What TradingKey 美股財經 said in this video:
- AMDbullishlong-termAMD has the most complete product portfolio, including EPYC server CPUs, Instinct GPUs, and Pensando networking, allowing it to offer full AI systems. As AI agents raise CPU density per system, AMD does not need to challenge NVIDIA's GPU dominance; its EPYC and next-gen Venice chips alone can capture the large new CPU market. Even as a stable second supplier it can earn significant profits, but risks include Venice production timing, TSMC advanced packaging capacity, and an already-high valuation that may leave investors unsatisfied if results merely meet expectations.Source
- INTCneutralmid-termIntel appears to be the most direct beneficiary of the current CPU shortage, with its Xeon processors having a large installed base in enterprise and cloud markets. If customers urgently need CPUs, Intel could most easily raise prices and expand capacity to reflect profits in earnings, and as a long-time laggard even slight improvement could trigger an explosive stock rebound. However, shortage does not equal an earnings boom; if capacity and execution fail to deliver, cloud giants may turn to AMD or accelerate adoption of ARM architecture. Investors must watch next few quarters' server chip shipments, average selling prices, and data-center operating profit to confirm, otherwise the shortage is just a facade for capacity constraints.Source
- ARMbullishlong-termARM's most attractive point is the long-term trend of cloud players developing their own CPUs. Its architecture is power-efficient and highly customizable, perfectly matching data-center needs; especially as AI agents run 24/7 in the background, data centers now weigh tasks completed per unit of electricity, fully leveraging ARM's power-efficiency strength. However, more ARM-architecture servers do not mean ARM the company captures a proportional share of chip profits, since its revenue mainly comes from licensing fees and royalties while hardware profits stay with cloud giants. Thus, although ARM stands in the right position on a strong trend, earnings elasticity may be smaller than imagined and valuation may already be too expensive.Source
- AI Capexpositive for stocksmid-termIf the CPU cycle is real, AI agents running in the background need large memory to store dynamic data and frequently read/write files, so demand for higher-capacity DRAM and enterprise SSDs will rise rigidly, and their results may even react earlier than CPUs themselves. If data transfer speeds lag, demand for high-speed network cards and switching equipment will also rise, while denser chips increase power draw and heat, pushing cooling systems from air cooling to liquid cooling upgrades. However, these second-layer supply-chain stories are one layer removed from front-end applications; do not buy blindly, but watch whether companies really win new orders and whether gross margins genuinely improve.Source