<p><strong>英伟达</strong>(NVDA) 可能是过去三年人工智能进步浪潮中最大的受益者。该公司的市值现已超过 4.3 万亿美元，成为全球最有价值的公司。这得益于其市场领先的图形处理单元（GPU）的持续需求，这些 GPU 对于大型语言模型的训练和推理至关重要。</p>
<p>英伟达的财务数据显示，这是一家仍在以惊人的速度增长销售和利润的庞大公司。但最近的一些发展让人对英伟达能否继续以如此快速的速度增长产生了疑问。这包括其最大客户之一与英伟达最强劲的竞争对手达成的最新交易。</p>
<p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MzI1NjIvZ2V0dHlpbWFnZXMtYWktYXJ0aWZpY2lhbC1pbnRlbGxpZ2VuY2UtY2lyY3VpdC1jaGlwLmpwZyZ3PTcwMA?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" width="700" height="393" original-src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MzI1NjIvZ2V0dHlpbWFnZXMtYWktYXJ0aWZpY2lhbC1pbnRlbGxpZ2VuY2UtY2lyY3VpdC1jaGlwLmpwZyZ3PTcwMA"/></p>
<p>图片来源：Getty Images。</p>
<h2>英伟达面临的主要风险</h2>
<p>虽然英伟达的销售因大型科技公司不惜一切代价保持在人工智能开发的前沿而激增，但这导致只有少数客户占据了其收入的绝大部分。英伟达透露，第二季度其收入的 39% 来自仅两个直接客户，85% 来自六个客户。</p>
<p>当然，能够投入必要资本建设数据中心的公司数量有限。许多这样的公司，如 <strong>微软</strong>，拥有数十个大型客户通过其云计算平台支付使用英伟达的芯片。因此，英伟达的客户集中风险并不是那么大。不过，这仍然是投资者需要考虑的一个重要风险因素。</p>
<p>这一点最近引起了关注，因为 <strong>OpenAI</strong> 作为使用英伟达芯片的最大公司之一， reportedly 正在与 <strong>博通</strong>(AVGO -2.69%) 合作开发定制的人工智能加速器芯片。该报告与博通披露其拥有一个新的合格定制人工智能芯片客户的消息相吻合，并且已经获得了 100 亿美元的订单。</p>
<p>如果 OpenAI 确实是博通的新客户，这将代表该公司的重大转变。作为生成式人工智能的领导者，OpenAI 正在推进星际之门项目，该项目计划在 2025 年至 2028 年间投资 5000 亿美元用于人工智能基础设施。其在公告中提到的技术合作伙伴名单中显著包括了英伟达，但没有博通。</p>
<h2>转变已经开始</h2>
<p>OpenAI 可能不是唯一一家寻求使用更多定制硅解决方案并减少对英伟达依赖的公司。博通首席执行官 Hock Tan 表示，该公司在另外三家定制人工智能芯片客户中持续获得市场份额，这些客户包括 <strong>Meta Platforms</strong>、<strong>Alphabet</strong> 和字节跳动。此外，微软 reportedly 正在计划大幅提升其下一代定制加速器的使用，这可能在 2027 年带来 100 亿到 120 亿美元的订单，Fubon Research 表示。</p>
<p>这一消息还跟随了 OpenAI 最近与 Alphabet 的 Google Cloud 签署的一项协议，使用其由博通设计的张量处理单元（TPU）。谷歌最近发布的研究估计，其基于文本的人工智能提示的平均能耗仅为 0.24 瓦时，称其为 “远低于许多公共估计”。它还指出，过去一年效率提高了 33 倍。研究人员给出的效率原因之一是谷歌的 TPU 芯片。考虑到 OpenAI 的规模，转向非英伟达架构可能会节省大量成本。</p>
<p>投资者面临的问题是：市场是否已经充分考虑了英伟达可能面临的市场份额损失以及其他芯片制造商的潜在收益？英伟达的股票目前交易约为未来收益预期的 38 倍，这仍高于自 2023 年初以来的平均价格。尽管如此，英伟达不太可能继续经历像去年和前年那样的盈利增长。</p>
<p>但这并不意味着博通现在就便宜。由于赢得了一份大合同，博通的股票已经上涨至超过 50 倍的收益。尽管定制人工智能芯片是收入和利润的重要驱动因素，但需要记住的是，目前其约一半的半导体销售与人工智能相关，并且它还有一个相对成熟的软件业务。随着时间的推移，人工智能支出将对博通的整体业绩产生更大影响，但目前，它的估值就像其所有收入都来自于与人工智能相关的支出。</p>
<p>因此，目前这两家芯片制造商的估值都显得昂贵。如果必须选择一家公司，尽管市场估值，博通的上行潜力似乎比英伟达更强。不过，仍然有许多其他人工智能股票的价格对投资者更具吸引力。</p>

