AI算力竞赛远未结束：从 𝑁 𝑉 𝐷 𝐴 到 NVDA到SPCX的产业级重构 The AI Compute Race Isn't Over: From NVDA to SPCX, an Industry-Wide Reconstruction
你好，希望你今天一切都好。 Hello, hope you're having a great day.
近期市场波动较大，团队分享操作机会的频率相对减少。在此想和各位会员说明：团队一直关注市场趋势变化，在方向不明朗、波动剧烈的阶段，团队选择耐心等待趋势真正明确后才进行部署，不会为了发布内容而频繁操作。
Recently, the market has been particularly volatile, and the team has shared trade opportunities less frequently as a result. We want to explain to all members: the team has been continuously monitoring market trend changes, and during periods of unclear direction and high volatility, the team chooses to wait patiently until a trend becomes genuinely clear before deploying — not trading frequently simply for the sake of publishing content.
真正有效的机会，往往需要等待确定性足够高的时刻才会出现。盲目追逐每一次波动，反而容易增加不必要的风险。有时候，不操作本身就是最好的操作。
Truly effective opportunities tend to emerge only when there is sufficient certainty. Blindly chasing every fluctuation tends to add unnecessary risk instead. Sometimes, not acting is itself the best action.
团队的每一次分享，都是建立在充分评估市场环境之后才做出的决定，目的是为了保护大家的资金，而不是单纯为了满足更新频率。感谢大家的信任与耐心。
Every recommendation from the team is a decision made only after thoroughly assessing market conditions — the goal is to protect your capital, not simply to satisfy a certain update frequency. Thank you for your trust and patience.
今天团队分享了Rocket Lab ·后市策略请会员平台查看详情
Today the team shared a follow-up strategy on Rocket Lab — please check the member platform for details.
至于市场方面，今天早上最受关注的新闻是美联储主席凯文·沃什在杰克逊霍尔会议上的讲话。如果你想看热闹，那么关注黄仁勋、苏姿丰、亚历克斯·卡普、埃隆·马斯克和杰米·戴蒙这些人——无论美联储下一步做什么，他们领导的公司都有盈利潜力——会更有娱乐性和价值。投资建议：别再试图预测美联储的下一步行动了。你预测不了，说实话，他们也预测不了。如果说有什么能提醒你保持专注，那就是现在了。世界上最优秀的CEO们始终专注于赚钱，无论发生什么。而且，这也是投资者应该思考的重点。
As for the broader market, the most closely watched news this morning was Fed Chair Kevin Warsh's remarks at the Jackson Hole conference. If you want something more entertaining and worthwhile to follow, keep an eye on figures like Jensen Huang, Lisa Su, Alex Karp, Elon Musk, and Jamie Dimon — whatever the Fed does next, the companies they lead still have earnings potential. Investment takeaway: stop trying to predict the Fed's next move. You can't predict it, and honestly, neither can they. If anything should remind you to stay focused, it's this moment. The world's best CEOs stay focused on making money no matter what happens. And that, arguably, is what investors should really be thinking about.
NVDA 财报再度证明：AI 算力需求远未见顶 NVDA Earnings Prove Again: AI Compute Demand Is Far From Peaking
NVDA 最新财报再次交出超预期成绩单，二季度营收达到 962.2 亿美元，数据中心营收更是同比翻倍至约 890 亿美元，均明显高于市场预期。
NVDA's latest earnings once again beat expectations, with second-quarter revenue reaching $96.22 billion and data center revenue roughly doubling year-over-year to approximately $89 billion, both well above market forecasts.
但真正让市场重新兴奋起来的，并不只是这一次财报超预期，而是管理层给出的中长期增长信号。
But what really re-energized the market wasn't just this one earnings beat — it was the medium-to-long-term growth signal management provided.
