概念词典 · 宏观
AI 资本支出
Mag7 等科技巨头投入 AI 数据中心、GPU、电力、光互联等基础设施的总额，是上游产业链景气度的最直接先行指标。
别名： AI capex／ AI 资本支出／ AI capital expenditure／ capex／ 资本支出
是什么
AI capex（AI 资本支出） 指各大科技公司投入 AI 基础设施的资金总额，包括：
- 购买 GPU / TPU / ASIC 芯片
- 建设新数据中心（土地、电力、冷却）
- 光互联设备 / 网络设备
- HBM 内存等核心零部件
2026 年，Mag7 中 $MSFT、$META、$GOOGL、$AMZN 四家合计 AI capex 突破 3000 亿美元，并且每季度财报都在上修指引。
为什么是上游产业链最重要的”先行指标”
Mag7 的 capex 计划公布后，约 6-12 个月会传导到上游：
Mag7 财报 capex 指引上修
↓
NVIDIA 拿到更多 GPU 订单
↓
TSMC 4nm/3nm 产能排队更长
↓
HBM (Micron/Hynix/Samsung) 供不应求
↓
光互联 (Coherent/SIVE/AAOI) 订单暴增
↓
最上游 InP 衬底 (AXTI) 出现真正的瓶颈Serenity 的关键判断
他对 AI capex 的态度是 “指数级增长且没有放缓迹象”：
“Capex is ramping up. Exponentially.
And flowing directly down to: Neoclouds → Connectivity → Energy → Semi/foundries.”
他常用 “AI capex 一旦放缓，整个供应链估值要重估” 作为最大尾部风险——但截至 2026Q2，他认为这个风险还没出现。
看 capex 的几个细节
- 看绝对额变化（如 Microsoft Azure 季度 capex 同比 +50%）
- 看资本支出占营收比例（健康范围还是过度投入？）
- 看构成（GPU vs 数据中心建设 vs 网络）—— GPU 占比上升说明算力还在加速
提到此概念的推文 (271)
按时间倒序 · 滚到底自动加载更多
-
驳斥克拉默观点，基于基本面看好内存与存储股，中期看多 Leopold 持仓。
今天看到吉姆·克拉默（Jim Cramer）支持对冲基金做空 Leopold 关于 $SNDK / $DRAM / SK Hynix 的“胡言乱语”式看涨观点。同时克拉默称 $CRWD 为“必买”。我不确定克拉默是否看到一家被贴上“网络安全”标签的 2410 亿美元市值公司就认为其安全，而忽略了所有基本面。但就 CrowdStrike 而言，这是一家 2028 财年远期市盈率（Forward P/E）约 145 倍、营收同比增长 26% 的公司，克拉默却认为它是“必买”。然后看看 Leopold 做多的 AI 资本支出受益主题。像 $SKHY / 三星在 $DRAM 领域……分析师预测 2027 财年远期市盈率仅约 2.8-3 倍。-> 三星签订至 2031 年的长期协议（LTA）（70% 产能）以确保耐用性…… -> SK Hynix 提供最低价格底线 + 无价格上限类型的合同。-> 未来一年内存短缺将更加严峻。-> DRAM/NAND 占超大规模云厂商资本支出的比例更高（Trendforce 估计为 68%）。-> 2027 年超大规模云厂商资本支出预测全部上调。至于 $SNDK，2030 年自由现金流（FCF）利润率预计达 50%，毛利率达 80%。我唯一不同意 Situational Awareness 持仓的是 $CRWV，因为债务利息侵蚀自由现金流，但我可以对 $BE、$INTC、$AMD 等提出类似的看涨论点。短期尚不确定，但中期来看：鉴于压倒性的基本面，很难不认为 Leopold 的持仓会走出与克拉默截然相反（Inverse Cramer）的走势。
英文原文
I've seen comments from Jim Cramer today supporting hedge funds that short Leopold's "nonsense" $SNDK / $DRAM / SK Hynix type calls. As well Cramer calling $CRWD a "Must Buy". I'm not sure if Jim Cramer sees a $241B company labelled "Cybersecurity" and thinks of it as safe... while ignoring all their fundamentals. But with Crowdstrike, you have a ~145x FY2028 fwd P/E company, growing revenue 26% Y/Y. That Cramer thinks is a "must buy". And then you turn your eyes to the AI capex beneficiary theme that Leopold is long on. With names like $SKHY / Samsung in $DRAM... Both FY2027 ~2.8-3x fwd P/E off analyst projections. -> Samsung LTAs to 2031 (70% capacity) for durability... -> Minimum price floor + no price ceiling type contracts from SK Hynix. -> Memory shortage much tigher for the next year. -> DRAM/NAND % of hyperscaler capex goes higher (est. 68% per Trendforce) -> 2027 hyperscaler capex projections all hiked. Then with $SNDK, 50% FCF margins, 80% gross margin projections into 2030. Only Situational Awareness position I'm in disagreement with is $CRWV due to debt interest eating FCF, but I can make a similar bullish argument with $BE, $INTC, $AMD, and others. Not quite sure about short-term, but over the medium run: It's hard to not see Leopold's positions pulling an Inverse Cramer due to overwhelming fundamentals.
-
Anthropic报告预示AI将重塑经济结构，利好资本密集型AI硬件板块。
Anthropic 的《经济情景》报告读起来很有趣：1. Anthropic 的“极端”情景显示 2030 年 GDP 增长达 15.4%，约为当前增速（~2.1% Y/Y）的 7.3 倍，这与 Elon Musk 关于 AI 将超越国债的说法一致。2. 几乎所有分歧发生在 2027 年之后。模型显示 2026-2027 年分离度较小，随后曲线在 2028-2030 年急剧弯曲。我的解读是：2026-2027 为基建年，2028-2030 为经济加速期。这与 $AMZN / $META / $GOOGL 的资本支出及短期 FCF 逆风（但 2028 年后预计巨额盈利）相吻合。3. 极端情景下，知识工作者失业率升至 17.9%，整体失业率 11.9%。“个人层面意味着程序员和客服可能需转行做电工和护士，这些受 AI 影响较小。”计算机科学专业 GG。4. 非认知工资：极端情景下比基准高 33.6%。电工/建筑类劳动力变得极具价值？（但这尚未计入人形机器人...）而 2030 年知识工作者工资比基准低 11.5%。5. 总资本收入比无 AI 时高约 81%。生产力爆炸 -> 总劳动收入几乎不变 -> 增量 GDP 几乎全归资本。因此资本所有权而非劳动... 使得计算、电力、半导体、网络、机器人等股票变得非常重要... TLDR: - 经济增长起飞 - 需思考的工作 = 失业 + 降薪 - 体力劳动更重要，如电工 - AI 股票起飞。
英文原文
Anthropic's Economic Scenario publication was a pretty interesting read: 1. Anthropic's "extreme" scenario has GDP growth reaching 15.4% in 2030 So around 7.3x current rates (~2.1% Y/Y), which is consistent with what Elon Musk is saying about outgrowing the national debt with AI. 2. Almost all of the divergence happens after 2027. Their model shows relatively little separation in 2026-2027, then the curves bend sharply into 2028-2030. So my read is that it points to 2026-2027 as buildout years and 2028-2030 as economic acceleration. Kinda gets mapped as well when you look at $AMZN / $META / $GOOGL capex, and short term FCF headwinds (but massive projected profitability post 2028) 3. In an extreme scenario, unemployment of knowledge workers goes to 17.9%. With 11.9% economy-wide unemployment "At the individual level, it means coders and call service center agents may have to switch to jobs like electrician and nurse, which are less exposed to AI." GG computer science majors. 4. Non-cognitive wages: +33.6% vs. baseline in an extreme scenario. So your electricians/construction-type labor becomes extremely valuable? (but this isn't factoring in humanoids yet...) Whereas -11.5% for knowledge workers in 2030 vs. baseline. 5. Aggregate capital income is ~81% higher than it would have been without AI Productivity explodes -> aggregate labor income barely changes -> essentially all of the incremental GDP accrues to capital. So that makes ownership of capital rather than labor... so equities like compute, power, semis, networking, robotics, etc. become pretty important... TLDR: - Economic growth go brr - Jobs that require thinking = unemployment + lower wages - Physical labor = more important, like electricians. - AI equities go brrr.
