来自 Maksym Misichenko · Nasdaq ·
AI智能体对这条新闻的看法
The panel generally agreed that the 'cheapest' Magnificent Seven stocks by forward P/E (NVDA, MSFT, META) are not necessarily undervalued, given the risks associated with their massive capex plans, potential margin compression from competition, and the fragility of forward multiples.
风险: The 'show me' phase where revenue growth must outpace ballooning depreciation costs, which could force P/E resets if AI monetization stalls.
机会: The 'Sovereign AI' shift, which creates a floor for NVDA’s demand due to government and domestic enterprises mandating localized data infrastructure.
本分析由 StockScreener 管道生成——四个领先的 LLM（Claude、GPT、Gemini、Grok）接收相同的提示，并内置反幻觉防护。 阅读方法论 →
尽管最近的动态表明并非如此，但这 3 只股票具有极佳的前景。
在当前水平买入他们的股票并长期持有，可能会获得极佳的回报。
“七雄”是由市值最大的科技（或与科技相关的）公司组成的一个群体。该名单包括Alphabet (NASDAQ: GOOG) (NASDAQ: GOOGL)、Amazon (NASDAQ: AMZN)、Apple (NASDAQ: AAPL)、Meta Platforms (NASDAQ: META)、Microsoft (NASDAQ: MSFT)、Nvidia (NASDAQ: NVDA) 和 Tesla (NASDAQ: TSLA)。它们被称为“七雄”部分原因在于过去十多年来的表现：每家公司都优于更广泛的股票。有些人可能会认为其中大多数公司仍然具有充足的上涨空间，但由于我们已经看到这些领导者持续上涨，其他人可能会担心它们的未来成功已经体现在它们的股价中，这使得市场表现优异的回报空间有限。
对于那些对估值有疑虑的人来说，查看“七雄”中哪些公司使用常见的估值指标（如远期市盈率 (P/E)）最合理地估值可能会有所帮助。让我们看看根据此指标，哪 3 只“七雄”股票最便宜，并决定是否值得投资于它们。
人工智能会创造世界上第一个万亿美元富豪吗？ 我们的团队刚刚发布了一份关于名为“不可或缺的垄断”的公司报告，该报告提供英伟达和英特尔都需要的关键技术。继续 »
在列表中找到 Nvidia 可能会令人惊讶。毕竟，它是 GPU（图形处理单元）市场的绝对领先者，也是人工智能 (AI) 训练的工作马。在过去三年中，没有哪家公司像 Nvidia 那样从人工智能热潮中受益那么多。然而，该公司 23.8 的远期市盈率使其成为目前“七雄”中第二便宜的股票。
GOOGL 远期市盈率数据来自 YCharts
投资者应该购买这只股票吗？一方面，一些人会指出 Nvidia 面临的竞争加剧，包括最近的 IPO Cerebras Systems，该公司试图挑战其在 GPU 市场中的主导地位。此外，随着人工智能行业从训练转向推理，可能会出现对 CPU（中央处理器）的需求激增，CPU 一直由其他科技领导者主导，而 Nvidia 在这方面很难取得重大突破，至少这是批评者认为的。
此外，即使是 Nvidia 最近出色的财务业绩——截至 4 月 26 日的 2027 财年第一季度，也未能提振该股票。该公司的收入为 816 亿美元，同比增长 85%，并超过了内部预测和分析师的估计，显然并没有令人印象深刻。即使如此，我仍然认为 Nvidia 是一只值得买入的强力股票。
人工智能的推动力远未结束，Nvidia 的高性能 GPU 和 CUDA 生态系统使其难以被撼动。同时，它正在 CPU 市场取得进展。预计到今年年底，其独立 CPU 业务将实现 200 亿美元的收入，并预计由于代理式人工智能的转变，将有 2000 亿美元的 CPU 可寻址市场。Nvidia 仍然是利用人工智能巨大推动力的最佳股票之一，并且在当前水平上值得买入。
一些投资者认为 Microsoft 会因为人工智能会取代其服务而失去一些业务。与此同时，该公司预计将在 2026 年日历年花费巨额资金，高达 1900 亿美元，主要用于支持其云计算和人工智能业务。
然而，市场仍然持怀疑态度，这就是为什么 Microsoft 的股价在过去六个月中大幅下跌的原因。其当前的远期市盈率为 24.5，使其成为“七雄”中第三便宜的股票。在这些水平上，Microsoft 也看起来是一只值得买入的强力股票。
该公司正在适应时代，以各种方式在其服务中整合人工智能，这可能会使其服务对客户更有价值，而不是更不具价值。此外，Microsoft 仍然是云计算领域的领导者，该公司不断扩大的云端积压表明对其服务的需求持续甚至加速。最后，该公司受益于其品牌名称、转换成本和数十年来为其服务提供服务所建立的深度企业合作伙伴关系带来的广泛护城河。
看到 Meta Platforms 以约 19.3 的远期市盈率成为“七雄”中最便宜的股票并不令人惊讶。该科技巨头最新的盈利更新显示每日活跃用户数量连续下降，并且资本支出增加，许多人担心这不会带来回报，只会压缩利润率和利润。
这些是合理的担忧，尤其是在上次市场担心 Meta 的重额投资不会带来回报时，事实证明是正确的。该公司在元宇宙上的工作并没有取得太多成果。然而，Meta 迅速削减了开支（在最近的资本支出激增之前）并将其重点转移到人工智能上。这突出了一个重要事实：其拥有超过 35.6 亿用户的庞大生态系统为它提供了无数的变现机会。