<p>英伟达作为人工智能技术的领导者，面临潜在的增长挑战，因为其主要客户之一 OpenAI 据报道正在与竞争对手博通开发定制的人工智能芯片。这一变化可能会影响英伟达的收入，因为其收入在很大程度上依赖于少数关键客户。OpenAI 与谷歌云在张量处理单元上的合作进一步突显了竞争格局。尽管英伟达的股票估值为前瞻性收益的 38 倍，但博通的估值因新合同已飙升至超过 50 倍的收益。两家公司看起来都很昂贵，但博通可能提供更强的上涨潜力</p>

<p><strong>Nvidia</strong>(NVDA) has been perhaps the single biggest beneficiary of the boom in AI progress over the last three years. The company's market cap now tops $4.3 trillion, making it the most valuable company in the world. That's driven by continued demand for its market-leading graphics processing units (GPUs), which are essential for large language model training and inference.</p>
<div class="lb-trans"><p><strong>英伟达</strong>(NVDA) 可能是过去三年人工智能进步浪潮中最大的受益者。该公司的市值现已超过 4.3 万亿美元，成为全球最有价值的公司。这得益于其市场领先的图形处理单元（GPU）的持续需求，这些 GPU 对于大型语言模型的训练和推理至关重要。</p>
</div><p>Nvidia's financials show a massive company that's still growing sales and profits at a breakneck pace. But a couple of recent developments have put a question mark on how long Nvidia can continue to grow at a torrid pace. That includes a recent deal by one of its biggest customers with one of Nvidia's fiercest rivals.</p>
<div class="lb-trans"><p>英伟达的财务数据显示，这是一家仍在以惊人的速度增长销售和利润的庞大公司。但最近的一些发展让人对英伟达能否继续以如此快速的速度增长产生了疑问。这包括其最大客户之一与英伟达最强劲的竞争对手达成的最新交易。</p>
</div><p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MzI1NjIvZ2V0dHlpbWFnZXMtYWktYXJ0aWZpY2lhbC1pbnRlbGxpZ2VuY2UtY2lyY3VpdC1jaGlwLmpwZyZ3PTcwMA" alt="" original-src="https://g.foolcdn.com/image/?url=https%3A%2F%2Fg.foolcdn.com%2Feditorial%2Fimages%2F832562%2Fgettyimages-ai-artificial-intelligence-circuit-chip.jpg&amp;op=resize&amp;w=700"/></p>
<p>Image source: Getty Images.</p>
<div class="lb-trans"><p>图片来源：Getty Images。</p>
</div><h2>The major risk facing Nvidia</h2>
<div class="lb-trans"><h2>英伟达面临的主要风险</h2>
</div><p>While Nvidia has seen its sales boom as a result of big tech companies spending whatever it takes to stay at the forefront of AI development, that has resulted in just a handful of customers making up the vast majority of its revenue. Nvidia revealed that 39% of its revenue in the second quarter came from just two direct customers, and 85% came from six customers.</p>
<div class="lb-trans"><p>虽然英伟达的销售因大型科技公司不惜一切代价保持在人工智能开发的前沿而激增，但这导致只有少数客户占据了其收入的绝大部分。英伟达透露，第二季度其收入的 39% 来自仅两个直接客户，85% 来自六个客户。</p>
</div><p>Of course, there are only so many companies in a position to spend the capital necessary to build out data centers. Many of those companies, like <strong>Microsoft</strong>, have dozens of big customers paying for access to those Nvidia chips through their cloud computing platforms. As such, Nvidia's customer concentration risk isn't quite that big. Still, it's a significant risk factor investors need to consider.</p>
<div class="lb-trans"><p>当然，能够投入必要资本建设数据中心的公司数量有限。许多这样的公司，如 <strong>微软</strong>，拥有数十个大型客户通过其云计算平台支付使用英伟达的芯片。因此，英伟达的客户集中风险并不是那么大。不过，这仍然是投资者需要考虑的一个重要风险因素。</p>
</div><p>That came to the forefront recently because OpenAI, one of the biggest companies using Nvidia's chips, is reportedly working on a custom AI accelerator chip with <strong>Broadcom</strong>(AVGO -2.69%). The report coincides with Broadcom's disclosure that it has a new qualified custom AI chip customer, and it's already secured $10 billion of orders.</p>
<div class="lb-trans"><p>这一点最近引起了关注，因为 <strong>OpenAI</strong> 作为使用英伟达芯片的最大公司之一， reportedly 正在与 <strong>博通</strong>(AVGO -2.69%) 合作开发定制的人工智能加速器芯片。该报告与博通披露其拥有一个新的合格定制人工智能芯片客户的消息相吻合，并且已经获得了 100 亿美元的订单。</p>
</div><p>If OpenAI is indeed Broadcom's new customer, that represents a major shift for the company. The generative AI leader is at the forefront of The Stargate Project, which is investing $500 billion between 2025 and 2028 in AI infrastructure. Its list of technology partners in the announcement notably included Nvidia, but not Broadcom.</p>