-
存储瓶颈未变，供需失衡加剧，基本面与短期价格常背离。
既然 $MU 到 $SNDK 都实现了大幅反弹。很高兴能提醒一下，存储瓶颈（Memory Bottleneck）并没有改变！连续波激光器（CW Lasers）、衬底（Substrates）或其他领域也是如此。但短期情绪（取决于价格/宏观）往往会变化。我认为许多供需失衡（Demand Imbalances）会比人们预期的更严重：- 今天日经新闻提到一些日本分销商称存储需求缺口已达 40-60%（比供应高出 67-150%）。随着普遍价格到年底上涨 50%。- $SPCX 未包含在 $1.3T 超大规模云厂商资本支出（Capex）数据中（我认为 Wells Fargo 等估计 AI 资本支出约 2630 亿美元），因此总资本支出数据可能会令人惊讶。- $SNDK 预计 80% 的毛利率将持续到 2030 年……（欢迎来到标普 100 指数）- 再次强调，现在对 2031 年有了长期的可见性。像 Samsung Memory 这样的公司波动性非常大……我的一些持仓上涨了 270%+，所以我发现更容易在波动时期持有。但无论如何，运营基本面并不总是与短期价格走势一致。同样的概念也可以应用于其他板块。
英文原文
Now that $MU to $SNDK have pulled a major recovery. Good to have a reminder that the memory bottleneck hasn’t changed! Same with CW lasers… substrates… or others. But short term sentiment (depending on prices/macro) often does. I think a lot of demand imbalances get worse than people expect: - There’s some Japanese distributors in Nikkei today saying memory demand deficit has reached 40-60% (67-150% higher than supply). With general prices going up 50% by years end. - $SPCX isn’t included in the $1.3T hyperscaler capex figures (I think Wells Fargo ets. ~$263b AI capex), so total capex figures might be surprising. - $SNDK expects 80% gross margins to continue into 2030… (welcome to S&P 100) - and once again you have long term visibility into 2031 now with companies like Samsung Memory is very really volatile… some of my positions are up 270%+, so I find it a bit easier to hold through volatile periods. But regardless, operating fundamentals doesn’t always align with short term price movements. Same concept can be applied to other sectors.
-
因管理层明确长期目标及新产能指引，博主修正了对混合制造资本支出的看空观点。
@Sigurd_Drivenes 我很高兴管理层终于对其长期目标提供了清晰的说明。 鉴于新的产能目标，我改变了对他们混合制造模型资本支出（Capex）的负面看法。
英文原文
@Sigurd_Drivenes I’m glad management finally provided clarity on their long term goals. Changed my negative view about their hybrid manufacturing model capex given new capacity targets.
-
持有IREN，看好其与Anthropic合作，但担忧巨额资本支出。
@KnightCapitalYT 我的意思是，我现在确实持有 1 股 $IREN，所以我很希望看到他们与 Anthropic 达成交易。 有泄露的内部备忘录显示，Anthropic 正在寻求在澳大利亚获得更多资源，而这恰好是 IREN 大力投资的国家。 不过，在当前宏观环境下，250-300 亿美元的资本支出(Capex)数字总是让人望而生畏，尽管他们已通过预付款获得了大量融资。
英文原文
@KnightCapitalYT I mean I do have 1 single share of $IREN now, so I'd love to see them get a deal with Anthropic. There were leaked memos suggesting that Anthropic was looking to secure more in Australia, and that just so happens to be the country Iren was heavily investing in. Buuut, the $25-30B capex number is always a bit daunting during current macro, despite securing a lot of financing via prepayments.
-
无视宏观波动，因AI资本支出强劲且需求失衡，坚持做多AI核心标的。
宏观环境显然在过去几个月削减了高贝塔AI的估值。但我试图专注于基本面。 如果你好奇我的看法： 我们看到了创纪录的长期协议(LTA)可见性： - 存储(至2031年)、光子学(至2029-2030年)等 - $NBIS、$SOL等，尽管资本支出巨大，但算力需求仍存在巨大失衡。 - 加上创纪录的预测（$NVDA 2027年增长70%，$SMCI新增服务器订单超600亿美元等）。 对比： 令人不安的宏观环境： - 沃什表示，如果核心通胀未明确回归2%，美联储“还有更多工作要做”。 - 7月PCE同比为3.7% - 7月就业报告新增就业减少2.3万，失业率4.1% - 10年期美债收益率高企/30年期创2007年以来新高 - 伊朗紧张局势持续 - 日元跌至160/韩国两次加息 且近期加息25个基点的概率为40/60。 这肯定会对融资产生影响，且第三/四阶效应太多，我个人无法完全映射。 我不是宏观专家，但我的看法是： AI建设将继续，预计明年上游投入超1.3万亿美元资本支出，流经许多瓶颈和卡点。 虽然市场削减了远期估值并转向当前盈利/防御性公司。 我不知道会发生什么，尤其是加息概率几乎五五开，且特朗普的帖子可能一夜之间改变情绪。 只要我没看到超大规模云厂商削减资本支出或需求波动的迹象。 尽管宏观波动，我保持做多，因为我想持有控制AI建设未来格局的公司。 如果宏观环境有任何好转迹象，高贝塔股通常恢复最快。
英文原文
Macro has clearly cut high beta AI valuations over the past few months. But I try to stay focused on fundamentals. If you're curious about my take though: We're seeing record visibility with - memory (2031), photonics (2029-2030), etc LTAs - $NBIS, solana:5aQcoaSyy7fnhvpRHasBUAMH2nL5tTjVrmFw6BeGmoon, etc. massive demand imbalance for compute despite capex spend. - combined with record projections ( $NVDA 70% growth 2027, $SMCI $60B+ new server orders, etc). vs. uncomfortable macro where: - Warsh said the Fed has "more work to do" if underlying inflation isn't clearly returning to 2%. - July PCE came in at 3.7% Y/Y - July jobs report payrolls fell by 23k with unemployment at 4.1% -10Y Treasury high/ 30Y highest since 2007 - Iran tensions continuing - Japan Yen weakened to 160/ 2 rate hikes in Korea And with rate 40/60 in favor of 25BPS near-term hikes. There's definitely implications on financing + too many third/fourth order effects for me to personally map. I'm not a macro expert but my take is: The AI buildout will continue w/ $1.3T+ capex poured upstream expected next year, flowing through many bottlenecks and chokepoints. While markets chopped off forward valuations and rotated toward currently profitable/defensive companies. I don't know what's going to happen, especially since rate hike odds are almost a coinflip, and Trump posts can change sentiment overnight. As long as I don't see indicators of hyperscalers cutting capex or demand wavering. I'm staying long despite macro volatility, since I want ownership of the names controlling the future of the AI buildout. And if there's any indicators of better macro climates, high beta are typically the fastest to recover.
