Sealand: The server CPU has entered a triple resonance period, and the market is expected to enter a price increase cycle.

Sealand's research report states that server CPUs are the core beneficiaries of AI computing power and data center upgrades, entering a cycle of stock upgrades, domestic substitution, and model iteration. This may bring about a reassessment of value, maintaining a "recommended" rating for the computer industry. The CPU market is experiencing multiple drivers such as super-node technology innovation, AI inferencing demand, and a prosperous AI computing power industry chain, which may lead to a price uptrend. In particular, the CPU market in China is more concentrated compared to overseas, and the supply-demand relationship may become more tense. Sealand's main points are as follows: - DeepSeek's introduction of the Engram module may boost CPU memory demand. On January 12, 2026, DeepSeek released a new paper proposing a second path to improve model efficiency beyond the existing "Mixture of Experts" (MoE) model - "Conditional Memory". The Engram architecture aims to solve the problem of efficient storage and retrieval of knowledge. - From HBM, DDR to CPU: The core components of AI computing power are being redefined. CPU is the central brain of computers and its architecture design pursues "generality" - able to handle various complex logic, branch judgments, and serial tasks. It is the foundation that supports all system operations. - Data centers entering an upgrade cycle, CPUs may face new growth opportunities. Top CPU manufacturers continue to update their products. AMD showcased the upcoming Venice series server CPUs at CES 2026. - Large-scale data center construction leads to tight supply of server CPUs. Both AMD and Intel have sold out their server CPU inventories in 2026, with most demand coming from large-scale data center operators. - Risk factors include technology landing falling short of expectations and substitution risks, intensified industry competition and weakened bargaining power, structural fluctuations in industry demand, capacity constraints, policy and international trade risks, dependence on high-end EDA tools and verification software imports.