<div class="lb-trans"><p>如果 OpenAI 确实是博通的新客户，这将代表该公司的重大转变。作为生成式人工智能的领导者，OpenAI 正在推进星际之门项目，该项目计划在 2025 年至 2028 年间投资 5000 亿美元用于人工智能基础设施。其在公告中提到的技术合作伙伴名单中显著包括了英伟达，但没有博通。</p>
</div><h2>The shift is already underway</h2>
<div class="lb-trans"><h2>转变已经开始</h2>
</div><p>OpenAI might not be the only company looking to use more custom silicon solutions and reduce its reliance on Nvidia. Broadcom CEO Hock Tan said it continues to gain share with its three other custom AI chip customers, which include <strong>Meta Platforms</strong>, <strong>Alphabet</strong>, and ByteDance. Additionally, Microsoft is reportedly planning a big step-up in usage for its next-generation custom accelerator, which could result in $10 billion to $12 billion in orders in 2027, according to Fubon Research.</p>
<div class="lb-trans"><p>OpenAI 可能不是唯一一家寻求使用更多定制硅解决方案并减少对英伟达依赖的公司。博通首席执行官 Hock Tan 表示，该公司在另外三家定制人工智能芯片客户中持续获得市场份额，这些客户包括 <strong>Meta Platforms</strong>、<strong>Alphabet</strong> 和字节跳动。此外，微软 reportedly 正在计划大幅提升其下一代定制加速器的使用，这可能在 2027 年带来 100 亿到 120 亿美元的订单，Fubon Research 表示。</p>
</div><p>The news also follows a recent deal that OpenAI signed with Alphabet's Google Cloud to use its Tensor Processing Units (TPU), also designed by Broadcom. Google recently published research estimating its average energy expenditure for a text-based AI prompt was just 0.24 watt-hours, which it described as &#34;substantially lower than many public estimates.&#34; It also noted a 33-times improvement in efficiency over the past year. Part of the reason the researchers gave for its efficiency is Google's TPU chip. Considering OpenAI's scale, it could stand to save a lot by moving off Nvidia's architecture.</p>
<div class="lb-trans"><p>这一消息还跟随了 OpenAI 最近与 Alphabet 的 Google Cloud 签署的一项协议，使用其由博通设计的张量处理单元（TPU）。谷歌最近发布的研究估计，其基于文本的人工智能提示的平均能耗仅为 0.24 瓦时，称其为 “远低于许多公共估计”。它还指出，过去一年效率提高了 33 倍。研究人员给出的效率原因之一是谷歌的 TPU 芯片。考虑到 OpenAI 的规模，转向非英伟达架构可能会节省大量成本。</p>
</div><p>The question for investors is this: Has the market fully priced in the potential share loss Nvidia faces and the potential gains at other chipmakers? Nvidia shares now trade at about 38 times forward earnings estimates, which is still above its average price since early 2023. That's despite the fact that Nvidia's unlikely to experience a continuation of earnings growth like it saw last year and the year before.</p>
<div class="lb-trans"><p>投资者面临的问题是：市场是否已经充分考虑了英伟达可能面临的市场份额损失以及其他芯片制造商的潜在收益？英伟达的股票目前交易约为未来收益预期的 38 倍，这仍高于自 2023 年初以来的平均价格。尽管如此，英伟达不太可能继续经历像去年和前年那样的盈利增长。</p>
</div><p>But that doesn't mean Broadcom is any bargain right now, either. Shares have climbed above 50 times earnings on the news that it won a big new contract. And while custom AI chips are a big driver of revenue and profits, it's important to remember that only about half of its current semiconductor sales are related to AI, and it has a relatively mature software business as well. Over time, AI spending will have a greater influence on Broadcom's overall results, but right now, it's valued as if all of its revenue comes from AI-related spend.</p>
<div class="lb-trans"><p>但这并不意味着博通现在就便宜。由于赢得了一份大合同，博通的股票已经上涨至超过 50 倍的收益。尽管定制人工智能芯片是收入和利润的重要驱动因素，但需要记住的是，目前其约一半的半导体销售与人工智能相关，并且它还有一个相对成熟的软件业务。随着时间的推移，人工智能支出将对博通的整体业绩产生更大影响，但目前，它的估值就像其所有收入都来自于与人工智能相关的支出。</p>
</div><p>As such, both chipmakers look expensive right now. If you had to pick one, the upside looks stronger for Broadcom than Nvidia despite the market valuation. Still, there are plenty of other AI stocks with more compelling prices for investors to consider.</p>
<div class="lb-trans"><p>因此，目前这两家芯片制造商的估值都显得昂贵。如果必须选择一家公司，尽管市场估值，博通的上行潜力似乎比英伟达更强。不过，仍然有许多其他人工智能股票的价格对投资者更具吸引力。</p>
</div>

One of Nvidia's Biggest Customers Just Struck a Massive Deal With Its Fiercest Rival

- Nvidia 市值超过 4.3 万亿美元，成为全球最有价值公司。  
- 39% 收入来自两家客户，85% 来自六家客户，客户集中风险显著。  
- OpenAI 可能转向 Broadcom，影响 Nvidia 市场份额，投资者需关注潜在变化。  

Motley Fool

英伟达

博通

OpenAI

英伟达的一个最大客户刚刚与其最强劲的竞争对手达成了一项重大交易