-
NVDA指引超预期，AI供应链瓶颈环节将迎来爆发。
提醒AI看空者：$NVDA 预测2027年超大规模云厂商支出将达1.3万亿美元。较2026年约8000亿美元大幅增长，尽管产能仍受限，营收增速仍达约70%，远超华尔街此前约45%的预期。这些资本支出流向AI供应链……从存储、网络到晶圆代工厂。尤其是那些未从旧世代中受益的新瓶颈/卡点环节。我们就像派对刚开始2小时，大家起初喝的是札幌生啤。有人往情境(Situational)的酒里加了烈酒，我们不得不暂停派对30分钟。但现在派对重启，桌上摆满了龙舌兰酒。我认为真正的狂欢即将开始。
英文原文
FYI to the AI bears: $NVDA projects $1.3T of hyperscaler spend for 2027. Up from ~$800B in 2026, alongside ~70% revenue growth despite remaining capacity constrained… far above prior Street expectations of ~45%. That capex flows across the AI supply chain… from memory and networking to foundries. Especially through the new bottlenecks/chokepoints that didn’t benefit from older generations. We’re like 2 hours into the party where people started off drinking Sapporo draft beer. Someone spiked Situational’s drinks and we had to pause the party for 30 min. But now it’s back on with tequila shots on the table. I think the real party is about to begin.
-
SIVE遭恶意做空，长期Capex增加是主要看空点。
@sokarsokk 混合了算法、空头以及数十个被重新启用或休眠的账户，专门用于散布关于 $SIVE 的负面消息。很难相信那些在 2016 年亚洲 ICO 期间用于营销的账户，或是新的 X 账户，会突然开始只发布 50 条针对 Sivers 的负面评论。尽管这些行为会导致短期波动，但基本面终将回归。我看到的实质性看空因素是，从 Fab-lite（轻晶圆厂）模式转向混合制造模式所带来的资本支出（Capex）增加的影响（但从长远来看这是不可避免的）。我还没看到多少人评论这一点。
英文原文
@sokarsokk Mix of algorithms, short sellers, dozens of repurposed/dormant accounts dedicated toward spreading negative news about $SIVE Hard to believe accounts used for marketing during 2016 Asian ICOs or new X accounts would just start posting 50 negative comments only about Sivers. Regardless of near term volatility these things cause, fundamentals will catch up. The materially bearish thing I see is implications of increased capex spend from going to a hybrid manufacturing model from fablite (but it’s inevitable in the long run) I haven’t really seen much people comment on that
-
SIVE获新晶圆产能及6家新客户，解决CW激光器瓶颈，博主看多。
$SIVE 财报电话会纪要刚出，简而言之： - 现在有6家新的可插拔模块厂商正在与 Sivers 合作。 这可能是继新晶圆厂产能分配之外，本次财报最大的新闻。 “3家处于Alpha样品评估阶段”，另外3家处于技术对接/供应评估阶段。” 这也解释了为什么他们12亿美元的营收管道大幅膨胀（因为这次飞跃比 Jabil 更大）。 - 预计与 $JBL 的初始生产订单将在2027年上半年下达，量产爬坡计划于2027年下半年进行。 冲啊，Jabil 是巨大的超大规模客户供应商，终于有了明确的量产营收时间表。 - “新的晶圆厂合作伙伴带来了现已可用的巨大产能” 这正是我想听到的。显然他们为此筹备已久，因为产能“现已可用”，这非常关键。 连续波激光器(CW lasers)严重短缺，而 $SIVE 引入了 Win Semi 之外的新供应源（也降低了风险）。 正如你从其他合格的 CW 厂商那里听到的（他们生产的任何东西都能卖光），随着 Sivers 的合作伙伴完成资格认证，我预计情况会相同。 - “长期产能模型中，三分之一的制造产能将内部化，而三分之二来自我们的晶圆厂合作伙伴” 对此反应冷淡，虽然他们最终需要像 $AAOI / $LITE 那样垂直整合是有道理的，但这意味着更多的资本支出（最好在纳斯达克上市后再进行）。 正如你在 ESMT + DDR2 瓶颈中看到的，仅在短缺期间通过无晶圆厂模式确保晶圆所带来的营业利润是惊人的。 而且用未来的现金流来资助这一点会更合理。 - 纳斯达克上市正在进行中。CFO 给出了非常像律师的回答，但一如既往地在正轨上。 - “我们目前不认为生产能力是瓶颈” 这很有意义，因为随着新的晶圆厂合作伙伴加入，意味着 $SIVE 正在为一个瓶颈市场中提供非常关键的 CW 激光器供应。 简而言之： - 新的实质性晶圆厂产能分配！（在 CW 激光器短缺期间非常、非常关键） - 除了 $JBL 之外的6家新可插拔模块厂商 - 来自 Jabil 的明确营收爬坡时间表 对可插拔模块 -> 非平面光学(NPO) -> 共封装光学(CPO) 的重点关注似乎是演进路径。唯一反应冷淡的是建立内部产能，但长期来看是积极的。 总之，读完纪要后我非常高兴，除了可能优先投资激光器产能而非 IP 收购（这在行业短缺期间也是可以接受的）。 6家新的可插拔模块厂商 + “现已可用”的实质性晶圆分配，这是一个相当惊人的披露。
英文原文
$SIVE ER transcript just dropped, TLDR: - 6 NEW pluggable players working with Sivers now. Probably the biggest news of the entire ER next to new foundry allocations. "3 are in alpha sample evaluation stage" and another 3 in technical engagement / supply assessment stage." And now it makes more sense why their $1.2B revenue pipeline ballooned (since this was a bigger leap than Jabil). - Initial production orders with $JBL expected in the first half of 2027, with the production ramp planned second half of 2027 LFG, Jabil is a massive hyperscaler supplier and finally got clear timelines on revenue from volume ramp. - "new foundry partner who has brought on tremendous capacity that is available now" This is what I wanted to hear. Apparently they've been in the works for this for awhile since it's "available now", and THIS IS VERY MATERIAL. CW lasers are in a massive shortage and $SIVE brought on new supply outside of Win Semi (also de-risks). As you hear with other qualified CW players (anything they make gets sold), so as Sivers partners finishes their qualifications, I'm expecting the same. - "long-term capacity model where one-third of manufacturing capacity will be internal, while two-thirds will come from our foundry partners" Lukewarm on this, it makes sense they need to be vertically integrated like $AAOI / $LITE eventually but implies more capex (better after NASDAQ listing). As you see with ESMT + DDR2 bottlenecks, the operating income they get from just securing wafers during shortages and doing fabless models is incredible. And it would make more sense to fund this with cashflow down the road. - NASDAQ listing ongoing. CFO gave a very lawyer like answer, but on track as usual. - "We do not see production capacity as a bottleneck at this point in time" This is very meaningful since with their new foundry partners, implies $SIVE is coming online with a very material CW laser supply to a bottlenecked market. TLDR: - NEW substantial FOUNDRY ALLOCATION! (very, very material during CW laser shortage) - 6 new pluggable players outside of $JBL - Clear revenue ramp timelines from Jabil Heavy focus on pluggables -> NPO -> CPO seems like the progression. Only lukewarm piece was building up internal capacity but it's long term positive. Anyway, very happy after reading the transcript aside from potential capex prioritizing laser capacity (which is fine too during an industry shortage) over IP acquisition. The 6 new pluggable players + substantial wafer allocation "available now" is a pretty insane disclosure.