英伟达罕见地提前给出了下一财年的收入展望，预计全年营收增长约 70%，远高于市场此前约 40%–44% 的预期。这意味着市场最担心的一个问题——AI 基础设施投资是否已经接近周期顶部——目前仍没有得到财报数据的验证。
Nvidia rarely gives forward guidance this early, but it projected next fiscal year revenue growth of approximately 70%, far above the market's prior expectation of roughly 40%–44%. This means the market's biggest worry — whether AI infrastructure investment is nearing a cyclical peak — remains unconfirmed by the earnings data so far.
数据中心依旧是核心增长引擎，而 Vera Rubin 新平台已经开始向客户出货，预计当前季度就将贡献约五分之一的数据中心收入。
Data center remains the core growth engine, and the new Vera Rubin platform has already begun shipping to customers, expected to contribute roughly one-fifth of data center revenue this quarter.
与此同时，英伟达仍然受到供应约束，尤其是内存和其他关键零部件成本上涨，将阶段性压缩毛利率。公司预计第四季度毛利率可能降至约 71%–72%，这也是当前财报中少数需要警惕的变量。
At the same time, Nvidia is still facing supply constraints, and rising costs for memory and other key components will compress gross margin for a period. The company expects fourth-quarter gross margin could fall to roughly 71%–72%, one of the few variables in this report worth watching closely.
中国市场则依然存在不确定性。H200 已经出现少量对华出货，但目前规模仍非常有限，英伟达也没有把中国数据中心收入纳入最新展望，因此短期内中国业务并不会成为整体业绩增长的核心驱动力。
Uncertainty persists in the China market. Small amounts of H200 shipments to China have begun, but volumes remain very limited, and Nvidia has not included China data center revenue in its latest guidance — so China will not be a core driver of overall growth in the near term.
真正让资金重新转向乐观的，是全球 AI 算力依旧供不应求。从超大规模云厂商，到 AI 实验室、Neocloud、企业客户以及主权 AI 项目，需求正在进一步扩散。英伟达甚至预计下一财年 AI 实验室客户将贡献约四分之一的整体业务。
What's really turning capital bullish again is that global AI compute remains in short supply relative to demand. Demand is broadening further — from hyperscale cloud providers to AI labs, Neoclouds, enterprise customers, and sovereign AI projects. Nvidia even expects AI lab customers to contribute roughly a quarter of overall business next fiscal year.
所以这份财报释放出的最重要信号是：AI 需求没有见顶，真正的瓶颈依然是供给。只要算力需求继续增长，NVDA 就仍然处在整个 AI 基础设施周期最核心的位置。
So the most important signal from this earnings report is this: AI demand has not peaked — the real bottleneck remains supply. As long as compute demand keeps growing, NVDA remains at the very center of the entire AI infrastructure cycle.
短期需要关注的，是毛利率下行、供应链成本以及中国市场的不确定性；长期真正决定估值的，则是 Rubin 平台能否继续推动 AI 算力需求进入下一轮增长。
In the near term, the things to watch are gross margin decline, supply chain costs, and uncertainty around the China market. Over the long run, what truly determines valuation is whether the Rubin platform can keep pushing AI compute demand into its next growth wave.
AI 还没有结束。至少从 NVDA 的订单、收入和未来指引来看，这场算力竞赛仍在加速。
AI is not over. At least based on NVDA's orders, revenue, and forward guidance, the compute race is still accelerating.
黄仁勋说"人工智能已经到了转折点"，这一表态值得关注。他接着表示，需求不仅强劲，而且正在加速增长——在如此庞大的规模下，新的人工智能实验室、开放模型和物理人工智能正处于黄金时代，所有这些都在全速运转，此外，预计明年收入增长将达到约 70%，而供应仍然受到限制。
Jensen Huang said "AI has reached an inflection point" — a statement worth noting. He went on to say that demand is not only strong but accelerating — at this scale, new AI labs, open models, and physical AI are all in a golden era and running at full speed. He also noted expected revenue growth of roughly 70% next year, with supply still constrained.
NVDA：这已经不是普通的半导体增长 $NVDA: This Is No Longer Ordinary Semiconductor Growth
目前的 NVDA，每一个季度新增的收入，都已经超过许多传统半导体公司一整年的收入。更夸张的是，这种增长并没有明显放缓。
Today, NVDA's incremental revenue in a single quarter already exceeds the full-year revenue of many traditional semiconductor companies. What's more striking is that this growth shows no clear sign of slowing.
