科技日报美股盘前20260710

TMT BREAKOUTJUL 10, 2026∙PAID Happy Friday.Big focus overnight continues to beMETA +3%while otherhyperscalers flat (more below on META).Semis -1%whileSoftware +0.5%andInternet+0.2%in the green. 周五愉快。隔夜市场焦点持续集中在META上涨3%，⽽其他超⼤规模企业持平（下⽂详述META）。半导体板块下跌1%，软件板块上涨0.5%，互联⽹板块微涨0.2%。 Relatively quiet morning. Asia mainly green overnight: TPX+0.39%,NKY+1.2%,HangSeng+0.6%,HSCEI+0.52%,SHCOMP-1%,Shenzhen-1.24%,Korea KOSPI+2.52%.Memory names mixed but Softbank +10%. 今早市场相对平静。亚洲市场隔夜普遍收涨：TPX涨0.39%，⽇经225涨1.2%，恒⽣指数涨0.6%，国企指数涨0.52%，上证综指跌1%，深证成指跌1.24%，韩国KOSPI涨2.52%。存储芯⽚板块涨跌互现，但软银集团涨10%。 Lots to get to so let’s get to it… 内容繁多，我们直接切⼊正题…… META We wrote aboutMETAin our EOD wrap yesterdayhere. Lots of sell side notes outovernight…a few snippets of some of the more interesting ones below… 我们昨⽇在收盘总结中已对META进⾏了分析（详⻅此处）。隔夜涌现⼤量卖⽅研报……以下摘录⼏份较具参考价值的观点⽚段⸺ META:Why Meta is selling AI models now META：Meta为何现在开始销售AI模型 Sources: When I gotAlexandr Wangon the phone ahead of Meta’s Muse Spark 1.1 release, Iwanted to know how serious Meta was about its new API business for sellingaccess to the model, which now puts it in more compeitition with OpenAI,Anthropic, and Google. Wang said that Meta sees the API as a real business, notjust a learning exercise. Tokens are “gigantic and growing very, very quickly,”he told me, and capturing even a slice of that “can be a very meaningfulbusiness even for Meta at our scale….I floated the theory that Meta built morecompute than it has inference demand, and the API soaks up the slack. Wang saidno; the model is worth sharing on its own, and watching how developers use it willshape what Meta builds next."…“We feel really optimistic about our ability to beat the frontier by the end of the year,” Wang told me.” 在Meta发布Muse Spark 1.1版本之前，我与亚历⼭⼤·王通电话时，想了解Meta对其通过API销售模型访问权限的新业务有多认真⸺这项业务如今让Meta与OpenAI、Anthropic和Google展开了更直接的竞争。王表示，Meta将API视为⼀项真正的业务，⽽不仅仅是学习实践。他告诉我，Token规模"庞⼤且增⻓⾮常⾮常迅速"，即便只分得其中⼀杯羹，"对Meta这样体量的公司来说，也能成为⼀项意义重⼤的业务……"我曾提出⼀个观点：Meta建设的算⼒已超过其推理需求，⽽API正好消化了这部分冗余。王否定了这⼀说法；他认为模型本身就值得分享，⽽观察开发者如何使⽤它，将塑造Meta下⼀步的研发⽅向。"我们对年底前跻身前沿阵营的能⼒感到⾮常乐观，"王告诉我。 META: Semianalysis: “In our view, the race for 3rd iscurrently between Meta and SpaceX, not Google.” META：Semianalysis分析："在我们看来，争夺第三名的竞争⽬前发⽣在Meta与SpaceX之间，⽽⾮Google。" META: BofA Reads Reuters Chip Memo as Positive —14GW 2026/27 Capacity Target Implies Costs Well BelowStreet; Buy,$835 PO META：美银将路透芯⽚备忘录解读为利好⸺2026/27年14GW产能⽬标意味着成本远低于市场预期；给予买⼊评级，⽬标价835美元 BofA flags the Reuters-reviewed internal memo showing Meta targeting 14GW ofcompute capacity additions across 2026–27 (6.5GW in 2026 vs. BofAe 2.6GW), whichif accurate implies ~$22B cost per GW (assuming$145B of capex this year)— lessthan half BofA’s$45B/GW estimate — suggesting engineered cost savings withsignificant positive economics vs. Amazon/Google Cloud annual revenues of$10–16Bper GW. 