-
NVDA指引受供应限制，云厂商CapEx上调利好高贝塔供应链。
$NVDA：“我们预计2028财年营收将增长约70%。这是一个受供应限制的展望。” 仅这一句话的重要性可能就超过了刚刚发布的整个业绩超预期？ 疯狂的向上修正：从预期的+43.9%上调至70%。 “随着云行业积压订单现已超过2万亿美元，前五大超大规模云服务商的资本支出预计将在2026年达到近8000亿美元，2027年达到1.3万亿美元。” 2027年达1.3万亿美元。（微软此前6月约为1.2万亿，因此这是上调预测） 是的，供应链中高贝塔部分的增长将会非常喜欢这一点。
英文原文
$NVDA: "We expect to grow revenue by approximately 70% in fiscal 2028. This is a supply constrained outlook" That statement alone is probably more important than the entire earnings beat just now? Insane upward revision beat from +43.9% expected to 70%. "With cloud industry backlog now greater than 2 trillion, CapEx by the top five hyperscalers is expected to reach nearly $800 Billion in 2026 and 1.3 Trillion in 2027" $1.3 Trillion in 2027. (MS was around $1.2T June, so this is upward projection revisions) Yeah the high-beta parts of the supply chain growth is going to like this.
-
NBIS 业绩超预期，需求可见度极高，电力指引上调，看多。
$NBIS 的前景再次令人难以置信。 营收 5.823 亿美元（同比 +454%，环比 +46%），AI 云业务调整后息税折旧摊销前利润 (Adjusted EBITDA) 利润率约 50%。手头现金 80.4 亿美元。 重申财务指引（2026 年营收：30-34 亿美元；2026 年底年度经常性收入 ARR：70-90 亿美元），我原本希望会上调指引，但这已经是巨大的增长了。 重点不在于本季度财务数据，而是这些信号： - “我们今天就能以这些条款卖出 2027 年的全部产能。”需求可见度极高。 - “4 份客户协议，每份合同价值平均超过 10 亿美元。” - 约 70% 的 Q2 交易包含客户预付款。 - 签约电力指引再次上调，从约 4+ GW 升至 5 GW。（这曾是相对于 $IREN 辩论中的担忧点，但 Nebius 的电力指引持续飙升）。 - 预计 2026 年预付款超过 90 亿美元，并表示拥有超过 400 亿美元的承诺订单。 - 每兆瓦年度合同价值持续飙升（Q3 短期容量交易 >4000 万美元/兆瓦）。 看空故事依然存在（例如 56.6 亿美元的高资本支出和当前 GAAP 净亏损），但需求极度强劲且可见度极高，AI 云调整后息税折旧摊销前利润 (Adjusted EBITDA) 利润率持续提升（24% -> 45% -> 50% 的演进）。 我们昨天通过各大超大规模云厂商 (Hyperscaler) 的财报和 $CRWV 的积压订单已经大致看透了这一点，很高兴在 Weebius（应为 Nebius 笔误）的财报中看到印证。
英文原文
$NBIS outlook is pretty insane yet again. $582.3M revenue (+454% Y/Y, +46% Q/Q) off ~50% adjusted EBITDA margin (AI Cloud). $8.04B cash on hand. Reiterates financial guide (2026 revenue: $3.0–$3.4B 2026 EoY ARR: $7–$9B), I was hoping for a raise, but it's already massive growth. Not really about this quarter financials, but these signals: - "We could sell our entire 2027 capacity on these terms today." massive visibility on demand - "4 customer agreements averaging over $1B+ each in contract value" - Around 70% of Q2 deals included customer prepayments - Contracted power guidance was raised again, from more ~4+ GW to 5 GW. (this used to be a concern vs. $IREN debate, but Nebius power guidance keeps going through the roof). - Expects $9B+ of prepayments in 2026 and says it has $40B+ of commitments. - annual contract value per MW keeps going brrr (>$40M/mw for Q3 short term capacity deals) Same bear story will always be there (eg. large $5.66B of capex with current GAAP net loss), but demand is extreme with massive visibility, with increasing AI Cloud adjusted EBITDA margins ( 24% -> 45% -> 50% progression) We kinda got this read through from all your hyperscaler cloud earnings and $CRWV backlog yesterday, but glad it's showing up in Weebius's earnings.
-
四大云厂商上调Capex指引，看好上游半导体及AI供应链瓶颈环节重估。
作为业绩复盘：$AMZN、$META、$GOOGL 和 $MSFT 将 2026 年联合资本支出(Capex)指引上调至约 7200-7450 亿美元。此前指引为 6950-7250 亿美元。亚马逊：2200 亿；谷歌：1950-2050 亿；Meta：1300-1450 亿；微软：1750 亿。我们已看到显著的降杠杆和零售/机构保证金清算（可能还会持续，谁知道呢）。但中长期来看，我不太明白为何有人看空上游半导体或新云厂商。鉴于每家超大规模云服务商都指出了算力短缺、云需求上升+定价权增强，或增加芯片/网络支出。我对许多上游半导体供应链的“瓶颈”论点是：当数万亿美元资本流入磷化铟(InP)衬底或存储（此前均被视为廉价大宗商品）甚至能源领域时。许多此前在电信周期中被视为无用甚至“卖马桶”的当前 AI 概念股。当拐点到来且资本支出流入其资产负债表时，估值将重估。正如过去几年 $NVDA GPU、今年存储、未来几年 CPU/MLCC、2027 年共封装光学(CPO)、2027 年玻璃衬底、2027 年 800V 等所见。
英文原文
As a earnings recap: $AMZN, $META, $GOOGL, and $MSFT guided a 2026 combined capex forecast to ~$720-$745 Billion. Up from $695-$725B Billion previously. Amazon: $220B Google: $195B-$205B Meta: $130B-$145B Microsoft: $175B We've already seen significant deleveraging and retail/institutional margin liquidations (maybe it continues for more time, who knows). But medium-long term, I'm not quite sure how anyone can be bearish the upstream semis or neoclouds. Given each hyperscaler has flagged either compute shortages, rising cloud demand + pricing power, or increased spending for chips/networking. My "bottleneck" thesis with many of these upstream semi supply chains is that when trillions in capital flows into things from InP substrates or memory (which were both treated as cheap commodities) or even energy. Lot of these current AI names that were treated as useless before in telecom cycles or even toilet sellers. Gets rerated when their inflection period hits and capex flows into their balance sheets. As seen with $NVDA GPUs past few years, memory this year, CPUs/MLCCs next few years, CPO in 2027, Glass Substrates in 2027, 800V in 2027, and so on.
