OpenAI Jalapeño 从 RTL 冻结到流片约 9 个月OpenAI's Jalapeño Takes About 9 Months from RTL Freeze to Tapeout
推特用户 gntalktalk 梳理芯片设计周期变化：OpenAI 自研芯片 Jalapeño 从 RTL 冻结到流片约需 9 个月，Google 正加速迭代 TPU；其提到 OpenAI 与 Amazon 前一天分别与 Synopsys 达成合作。据此其判断芯片设计门槛正大幅下降、设计产能本身在提升，一阶效应落在 EDA，二阶效应将传导到流片与制造，每一次新的 design release 或流片都需要新的光罩组，光罩需求随之上升。Twitter user gntalktalk outlined shifts in chip design cycles: OpenAI's in-house chip Jalapeño takes roughly nine months from RTL freeze to tapeout, while Google is accelerating its TPU iterations; the post notes that OpenAI and Amazon each struck partnerships with Synopsys a day earlier. On that basis the author argues chip design barriers are falling sharply and design capacity itself is rising, with the first-order effect landing on EDA and the second-order effect flowing into tapeout and manufacturing; each new design release or tapeout requires a new mask set, pushing photomask demand higher.