVIDIA and Meta advanced an open AI network based on the NVIDIA Spectrum-X Ethernet SuperNIC. NVIDIA and Cisco also announced what was described as the industry’s first 400Gb/s fabric module standard, USFP. In a separate update, NVIDIA and Google said Falcon Transport will be supported on future ConnectX-10 SuperNIC and BlueField DPU products.500
Morgan Stanle2026-10-12 14:45:07Morgan Stanley says market may be overestimating risks from delayed 800V rollout in AI data centersMorgan Stanley said in a new research note that investor concern over delays in deploying native 800V DC power systems in AI data centers may be overdone. The bank said Nvidia’s full transition to native 800V-powered AI server racks could slip to around 2028, instead of arriving with the Rubin Ultra generation that had been expected in the second half of 2027. Even so, Morgan Stanley argued that the broader shift in power architecture remains intact. The note said Nvidia’s latest roadmap still shows third-generation server racks at roughly 570 kilowatts using 54V DC distribution, while native 800V racks may not arrive until the later Feynman architecture, where rack power could reach about 1 megawatt. At the same time, 800V external power supply modules are still on track to ship, and the Power Center centralized power system expected in 2027 is still moving forward. Morgan Stanley added that valuation multiples for 800V-related companies have fallen back to about 6.5x forward EV/Sales from June highs, close to levels seen before Nvidia’s GTC event in March. Yet expected revenue growth for the next 12 months has risen from about 12% to 27%, and revenue estimates have been lifted by roughly 11% since the first reports of deployment delays. The bank said the Oct. 12-15 Open Compute Project Global Summit could act as a catalyst for semiconductor stocks tied to the theme.320