-
反驳媒体看空AI，指出巨头需求旺盛且韩国半导体大涨。
本周要点速览： 媒体：超大规模云厂商建设过剩，正在抛售多余算力。 -> $GOOGL, $MSFT, $META, 和 $AMZN：算力需求远超产能。资本支出(Capex)持续飙升。 机构：紧急加息或三次加息，行情就此终结。 -> Kevin Warsh：不会加息。 散户/媒体：AI因是泡沫而崩盘。 -> 城堡证券(Citadel)正寻求买入AI杠杆对冲基金头寸：👀 Jim Cramer：清仓所有杠杆型数据中心(DC)股票。 -> 韩国市场：SK海力士(SK Hynix) +30%，三星(Samsung) +26.81%
英文原文
Just a TLDR of this week: Media: Hyperscalers overbuilt and are selling excess compute. -> $GOOGL, $MSFT, $META, and $AMZN: Demand for compute far exceeds capacity. Capex go brrr. Institutions: Emergency rate hike or 3x rate hikes, it's all over. -> Kevin Warsh: No rate Hike. Retail/Media: AI is crashing because it's a bubble. -> Citadel looking to buy AI leveraged hedge funds positions: 👀 Jim Cramer: Sell all your leveraged DC stocks. -> Koreans: SK Hynix +30%, Samsung +26.81%
-
亚马逊资本支出上调及算力短缺证实，AXTI 产能扩张且供不应求，看多 AI 半导体及光通信供应链。
几条财报摘要，重点是我看好的 $AXTI 和 $AMZN： 亚马逊： - 将 2026 年资本支出上调至 2200 亿美元（此前为 2000 亿美元），部分原因是内存成本上升，这对 $MU 和三星是利好。 - 即使达到 2200 亿美元，亚马逊的产能仍无法满足 2026 年的所有需求；Jassy 预计 2027 年情况相同。 - 大多数 2027 年的新增产能已被预订，2028 年的大量产能也已预订。 主要浏览了上游半导体板块。$GOOGL、$META、$AMZN 和 $MSFT 均指出算力短缺。 几周前关于 Meta 等公司“算力过剩”以及超大规模云厂商削减支出的所有叙事……全是胡扯。 亚马逊财报对 AI 半导体交易非常看多。 AXTI： - AXT 计划在 2026 年将磷化铟(InP)产能翻倍，2027 年再翻倍。 - 预计这将使 AXT 成为“迄今为止全球最大的磷化铟生产商”——冲鸭。 - 第二季度营收为 4760 万美元，创 AXT 历史单季新高（来自数据中心应用的 InP 营收创纪录达 3070 万美元）。 - 营收环比增长 77%，同比增长 164%。 - InP 营收-产能目标：2026 年底每季度 6000 万美元。 - 2027 年底每季度 1.3 亿美元。 由于管理层表示他们将“寻找任何方式增加产能扩张”，1.3 亿美元的目标可能会进一步提高。 管理层称报告中的积压订单仍远高于 1 亿美元，但该数字已不能反映所有可用需求。 “无论我们增加产能的速度多快，客户需求持续超过供应。” 800G/1.6T 正在驱动当前周期，而 NPO/CPO 将其延伸至 2027 年之后。 此外，他们的目标是 50% 以上的毛利率：“我们绝对应该瞄准一个以五开头的数字。” 中国需求翻倍有余，且与 Casela、$COHR 和 $LITE 的协议并未实质性推动第二季度业绩。 这甚至还没有考虑我预测的随着 InP 衬底瓶颈加剧带来的巨大平均售价(ASP)上涨。 摘要： - 亚马逊算力需求过大，需要资本支出来满足，因此上游半导体起飞。 - AXT 是全球最大的 InP 衬底供应商，高毛利率，产能扩张，且供应跟不上需求。 看多来自亚马逊的需求端+资本支出。看多上游光通信供应链，因为需求过大。
英文原文
Few earnings TLDRs with my favorite $AXTI and $AMZN: Amazon: - Raised 2026 capex to $220B vs. prior $200B (partly due to higher memory costs, which is bullish on $MU to Samsung) - Even at $220B, Amazon will not have enough capacity to meet all 2026 demand; Jassy expects the same in 2027. - Most incoming 2027 capacity is already reserved, with substantial 2028 capacity also reserved. Mostly read through on upstream semis. $GOOGL, $META, $AMZN, and $MSFT all identified compute shortage. All the narratives a few weeks earlier was "excess compute" from Meta and others + hyperscalers cutting back on spend... All BS. Amazon earnings was very bullish on AI semi trade. AXTI: - AXT to double InP capacity during 2026. Then double again in 2027. This is expected to make AXT "by far the largest indium phosphide producer in the world." - LFG - Q2 rev was $47.6M, the highest quarterly revenue in AXT history (InP revenue reached a record $30.7M, from DC applications.) - Revenue increased 77% sequentially and 164% year over year. - InP revenue-capacity targets: $60M per quarter exiting 2026. - $130M per quarter exiting 2027 That $130m target could be hiked too since management stated they find "whatever ways to increase that capacity expansion" - Management said the reported backlog remains well above $100M, but that number no longer reflects all available demand. "Customer demand continues to outpace supply, no matter how fast we add capacity." 800G/1.6T is driving the current cycle, while NPO/CPO extends it beyond 2027. Also they're targeting 50%-plus gross margin: "We should definitely be targeting a number that begins with a five." China demand more than doubled, and their agreements with Casela, $COHR and $LITE did not materially drive Q2. This is not even considering my projected massive ASP hikes yet as InP substrates get more bottlenecked. TLDR: - Amazon too much compute demands, needs capex to fufill it, so upstream semis go brrr. - AXT world largest InP substrate supplier, high gross margins, expansion, and supply can't keep up with demand. Bullish on demand side from Amazon + capex. Bullish on upstream optical supply chains from too much demand.