最新季度营收达到约 962 亿美元，同比增长 106%；数据中心业务单季收入约 890 亿美元，同比增长 117%。下一季度营收指引更是达到约 1080 亿美元。
Latest-quarter revenue reached approximately $96.2 billion, up 106% year-over-year; data center revenue alone was about $89 billion for the quarter, up 117% year-over-year. Next-quarter revenue guidance is even higher, at approximately $108 billion.
这意味着英伟达正在创造一种过去很少见的商业规模：一家公司每隔三个月，就能增加数百亿美元级别的收入，同时仍然保持接近三位数的同比增长。
This means Nvidia is creating a scale of business rarely seen before: a company adding tens of billions of dollars in revenue every three months, while still maintaining near-triple-digit year-over-year growth.
而这背后真正发生的变化，是 GPU 已经不再只是芯片。它正在成为整个 AI 数据中心的核心基础设施。
What's really changing behind this is that the GPU is no longer just a chip. It is becoming the core infrastructure of the entire AI data center.
从训练到推理，从 GPU 到 CPU，从 NVLink 到网络，再到 CUDA 软件生态，英伟达正在不断扩大自己在 AI 计算栈中的覆盖范围。
From training to inference, from GPU to CPU, from NVLink to networking, to the CUDA software ecosystem, Nvidia is continuously expanding its footprint across the AI compute stack.
更重要的是，管理层预计下一财年收入仍有望增长约 70%，远高于市场此前约 44% 的预期。这意味着市场最担心的"AI 算力需求正在见顶"，至少目前还没有出现在数据里。相反，客户仍然在疯狂增加算力。
More importantly, management expects next fiscal year revenue to still grow by roughly 70%, far above the market's prior expectation of about 44%. This means the market's biggest fear — that "AI compute demand is peaking" — has, at least for now, not shown up in the data. On the contrary, customers are still aggressively adding compute capacity.
AWS 已经宣布未来部署约 200 万颗英伟达 GPU；超大规模云厂商、AI 实验室、企业以及主权 AI 项目都在继续扩大投入。
AWS has already announced plans to deploy approximately 2 million Nvidia GPUs going forward; hyperscale cloud providers, AI labs, enterprises, and sovereign AI projects are all continuing to expand their investment.
所以现在回头看 $NVDA 的增长曲线，会发现真正令人震撼的地方并不是某一个季度的财报。而是：两年前市场认为不可持续的增长速度，如今依然在持续。
So looking back at $NVDA's growth curve now, the truly striking part isn't any single quarter's earnings. It's this: the growth rate the market considered unsustainable two years ago is still continuing today.
这已经不是简单的"芯片周期"。这是计算基础设施正在被 AI 重新定义。
This is no longer a simple "chip cycle." This is computing infrastructure being redefined by AI.
有人会问，这种增长还能持续多久？答案最终取决于 AI 推理需求、模型规模、Agentic AI、Physical AI 以及全球数据中心资本开支能否继续扩张。
Some may ask, how long can this growth continue? The answer ultimately depends on AI inference demand, model scale, Agentic AI, Physical AI, and whether global data center capital expenditure can keep expanding.
但至少到目前为止，数据给出的答案非常明确：AI 算力需求还远没有结束。
But at least so far, the data gives a very clear answer: AI compute demand is far from over.
市场正在见证的，可能不仅仅是一家半导体公司的超级周期。而是一场计算基础设施史上极其罕见的产业级重构。
What the market is witnessing may not simply be one semiconductor company's super-cycle. It may be an extremely rare, industry-wide restructuring in the history of computing infrastructure.
英伟达依然是一家实力强劲的公司，且仍被视为一项被严重低估的投资。
Nvidia remains a powerhouse company, and is still regarded as a significantly undervalued investment.