美国银⾏指出，路透社审阅的内部备忘录显示，Meta计划在2026-2027年间增加14吉瓦的计算能⼒（2026年6.5吉瓦，⽽美银预期为2.6吉瓦），若属实，则每吉瓦成本约为~$22B（假设今年资本⽀出为1450亿美元）⸺不到美银估算的每吉瓦450亿美元的⼀半⸺这表明⼯程成本节约显著，与亚⻢逊/⾕歌云每吉瓦100-160亿美元的年收⼊相⽐，经济效益可观。 BofA views the Iris custom chip as a sentiment positive supporting capacity ROI, though unlikely to drive 2026 savings. 美银认为，Iris定制芯⽚作为情绪利好因素，可⽀持容量投资回报率，但不太可能推动2026年的成本节约。 META: KeyBanc Raises PT to$790 on Product CycleVelocity — Preferred Megacap Into 2H; Maintains OW META：KeyBanc因产品周期速度将⽬标价上调⾄790美元⸺⾸选⼤型股进⼊下半年；维持增持评级 KeyBanc lifts 2026/27E revenue 1% each to$254.6B/$308.1B and introduces 2028Eat$366.9B (4% above Street), arguing Meta’s AI product cycle — Muse Image’sstrong benchmark reception, Spark 1.1’s business-task performance (e.g., Harvey),and API pricing competition — can drive positive revisions plus P/E expansion. KeyBanc将2026/27年营收预期各上调1%⾄2546亿美元/3081亿美元，并⾸次给出2028年预期3669亿美元（⾼于市场预期4%），认为Meta的AI产品周期⸺Muse Image在基准测试中反响强烈、Spark 1.1在商业任务中的表现（如Harvey）以及API定价竞争⸺可推动盈利预期上修及市盈率扩张。 For 2Q, KeyBanc models in-line revenue of$60.2B with EPS 3% above consensus,and sees 3Q guidance of$61–64B with upside to typical q/q seasonality from AIproducts and wearables. KeyBanc’s checks show Family of Apps still generatinghealthy ROI despite macro/geopolitical ad-budget risk. KeyBanc预测第⼆季度营收为602亿美元，与市场预期持平，每股收益较市场共识⾼出3%；预计第三季度营收指引为610-640亿美元，受AI产品和可穿戴设备推动，环⽐季节性表现有望超预期。KeyBanc的调研显示，尽管存在宏观/地缘政治⼴告预算⻛险，Meta旗下应⽤家族仍能产⽣健康的投资回报率。 New$790 PT rolls to 20x 2028E EPS; bull case$1,065 at 25x if new AI revenuestreams gain traction. 新的790美元⽬标价对应2028年预期每股收益的20倍；若新的AI收⼊流获得市场认可，乐观情景下⽬标价可达1065美元（对应25倍市盈率）。 META: Benchmark Maintains Hold on Leaked 14GWMemo — Capacity Ambition Escalates but ROIC BridgeStill Missing META：Benchmark维持持有评级，基于泄露的14GW备忘录⸺产能⽬标升级但投资回报率路径仍不明确 Benchmark reads the leaked memo (14GW by 2027, capex up to$145B) as escalatingambition without addressing the ROIC visibility anchoring its Hold: the incremental7GW implies ~$350–400B of spend at$50–60B/GW, pointing to a step-functionguide raise or off-balance-sheet buildup, while Meta declines to quantify its cloud-capacity commitments. Benchmark认为这份泄露的备忘录（2027年实现14GW产能，资本⽀出⾼达1450亿美元）虽彰显了扩张雄⼼，但未能解决其维持持有评级所