-
超大规模云厂商资本支出强劲，计算资源稀缺，AI抛售被夸大。
以下是来自 $META 和 $MSFT 财报电话会议的一些要点： 微软： - 尽管资本支出增加，预计2027年将实现自由现金流(FCF)转正（这对由运营收入资助的AI建设极为有利） - “自由现金流为196亿美元，反映了更高的资本支出” - 季度资本支出为410亿美元，其中约2/3为“短期资产，主要是CPU和GPU” - 预计下一季度资本支出将超过500亿美元 - 资本支出指引约为1750亿美元，2027年资本支出大致相同。支出计划不变且符合预期。 - “将数据中心的预计使用寿命从15年延长至25年” - “我们将成为首批部署基于 $AMD Helios 和 $NVDA Vera Rubin 的下一代机架级AI基础设施的云提供商之一” - “客户需求持续超过可用容量” Meta： - 资本支出1300-1450亿美元（范围收窄），此前为1250-1450亿美元。 - Meta 收到的报价比其支付价格有“显著溢价”（计算资源稀缺，对 $IREN / $NBIS 等新云厂商有利） - 预计大部分计算资源（如埃尔帕索的1GW数据中心）将用于开发内部模型。 - Meta 在其核心业务中有多个投资回报率(ROI)为正的额外计算用途（内部使用，非Meta Compute） - “最后，我们认为整体行业产能将在可预见的未来保持紧张” - “历史上行业对AI采用浪潮的建设不足，使得现有产能，包括我们自己的，极具价值” - Susan Li 总结： - $MSFT 和 $GOOGL 主要在保持自由现金流为正或通过运营收入的同时，维持AI资本支出建设。 - $META 指出可用计算资源至少到2027年都远低于需求。$MSFT 也指出计算需求远超供应。 - 三大超大规模云服务商的资本支出基本与Google上调资本支出数字一致。 现在的AI抛售似乎被极度夸大了，超大规模云服务商继续按预期进行资本支出（微软保持自由现金流为正）甚至像 $GOOGL 那样上调。 计算资源稀缺在所有超大规模云服务商的财报电话会议中可见。
英文原文
Just some takeaways from $META | $MSFT earnings calls: Microsoft: - Expected to be FCF positive in 2027, despite the increase in capex (extremely positive for AI buildout that it's funded by operating income) - "Free cash flow was $19.6 billion, reflecting higher capital expenditures" - Quartely capex was $41B, roughly 2/3rds were "short lived assets, primarily CPUs and GPUs" - Expects capex spend will be over $50 billion for next quarter - Capex Guidance at ~$175 billion and 2027 capex roughly the same. Spending plans unchanged and in line. - "Extending the estimated useful life of our data centers from 15 to 25 years" - "We will be among the first cloud providers to deploy next generation rack-scale AI infrastructure based on $AMD Helios and $NVDA Vera Rubin" - "Customer demand continues to exceed available capacity" Meta: - Capex $130-$145 billion (narrowed range), from $125B-$145B. - Meta is receiving offers at a "significant premium" to what they paid for it (compute scarcity, positive for neoclouds like $IREN / $NBIS ) - Expects significant portion of compute (like the 1 GW DC in El Paso) to develop internal models. - Meta has multiple ROI-positive uses for additional compute across its core business (internally, not Meta Compute) - "Finally, we believe that overall industry capacity is going to remain tight for the foreseeable future" - "The industry has under-built historically for the wave of AI adoption, making existing capacity, including our own, extremely valuable" - Susan Li TLDR: - $MSFT and $GOOGL largely sustaining AI capex buildout while remaining FCF positive or through operating incomes. - $META flags available compute materially below demand at least through 2027. And $MSFT also flags compute demand far exceeds supply. - All three hyperscaler capex largely in line with Google hiking capex figures. AI selloff seems extremely overblown now, hyperscalers continuing capex in line (with Microsoft being FCF positive) or even hiked with $GOOGL. Compute scarcity is visible throughout every single hyperscaler ER.
-
博主认为AI需求加速信号明确，短期波动属去杠杆，看好光互连及存储板块。
我个人认为这是极短期的去杠杆行为，导致许多个股超跌。 - $BE 和 $TER 公布了超预期的业绩，Bloom 报告营收同比增长 166% + 利润率扩张 + 上调 2026 年指引... Teradyne 报告营收同比增长 104%，每股收益增长 300%+。 - 中国领先的光学玩家公布了惊人的初步财报，对西方光子学板块有极强的映射效应。 - 我认为财报季倾向于提醒市场 AI 加速仍在继续。我预计 $LITE、$SNDK / SK Hynix 以及板块龙头 ERs 将继续这一趋势。 - 但感觉市场更多是在奖励现有的加速，而非基于短期修正（如 $AMKR 所见）的未来一年增长。 - 总之，对我跟踪的主题来说看起来非常积极。 - 然后你有 $GOOGL 资本支出上调至 1950-2050 亿美元，这通常是超大规模云厂商最大的 AI 需求信号。 - 预测市场显示下次 FOMC 会议利率不变的概率为 77%。特朗普政府周一呼吁美联储降息。 - 所以今年三次加息的恐惧似乎有些夸大。 - 从 DUV 到 $CXMT 产能过剩的中国恐惧也被极度夸大。我们在每个行业都见过这种情况，直到人们记住激光或 HBM 瓶颈存在是有原因的。 - 就个人想法而言，看到所有东西一起涨跌是不健康的，尤其是这些主题。但我个人并不担心，因为我对我的持仓有充分信心，特别是像 $SIVE 或 $AAOI 这样的光互连领域。 - $META 计算和 $GOOGL GCP 利润率/增长显示出对 $NBIS 及类似新云业务模式的需求增加。 - $META 及其他超大规模云厂商关于与 $SNDK 和三星/SK Hynix、$MU 的长期协议（LTAs）的笔记指向结构性内存需求。 - 可以一直列举下去... - 但当 Jim Cramer 告诉每个保证金交易员在开盘时“全部卖出”，而新闻说存在 AI 泡沫时。 - 我讨厌别人说“它在跌所以卖出”。我相信重要的是看加速的营收/每股收益增长。如果营收加速与超大规模云厂商资本支出保持一致，那么该逻辑依然成立。 - 总之：我看到了 AI 需求/营收加速的迹象，超大规模云厂商资本支出是最重要的跟踪指标之一。 - 我猜我们会看到该主题广泛复苏。但我无法预测具体发生在哪一小时、哪一天或哪一周。
英文原文
I personally see it as extreme short term deleveraging that overshot many individual names. - $BE to $TER reported blowout results, with Bloom reporting 166% Y/Y rev growth + expanding margins + raised 2026 guidance... Teradyne reported 104% Y/Y revenue growth with 300%+ EPS growth. Your leading optical players over in China reported amazing preliminary earnings + extremely strong read through for the Western photonics sector. Think earnings season tend to remind markets about continued AI acceleration. I'm expecting $LITE, $SNDK / SK Hynix, and sector leader ERs to continue that trend. But it does feel like markets are rewarding existing acceleration more rather than future growth a year out over near-term revisions as seen with $AMKR. But as a TLDR, looks very positive for the themes I'm tracking. - Then you have $GOOGL capex raised to $195-$205B, which is typically the biggest AI demand signal from hyperscalers. - 77% 0bps change est. next FOMC decision from prediction markets. Trump administration on Monday called for the Fed to lower interest rates. So your 3x rate hikes fears this year seem kinda overblown. - Chinese fears from DUV to $CXMT overflooding are extremely overblown as well. We see this with every sector from time to time until people remember that lasers or HBM bottlenecks for a reason. As for personal thoughts, it was unhealthy seeing everything rise up or down together, especially with these themes. But I'm personally not worried since I have full conviction in my names, especially optical interconnects like $SIVE or $AAOI. $META compute and $GOOGL GCP margins/growth shows increasing demand for $NBIS and similar neocloud business models. $META + other hyperscaler notes around LTAs with $SNDK and Samsung/SK Hynix, $MU point to structural memory demand. Can go on and on... But when you have Jim Cramer telling every margined DC trader to "sell everything" at market open and news saying there's an AI bubble. I do hate it when others say "it's dropping so sell". My belief is that the important thing to look is accelerating revenue/EPS growth. And if that thesis stays in-tact with revenue acceleration in line with hyperscaler capex. TLDR: I see signs of AI demand / revenue acceleration, hyperscaler capex is one of the most important things to track. And my guess is that we'll see the theme recover broadly. But I can't predict what hour, day, or week that happens.