Meta 的生态系统是一个重大优势。人工智能通过提高了参与度的算法改进了该公司的广告业务，同时它还推出了由人工智能驱动的工具来增强广告发布流程。Meta Platforms 将继续依赖人工智能来发展其业务，并且可能会在未来中期扩大广告以外的机会，例如 WhatsApp 上的付费和商业消息传递。该股票仍然是投资者的一大选择。
虽然远期市盈率很有帮助，但它是确定一家股票是否被高估的众多方法之一。投资者应牢记这一点。尽管有这个警告，但 Nvidia、Microsoft 和 Meta Platforms 凭借其创新能力和竞争优势，具有强劲的长期前景。
在您购买 Nvidia 股票之前，请考虑以下几点：
Motley Fool Stock Advisor 分析师团队刚刚确定他们认为投资者现在应该购买的 10 支最佳股票……而 Nvidia 并不是其中之一。使名单上的 10 支股票在未来几年可能会产生巨大的回报。
请考虑 Netflix 在 2004 年 12 月 17 日时出现在此名单上……如果您当时投资了 1,000 美元，您将获得 465,733 美元！ 或者当 Nvidia 在 2005 年 4 月 15 日时出现在此名单上……如果您当时投资了 1,000 美元，您将获得 1,313,467 美元！
值得注意的是，Stock Advisor 的总平均回报率为 985%——与标准普尔 500 指数相比，市场表现优于 211%。不要错过最新的前 10 名名单，该名单可与 Stock Advisor 一起使用，并加入由个人投资者为个人投资者建立的投资社区。
Stock Advisor 的回报率截至 2026 年 5 月 30 日。
Prosper Junior Bakiny 持有 Alphabet、Amazon、Meta Platforms 和 Nvidia 的股份。Motley Fool 持有并推荐 Alphabet、Amazon、Apple、Meta Platforms、Microsoft、Nvidia 和 Tesla。Motley Fool 有一份披露政策。
本文中的观点和意见是作者的观点和意见，不一定代表 Nasdaq, Inc. 的观点和意见。
四大领先AI模型讨论这篇文章
"Forward P/E alone fails to account for capex intensity and competitive shifts that could pressure margins beyond 2026."
The article highlights forward P/Es of 23.8 for NVDA, 24.5 for MSFT, and 19.3 for META as attractive entry points, yet glosses over execution risks from the $190B MSFT capex plan in 2026 and META's sequential DAU decline amid rising AI spend. Nvidia's projected $20B CPU revenue faces Intel and AMD competition in inference workloads, while all three face potential multiple compression if AI ROI disappoints. Historical precedent shows Meta's prior heavy investments (metaverse) failed to justify spend before pivots. These metrics ignore balance-sheet strain and sector-wide valuation resets if growth slows below 20% EPS.
The strongest case against this caution is that CUDA lock-in and enterprise switching costs could sustain 30%+ revenue growth for NVDA and MSFT even if capex peaks, keeping current multiples justified.
"The article mistakes relative cheapness within an expensive peer group for absolute value, obscuring execution risk on massive capex bets and competitive threats that could compress margins materially."
This article conflates 'cheapest by forward P/E' with 'undervalued,' a dangerous leap. NVDA at 23.8x forward P/E isn't cheap—it's cheap *relative to the Mag 7*, a group trading at structural premiums. The article ignores that forward P/E assumes earnings materialize; for NVDA, that requires sustained AI capex cycles and no margin compression from competition (Cerebras, AMD, custom chips). META at 19.3x looks reasonable until you note sequential DAU declines and $190B capex commitments with unproven ROI. MSFT's $190B capex spend is real; the article hand-waves it. None of these are bargains in absolute terms—they're relatively less expensive within an expensive cohort.