$SPCX 真正的估值重心，已经从火箭转向 AI $SPCX: The Real Center of Valuation Has Shifted From Rockets to AI
摩根士丹利维持 $SPCX 超配评级，并给出 300 美元目标价。更疯狂的是，机构模型预计 SpaceX 到 2040 年的年收入可能达到约 3.5 万亿美元。
Morgan Stanley maintains an overweight rating on $SPCX with a $300 price target. Even more striking, the firm's model projects SpaceX's annual revenue could reach approximately $3.5 trillion by 2040.
马斯克直接回应：这个数字不需要等到2040年，按他的估计，2033年左右就可能达到。整整提前7年。
Musk responded directly: that figure doesn't need to wait until 2040 — by his estimate, it could be reached around 2033. A full seven years earlier.
但真正值得关注的，并不是谁的预测更大胆，而是SpaceX 到底凭什么支撑这么大的收入规模？答案已经越来越清晰：底层是 Starship，中间是 Starlink，最上层是 AI。
But what really matters isn't whose forecast is bolder — it's what could actually support revenue on this scale for SpaceX. The answer is becoming increasingly clear: the foundation is Starship, the middle layer is Starlink, and the top layer is AI.
市场现在赌的已经不是 SpaceX 还能发射多少枚火箭，而是 Starship 能否真正把入轨成本、复用率和发射频率做到工业化水平。
The market is no longer betting on how many more rockets SpaceX can launch, but on whether Starship can truly bring launch cost, reusability, and launch frequency to an industrial scale.
如果 Starship 能够大规模复用，并把单位入轨成本持续压低，那么它就不再只是一个火箭项目，而会变成整个 SpaceX 商业模式的基础设施。
If Starship can be reused at large scale and keep driving down the cost per launch, it stops being just a rocket program and becomes the infrastructure underpinning SpaceX's entire business model.
这也是为什么路易斯安那州新 Starbase 如此重要。SpaceX 计划投入约 1000 亿美元建设新的大型发射基地，最终规划 5 个发射综合体、10 个发射台，并以每年数千次 Starship 发射能力为目标。建设预计 2027 年启动，首次发射目标在 2029 年左右。
This is also why the new Starbase in Louisiana matters so much. SpaceX plans to invest approximately $100 billion in a new large-scale launch site, ultimately planning 5 launch complexes and 10 launch pads, aiming for a capacity of thousands of Starship launches per year. Construction is expected to begin in 2027, with the first launch targeted around 2029.
这不是简单增加几个发射台。本质上是在为未来的高吞吐量太空经济提前建设基础设施。
This isn't simply about adding a few more launch pads. It is, in essence, building infrastructure ahead of time for a future high-throughput space economy.
因为只有发射成本足够低、频率足够高，Starlink 才能持续扩大卫星规模，未来轨道数据中心、AI 算力基础设施以及其他太空业务才真正具备商业化基础。
Because only when launch costs are low enough and frequency high enough can Starlink keep expanding its satellite constellation — and only then do future orbital data centers, AI compute infrastructure, and other space businesses have a real commercial foundation.
而现在最赚钱的，其实已经不是SpaceX的火箭业务。2026 年第二季度，SpaceX 营收约 78 亿美元，同比增长 92%。其中 Starlink 所在的连接业务贡献约 43 亿美元，同比增长 66%，已经成为公司的现金流核心。Starlink 用户数达到 1200 万，同比增长一倍，而 ARPU 约为 66 美元。
And right now, the most profitable part of SpaceX is actually no longer its rocket business. In the second quarter of 2026, SpaceX revenue was approximately $7.8 billion, up 92% year-over-year. Of that, the connectivity segment (Starlink) contributed approximately $4.3 billion, up 66% year-over-year, and has become the company's core cash flow driver. Starlink subscribers reached 12 million, doubling year-over-year, with ARPU of approximately $66.
这意味着SpaceX正在完成一个非常重要的结构变化：火箭负责降低成本，Starlink负责创造现金流，AI负责打开估值上限。而AI可能才是未来最大的变量。
This means SpaceX is undergoing a very important structural shift: rockets bring down costs, Starlink generates cash flow, and AI opens up the ceiling on valuation. And AI may well be the biggest variable going forward.