-
GOOGL上调资本支出指引后股价大跌，博主猜测是去杠杆所致。
@wannabepanacea 我最近刚退出（市场），正在从两轮手术中恢复。鉴于 $GOOGL 上周上调了资本支出指引，这一跌幅至少可以说是令人惊讶。我需要去查一下加息概率，但我猜测这只是去杠杆。
英文原文
@wannabepanacea I've been out if it recently, just recovering from 2 rounds of surgery. Drop was been surprising to say the least given $GOOGL hiked capex guidance last week. Will need to check on rate hike odds, but my guess is just deleveraging.
-
英伟达拟担保OpenAI融资，博主质疑循环融资健康度但看好资本支出受益者。
英伟达正在洽谈为 OpenAI 的数据中心融资提供高达 2500 亿美元的担保。 - 软银旗下的 SB Energy 开发俄亥俄州 10GW 园区 - OpenAI 签署长期租赁协议 - $NVDA 担保 2500 亿美元融资 - OpenAI 正在讨论单独协议，购买高达 3500 亿美元的英伟达芯片。 不太确定这种循环融资长期来看是否健康，但鉴于 AI 基础设施支出的巨大规模…… 感觉资本支出受益者会非常高兴。
英文原文
Nvidia in talks to backstop $250B worth of OpenAI DC financing. - Softbank’s SB energy develops 10 GW Ohio campus - OpenAI signs long term lease - $NVDA guarantees $250B in financing - OpenAI discussing separate agreement to buy up to $350B in Nvidia chips. Not quite sure if this circular financing is healthy long term, but given the massive size of AI infrastructure spending… Feels like capex beneficiaries will be extremely happy.
-
AI资本支出利好上游半导体，但市场担忧谷歌自身支出。
@haranguetan_ 在评论之前，具备一点点阅读理解能力会有所帮助。 AI资本支出(Capex) -> 流向内存、光子学、晶圆代工厂等上游半导体领域。这些参与者乐在其中。 $GOOGL 本身正在花钱，所以市场可能不喜欢他们（指谷歌）这一点。 但对于接收这些资金的参与者来说，这是积极的。
英文原文
@haranguetan_ Having an iota of reading comphrension helps a bit before commenting. AI Capex -> goes to upstream semis from memory, photonics, foundries, and others. These players are happy. $GOOGL itself is spending the money, so markets might not like that for them in specific. But it's positive to the players that receive that capital.
-
博主强调相关公司上调资本支出是核心论点。
@200pipss 他们刚刚确实上调了前瞻资本支出(capex)，这正是这条帖子的重点。
英文原文
@200pipss They literally just hiked forward capex, which was the point of the post.
-
超大规模云厂商上调AI资本支出，利好光网络板块财报表现。
@200pipss 超大规模云服务商上调人工智能资本支出指引，这一趋势尤其将在光网络（Optical Networking）板块的财报中体现出来。观察并学习……
英文原文
@200pipss Hyperscalers upping AI capex revisions is especially going to show up on optical networking earnings. Watch and learn...
-
GOOGL上调26财年资本支出指引，利好上游半导体供应链。
$GOOGL 更新 2026 财年资本支出(capex)指引： 上调至 1950 亿-2050 亿美元。此前为 1800 亿-1900 亿美元。 预计 Google 在 2027 年也将大幅增加资本支出。 这或许是超大规模云服务商对上游半导体最看多的解读。 https://t.co/dXiI0Ien8X
英文原文
$GOOGL updates FY 2026 capex guidance: To $195B-$205B. Up from $180B-$190B. Google is also expected to significantly increase capex in 2027. This is perhaps the most bullish read through on upstream semis from a hyperscaler. https://t.co/dXiI0Ien8X
-
GOOGL用户激增致算力紧缺，AI资本支出难放缓
$GOOGL 现在拥有 9.5 亿 Gemini 月活用户，每分钟处理 220 亿 API 令牌(API Token)。 相比之下，2 月份仅为 7.5 亿。 几个月内新增 2 亿用户，增长荒谬惊人。这解释了为何他们算力耗尽，进而削减了对 $META 等公司的配额分配。 鉴于此，AI 资本支出(AI Capex)不太可能放缓。
英文原文
$GOOGL now has 950M monthly Gemini users and processes 22B API tokens/min. Compared to 750M back in February. Absurd growth adding 200 million more users in a few months. Kinda explains why they ran out of compute, then cut allocations to $META and others. Very unlikely AI capex will slow down given.
-
GOOGL云业务超预期利好AI需求，支撑LITE等光网络板块。
$GOOGL 今日公布财报： 每股收益(EPS)：实际 $9.11 vs 预期 $2.91 - 散户对 EPS 大幅超预期的反应某种程度上只是会计噪音，因为 $SPCX、Anthropic 等可能是主要贡献者。 营收：实际 $1197 亿 vs 预期 $1169.8 亿 Google Cloud 营收：$247.7 亿，同比 +82%，vs 预期的 63-65% ($222-$226 亿) - 这是巨大的超预期，也是目前最重要的部分。 Cloud 营业利润：~$88.1 亿，暗示利润率 35.6%，vs Q1 的 32.9%。 资本支出(Capex)：$449.2 亿 vs ~$441.5 亿（基本符合预期）。 所以初步解读是 Google Cloud 增长加速且利润率扩张，这对 AI 需求确实是看多(bullish)信号。 他们此前 $180-$190 亿的资本支出指引已经非常庞大，只要我们在财报电话会 + 前瞻指引中看到围绕这些数字... 对于 $LITE 和其他光学/网络交易来说应该没问题了。（例如上次财报，他们说数据中心和网络将占资本支出的 ~40%）。 超大规模云厂商(Hyperscaler)的财报电话会记录可能是 AI 资本支出交易中最需要关注的。 那应该在 3 分钟后开始。
英文原文
$GOOGL reported earnings today: EPS: $9.11 actual vs $2.91 expected - retail reactions to EPS blowout is kinda just accounting noise, since $SPCX, Anthropic, and others were likely large contributors. Revenue: $119.7B actual vs. $116.98B expected Google Cloud Revenue: $24.77B +82% Y/Y, vs. 63-65% expected ($22.2-$22.6B) - this is a very large beat and most material part so far. Cloud Operating Income: ~$8.81B, implying 35.6% margin vs. 32.9% Q1. Capex: $44.92B vs ~$44.15B (basically in line). So initial read through is Google Cloud accelerating growth with expanding margins is genuinely bullish for AI demand. Their former $180-$190B capex guidance is already extremely large, and as long as we get around these numbers in the earnings call + forward projections... Should be good to go for $LITE and the other optics/networking trade. (eg. last earnings, they said DC and networking would be ~40% of capex spend). Hyperscaler earnings transcripts are probably the most important thing to pay attention to with the AI capex trade. And that should be in 3 minutes.