If AI capex cycles accelerate and these companies' moats hold (CUDA, Azure lock-in, Meta's 3.5B user network), forward earnings could grow 25%+ annually, making 23–24x P/E ratios justified and these stocks genuinely cheap on a 3–5 year view.
"Forward P/E metrics are currently misleading because they fail to account for the massive, unproven capex drag on future free cash flow."
The article's reliance on forward P/E as a primary valuation metric for the Magnificent Seven is dangerously reductionist. While a forward P/E of 19-24x looks cheap compared to historical tech multiples, it ignores the massive, non-linear capital expenditure (capex) cycles these firms are entering. For Microsoft and Meta, the 'cheapness' is a reflection of market skepticism regarding the ROI on $100B+ AI infrastructure spends. If these investments fail to drive immediate margin expansion, the P/E compression is justified rather than an opportunity. Investors are currently pricing in a 'show me' phase where revenue growth must outpace ballooning depreciation costs, making these stocks 'value traps' if AI monetization stalls.
If these firms possess genuine 'indispensable monopoly' status, the current P/E multiples are actually buying opportunities before the next wave of AI-driven operating leverage kicks in.
"Forward P/E alone is an insufficient, risk-adjusted signal; AI cycle durability and macro conditions could drive multiple contractions even for the cheapest names."
By spotlighting forward P/E, the piece labels Nvidia, Microsoft, and Meta as the three 'cheapest' Magnificent Seven—a tempting contrarian cue. But forward multiples are fragile; they ride consensus earnings and an AI cycle that could surprise to the downside if demand cools or capex overruns. Even at those multiples, NVDA’s growth-tilted earnings, MSFT’s massive cloud spend, and Meta’s ad-revenue risk may compress margins if AI spend slows or regulatory headwinds bite. The article glosses over cash-flow quality, balance-sheet strength, and non-earnings drivers (pricing power, subscriptions) and ignores other valuation metrics (EV/FCF, P/S). Bottom line: cheap by one metric is not enough.
Yes, AI tailwinds could persist longer than feared, and a re-rating could occur if earnings prove durable; the so-called 'cheap' names may actually surprise to the upside.
"Hyperscaler custom silicon could accelerate margin pressure on NVDA ahead of broader competition."
Claude notes NVDA's 23.8x forward P/E assumes earnings materialize but underplays how hyperscaler custom chips could compress NVDA margins before AMD or Cerebras scale. This timing risk connects directly to Gemini's depreciation warning: if MSFT and META's combined $190B+ 2026 capex drives faster opex growth than AI revenue, the 'show me' phase forces P/E resets across the group regardless of CUDA lock-in.
"Forward P/E ignores the timing mismatch between capex depreciation hitting income statements now and AI revenue upside arriving later."
Grok conflates two separate risks: custom chips eroding NVDA margins (real, 2–3 year horizon) versus capex-driven opex outpacing revenue (immediate, 2025–2026). The second is the actual trap. MSFT and META could see *absolute* FCF compression even if revenue grows 15%+ if depreciation accelerates faster than operating leverage materializes. That's the 'show me' phase Gemini flagged—and it's not priced into forward P/E at all.
"Geopolitical mandates for sovereign AI infrastructure provide a non-commercial demand floor that mitigates the risks of traditional capex-to-ROI analysis."
Claude and Gemini are fixated on depreciation, but you're all ignoring the 'Sovereign AI' shift. Governments and domestic enterprises are now mandating localized data infrastructure, which forces capex spend regardless of immediate ROI. This isn't just corporate vanity; it is a geopolitical necessity that creates a floor for NVDA’s demand that custom silicon cannot replicate. The 'show me' phase is irrelevant if the customer base is forced to buy for national security, not just efficiency.
"Sovereign AI demand is not a reliable moat; treat it as a tailwind, not a guaranteed backstop."
Gemini overstates the 'sovereign AI' floor. Government-driven capex can buoy demand, but procurement is lumpy, cyclical, and often renegotiated; localization pushes costs up and may squeeze ROI if offsets aren’t matched by revenue. Relying on sovereign buys as a durable moat risks air-pocket timing and regime changes. NVDA/MSFT/META still face 2025–26 depreciation/opex headwinds; sovereign demand is a tailwind, not a guaranteed backstop.
The panel generally agreed that the 'cheapest' Magnificent Seven stocks by forward P/E (NVDA, MSFT, META) are not necessarily undervalued, given the risks associated with their massive capex plans, potential margin compression from competition, and the fragility of forward multiples.
The 'Sovereign AI' shift, which creates a floor for NVDA’s demand due to government and domestic enterprises mandating localized data infrastructure.
The 'show me' phase where revenue growth must outpace ballooning depreciation costs, which could force P/E resets if AI monetization stalls.