SpaceX 正在持续增加 AI 算力基础设施投入，从地面数据中心，到未来潜在的轨道算力，AI 业务正在从一个远期故事逐渐变成公司资本开支的重要组成部分。
SpaceX is continuing to increase its investment in AI compute infrastructure — from ground-based data centers to potential future orbital compute — and AI is gradually shifting from a distant story into a significant part of the company's capital expenditure.
这也解释了为什么摩根士丹利的$300目标价里，真正值得关注的并不是传统 Space 业务，而是 Starlink 和企业 AI 所贡献的估值空间。换句话说：未来的 $SPCX，可能不是一家"更大的火箭公司"，而是一家拥有全球卫星网络、全球发射能力以及超大规模 AI 基础设施的综合平台。
This also explains why, within Morgan Stanley's $300 price target, the part really worth watching isn't the traditional space business, but the valuation contribution from Starlink and enterprise AI. In other words: the future $SPCX may not be "a bigger rocket company," but an integrated platform combining a global satellite network, global launch capability, and hyperscale AI infrastructure.
但这里最大的风险同样非常清楚。SpaceX 正在用巨额资本开支换取未来增长，而收入和现金流暂时还没有完全跟上投资规模。第二季度资本开支已经达到约 184 亿美元，而经营现金流约 35 亿美元，资金缺口依然巨大。
But the biggest risk here is equally clear. SpaceX is trading massive capital expenditure for future growth, and revenue and cash flow have not yet fully caught up with the scale of investment. Second-quarter capex reached approximately $18.4 billion, while operating cash flow was about $3.5 billion — a substantial funding gap remains.
所以接下来真正需要盯住的，不是一个遥远的3.5万亿美元收入预测。而是四组越来越重要的数据：
So what really needs watching going forward isn't a distant $3.5 trillion revenue forecast. It's four increasingly important sets of data:
Starship：发射频率与复用率能否持续提升。 Starship: whether launch frequency and reusability can keep improving.
Starlink：企业、政府、航空和海事客户能否成为新的增长引擎。 Starlink: whether enterprise, government, aviation, and maritime customers can become a new growth engine.
AI：算力规模能否真正转化成收入，而不只是不断增加CapEx。 AI: whether compute scale can truly convert into revenue, rather than just continuously rising CapEx.
现金流：经营现金流能否逐渐追上巨额资本开支。 Cash flow: whether operating cash flow can gradually catch up with massive capital expenditure.
如果这四个变量同时兑现，那么SpaceX的估值框架可能会发生根本变化。因为那时候，市场定价的就不再只是火箭。而是一套从"入轨运输 → 全球连接 → AI 算力"的完整基础设施体系。
If all four of these variables come through together, SpaceX's valuation framework could fundamentally change. Because at that point, the market would no longer be pricing just rockets. It would be pricing a complete infrastructure system spanning "orbital transport → global connectivity → AI compute."
这才是 $SPCX 最大的长期想象空间。
This is $SPCX's greatest long-term potential.
团队正在密切关注 $SPCX 的买入机会。
The team is closely watching for a buy opportunity in $SPCX.
结论 Conclusion
市场中的成功并非来自偶尔的举动。它源于你持之以恒的努力。买最好的，忽略其余的——遵循基本原则，利润自然会水到渠成
Success in the market doesn't come from occasional action. It comes from consistent, sustained effort. Buy the best, ignore the rest — stick to the fundamentals, and the profits will follow naturally.
你一定可以做到——这一点毋庸置疑。和往常一样，祝愿大家度过美好的一天。
You can absolutely do this — there's no question about it. As always, wishing everyone a wonderful day.
感谢你成为Wealth Club的一员！ Thank you for being a member of Wealth Club!
祝你投资顺遂，生活丰盛。 Wishing you prosperous investments and an abundant life.
财富俱乐部 Wealth Club