-
AI价值非线性质疑利润论，资本支出将持续高增
老实说，在 X 上随意发布新颖的信息发现才是做研究该有的样子。这似乎是斯梅尔（Smale）提出的“下个世纪数学问题”列表中的第 16 项？这在一定程度上表明，AI 产生的价值并不随 token 成本线性增长……看到这一点后，空头可能仍会质疑：“但 AI 应用后的利润究竟从何而来？”AI 与资本支出（Capex）的投入大概率将继续迅猛增长。
英文原文
TBH, casually posting novel information discovery to X is how research should be done. Apparently this was #16 on the Smale's list of Mathematical Problems for the Next Century? Kinda shows that the value generated from AI does not scale linearly with token costs... After seeing this, bears are still probably going to say: "B-but where's the profit coming from after AI usage?" AI + capex is likely going to keep going brrr.
-
AI需求旺盛致模型产能不足，预示资本支出将持续增长。
周日闻到更多人工智能资本支出(AI capex)的气息，这种感觉无与伦比。如果领先的尖端模型(frontier models)因需求过旺而产能不足，可能还有很长的路要走。 https://t.co/d8mnuAQdLz
英文原文
There’s nothing quite like the scent of more AI capex on a Sunday. Probably some ways to go if leading frontier models are out of capacity from too much demand. https://t.co/d8mnuAQdLz
-
高盛上调中际旭创目标价及盈利预测，看好硅光子及光模块产业链前景。
谢谢，喜欢读评论！ 高盛将中际旭创(Innolight)的目标价上调至2581元人民币。较当前估值约有163.6%的上涨空间。 但最大亮点是其2026-2028年的盈利预测大幅上调，分别增长65%/108%/119%，基于： - 硅光子模块出货量大幅增加 - 扩展、扩容、跨领域出货量增长 - 1.6T/3.2T提升混合平均售价(ASP)和利润率 - AI资本支出增加 这通常对光通信板块有重大传导效应，因为： 我倾向于将中际旭创视为类似台积电($TSM)的半导体资本支出风向标，用以观察光子学领域的状况。 例如，硅光子渗透率提高意味着连续波激光器(如$SIVE)和SOI晶圆($SOI)的总可寻址市场(TAM)扩大。未来代际产品的ASP上涨对其他光模块厂商也是利好。 我认为报告之外的另一个要点是，中际旭创在12日的电话会议中表示，800G需求的增长比预期更显著。 这反过来预示着$AAOI和$LITE等公司在下一次财报季将面临更多需求。 简而言之：高盛在大幅回调期间给出了高盈利预测。 从根本上说，当财报季(ER)到来时，更广泛的光子学生态系统应该会感到高兴。
英文原文
Thanks, love reading the comments! Goldman Sachs raised Innolight PT to 2581 RMB. Roughly 163.6% upside from current valuations. But the largest thing is its 2026–2028 earnings est revision raised by a whopping 65%/108%/119%, based on: - Much higher silicon photonics module volumes - scale out, scale up, scale across volumes - 1.6T/3.2T lifting blended ASP + margins - increase in AI capex This is typically very material read through on the optical sector since: I tend to think of Innolight as a $TSM (semi capex) type read on how the photonics landscape is doing. Eg. Higher silicon photonics penetration means more TAM for cw lasers like $SIVE (cw) / SOI wafer demand for $SOI. ASP hikes for future gen is positive for the other optical markers too. Think my other takeaway outside the report was Innolight stating 800g demand was growing more significantly than expected from their transcript on the 12th. Which in turn signals more demand for names like $AAOI to $LITE next earnings. TLDR: GS gives high earnings projections during a time of massive corrections. Fundamentally, broader photonics ecosystem should be happy when it’s ER time.
-
推测苹果为求自力更生将增加AI资本支出，超出市场预期。
@SVTrivo 关于 $AAPL 正在采购哪些公司，目前尚无泄露消息。无论如何，我猜测在经历了 $GOOGL 和 $META 与 OpenAI 的事件后，他们正试图变得更加自力更生。这将预示着人工智能资本支出(AI Capex)的增加，而市场绝对没有预期到像苹果这样的公司会有此举。
英文原文
@SVTrivo No leaks yet on what companies $AAPL is buying. Regardless, my guess is that they're trying to become more self-reliant after what happened with $GOOGL + $META and OpenAI. Which would signal AI capex hikes, which markets definitely aren't expecting from someone like Apple.
-
苹果或并购AI芯片公司并大幅上调资本支出，利好光网络及代工。
$AAPL 看起来打算收购 AI 芯片公司来运行 AI（来源：The Information）。目前，$MSFT、$META、Amazon、Google 正在承担 AI 资本支出。但如果苹果在并购后加入其他公司呢？一种可能的场景是，他们大幅上调资本支出以建设自己的 AI 基础设施。因为他们可能目睹了 Google 因算力限制切断 Meta 的情况……或者当你与 OpenAI 合作开发大语言模型时会发生什么。然后他们意识到拥有自己的基础设施有多么重要。从光学网络到晶圆代工厂，这一场景在主题上会相当看多，而且是市场意想不到的？让我们看看会发生什么。
英文原文
$AAPL looks to acquire AI chip companies for running AI (Source: The Information) Right now, $MSFT, $META, Amazon, Google are carrying AI capex spend. But what if Apple joined the others after M&A? A possible scenario is that they revise capex largely upward for their own AI buildout. Since they probably witnessed Google cutting off Meta from compute constraints... or what happens when you partner with OpenAI for LLMs. Then learned how important it is to have your own infrastructure. This scenario would be quite bullish thematically from optical networking to foundries and something markets would not expect? We'll see what happens.
-
Meta 斥资百亿建数据中心扩 AI 算力，反驳媒体看空论调。
$META 将在加拿大建设一座耗资 100 亿美元的新数据中心，以扩大 AI 算力。媒体将“Meta 计算”框架化为“过度建设”和“削减资本支出”的说法，现在看来真是大错特错。
英文原文
$META to build a new $10B DC in Canada to expand AI capacity. So much for the media framing of “Meta Compute”as overbuilding and cutting capex. https://t.co/0QgSls0s9o