LPO(线性驱动可插拔)技术
Search documents
AI 的尽头是光:1年600%的LITE、COHR还能涨多久?
3 6 Ke· 2026-02-11 09:29
划重点 ① 2026 年,全球算力基建正式跨入"连接为王"的通信大年。当 AI 集群规模冲向更大级别,1.6T 速率让传统铜缆因电阻热效应撞上物理天花 板,光电转换从备选项跃升为唯一解。这场由大模型迭代倒逼出的"光速互联"革命,正开启一场比 2000 年光纤时代更具实力的价值重估。 ② LITE 靠垂直整合实现从"零件差价"到"系统溢价"的跨越;COHR 通过精简业务、聚焦核心资产以收割 AI 最肥美的利润;黑马 AAOI 则凭 借 LPO 方案的成本优势与微软深度绑定,展现极高弹性。硅光子、CPO 与 LPO 三线并进,核心目标只有一个:消灭能耗与延迟,重塑 1.6T 时代的定价权。 ③ 不同于 2000 年的供给泡沫,2026 年是需求倒逼下的刚需基建。虽然估值波动与技术更迭仍是变量,但投资逻辑已发生质变:核心不再是算 力多强,而是连接能效多高。拥有芯片自研能力与巨头订单的厂商,正接过 GPU 的火炬,点亮 AI 时代的下半场红利。 站在 2026 年来看,全球算力基建的叙事正发生一场深刻的权力交接。如果说 2024 至 2025 年是属于英伟达和芯片代工厂的"GPU 盛世",那么 2026 年则 正式开 ...
AI 的尽头是光:1 年 600% 的 LITE、COHR 还能涨多久?
RockFlow Universe· 2026-02-10 10:32
Core Insights - By 2026, global computing infrastructure will enter a "connection is king" era, driven by the limitations of traditional copper cables and the necessity for optical solutions to meet the demands of AI clusters [3][5][6] - Companies like LITE, COHR, and AAOI are leveraging vertical integration and strategic focus to transition from component pricing to system premium, aiming to eliminate energy consumption and latency [3][6][20] - Unlike the supply bubble of 2000, the 2026 infrastructure is driven by demand, with a shift in investment logic focusing on connection efficiency rather than raw computing power [3][28] Group 1: Physical Limits and Optical Solutions - The physical limitations of copper cables become apparent as signal rates reach 1.6T, leading to significant energy loss and heat accumulation in large AI clusters [9][10] - Optical technology is no longer an option but a necessity, with AI architects prioritizing connection density over raw computing power [11][12] - The share of connection components in total capital expenditure for data centers has surged from 15% to over 30% in three years [12] Group 2: Key Technologies and Innovations - Silicon photonics has crossed the 50% market share threshold in optical transceivers, becoming a critical weapon for cost reduction and production efficiency [14][15] - CPO technology aims to reduce energy demands by integrating optical engines directly with GPU or switch chips, potentially lowering energy needs by 50%-65% [16][17] - LPO technology offers a transitional solution for companies seeking cost-effective options without immediately switching to expensive CPO systems [18][19] Group 3: Competitive Landscape - LITE has transformed from a component supplier to a system provider, with 60% of its revenue linked to AI infrastructure and a revenue growth rate of 65% [23] - COHR is focusing on dual development in AI communication and power semiconductors, having divested non-core assets to concentrate on AI data centers [24] - AAOI, benefiting from deep ties with Microsoft and a flexible LPO approach, is positioned as a strong competitor in the mid-market segment [25] Group 4: Historical Context and Future Outlook - The current market dynamics differ fundamentally from the 2000 bubble, as the 2026 scenario is characterized by demand-driven infrastructure rather than supply chasing demand [28] - Investment focus is shifting from who has the strongest computing power to who can connect the most computing power with the least energy consumption [30]