评级调整：美国IT硬件 - 硬盘与内存：对谷歌TurboQuant的担忧被夸大 - 将西部数据（WDC）上调至跑赢大盘，但仍偏好希捷（STX）和闪迪（SNDK）（均为跑赢大盘）

HDDs & Memory: Google's TurboQuant fears overdone - UpgradingWDC to Outperform but still prefer STX and SNDK (both OP) HDD and Memory stocks have sold off significantly due in part to fears from Google’sTurboQuant report. This however, should have zero impact on HDD demand and negligibleimpact on NAND demand. Given the stock sell-off we see this as an attractive entry point forSTX, WDC and SNDK and upgrade WDC to Outperform. Mark C. Newman+1 212 845 7822mark.newman@bernsteinsg.com April Li+1 917 344 8339april.li@bernsteinsg.com TurboQuant sell-off provides attractive entry point.On March 24, 2026, GoogleResearch unveiled TurboQuant, a novel compression algorithm that was reported to achieve6x reduction in KV cache memory and up to 8x inference performance gains on NvidiaH100 GPUs with zero accuracy loss (see report here). TurboQuant compresses only theKV cache — a specific component used during inference (and not training workloads). Itdoes not compress model weights, training data, or any data at rest. Its potential impacttherefore is mainly to GPU HBM or system DRAM, with some indirect impact to NANDwhich is only used to offload cold caches. There is zero impact to HDD demand. Phoebe Sun+1 917 344 8481phoebe.sun@bernsteinsg.com Attractive entry points for STX, WDC and SNDK given ~20% sell off from recenthighs.Despite the fact that this has zero impact on HDD and negligible impact on NAND,stocks in our coverage have sold off dramatically with WDC, STX and SNDK down 21%,17% and 26% from recent highs respectively, and most of this decline was since theTurboQuant blog on March 24th. Demand and pricing continue to be better than expected on the back of AI.AIworkloads, richer content creation, longer data retention requirements, and stricter datasovereignty standards are aiding both demand and ASP. NAND price rally also providesmore headroom for HDD ASP. We now expect the combined revenue of WDC and STX togrow at a 24% CAGR from FY25 to FY30, supported by 24% bits growth and stable flatASP. This compares favorably with our prior model assumptions of 18.7% bits growth and-3.6% ASP decline. WDC’s innovation day suggests a solid ePMR roadmap in the short-term but also asoft push-out of HAMR.While we were impressed by WDC’s ePMR (legacy tech) roadmapwhich effectively stretches their legacy tech 1-2 years longer than expected, we believethis was effectively a soft push out of HAMR due to lack of tech readiness. On net thoughwe see this as anattractive entry point for WDC, upgrade to Outperform, raisingestimates and TP to $340 on 20x FY28 EPS (vs. 18x previously) with 5 year EPSCAGR of 46% (FY25-30E). STX’s HAMR advantage to become meaningful in FY27 and FY28.Our HAMRthesis remains essentially unchanged but has yet to play out. While WDC is emphasizingits dual track roadmap, STX is all-in on HAMR and we believe the higher areal densityimprovements from HAMR will drive accelerating cost declines. This will play moremeaningfully in FY27 and beyond.Reiterate STX Outperform, increasing TP to $620 on21x FY28 EPS(multiple unchanged), and remains our top pick in HDD and indeed in oursector on 12month+ horizon. With 5 year EPS CAGR of 51%. BERNSTEIN TICKER TABLE INVESTMENT IMPLICATIONS Weupgrade WDC to Outperformon the back of the recent sell-off which provides an attractive entry point. Our new TP forWDC is $340, which is 20x (vs. 18x before given higher EPS growth) our FY28 EPS. PE of 20x is inline to lower than recentmarket multiple and arguably conservative given its 5 year EPS CAGR of 46% (FY25-30E). Wereiterate Outperform for STXand increase TP to $620 on 21x (mutliple unchanged) our FY28 EPS of $29.5. We believeSeagate's strong fundamentals, HAMR leadership, and >50% EPS CAGR (FY25-30E) easily justifies a 21x P/E, and arguablymuch higher. Wereiterate Outperform for SNDK, TP $1000 on 11x FY27 EPS (unchanged) Table Of Contents Part I: Google’s TurboQuant Sell-off Provides Attractive Entry Point....................................................................................................................3Google’s TurboQuant - Another DeepSeek moment?......................................................................................................................................... 3Attractive entry points for STX, WDC and SNDK given ~20% sell off from recent highs....................................................................... 3Part II: The Data Explosion and AI demand drives Higher HDD Shipments........................................................................................................ 4Part III: HDD ASP Has Room To Grow Further as NAND Price Rallies................................................................................................................. 6Part IV: Tech Roadmap update and impact on Cost and profitability....................................................................................................................9Part V: WDC - Upgrading to Outperfor