Tracked topic
… AMD’s David McAfee said to PC Gamer, in effect, that additional capacity offers hope, but a noticeable easing is more likely to be associated with expansion toward 2028. Tom’s Hardware also reports that the industry does not expect a rapid normalization of DRAM and NAND in the short term. …
… For large AI models, the GPU may still compute at very high speed – but if data or model weights are not available in time at the memory interface, the accelerator waits. That is why HBM has become one of the most strategically important building blocks of the entire AI hardware market. …
… However, the stated AI targets suggest Apple’s own server infrastructure even more strongly. The Verge notes that Apple’s discontinued car project influenced the development of its own AI hardware. …
… Report Reuters Classifies NVIDIA’s AI PC push with RTX Spark and names target groups, OEM support, as well as open demand and cost questions. Report Tom’s Hardware / NVIDIA Documents Jensen Huang’s Computex statements on RTX Spark, AI PCs, and broad support from PC manufacturers. …
… The barely visible details come from an AMD document that is apparently only limitedly accessible, but whose public preview already contains key specifications. Tom’s Hardware and other hardware media have picked up the find. …
… Especially with AI hardware and server components, the market is likely to absorb higher manufacturing costs more readily than in the price-sensitive consumer segment. For classic PC hardware, the effect is therefore more indirect, but not irrelevant. …
… Dass der Standard bereits 2026 finalisiert wurde, zeigt vor allem eines: Der Druck aus dem AI-Sektor ist hoch genug, um Entwicklungszyklen weiter zu beschleunigen. Fazit PCIe 7.0 ist kein Feature für den nächsten Gaming-PC, sondern ein Infrastruktur-Upgrade für die kommenden Jahre der AI-Expansion. …
… Report Tom’s Hardware Explains the earlier expansion of the Shared GPU Memory Override function and the difference between system RAM and dedicated VRAM. …
… The new report therefore fits a clear pattern: Anthropic is scaling not only Claude, but the entire hardware base behind it. Until now, the simpler path for AI start-ups was: rent cloud capacity, train models, sell product. This model is reaching its limits at the largest AI labs. …
… DRAM and HBM have become more strategically valuable due to AI demand. New capacity is urgently needed. If permitting procedures stall or additional requirements become necessary, this may affect long-term expansion plans. …