<p>在设计用于支持人工智能（AI）工作负载的半导体芯片时，有两家公司远远领先于其他公司：<strong>英伟达</strong>（Nvidia）（NVDA -4.36%）和 <strong>博通</strong>（Broadcom）（AVGO -2.53%）。英伟达在 AI 芯片领域是最大的早期赢家，但博通正在逐渐获得动力。</p>
<p>让我们看看在当前市场抛售期间，投资者应该关注哪只股票。</p>
<h2>两种不同的策略</h2>
<p>英伟达和博通在 AI 芯片市场采取了两种截然不同的策略。英伟达在图形处理单元（GPU）方面是明显的市场领导者，市场份额超过 80%。GPU 最初是为了加速视频游戏中的图形渲染而设计的，但由于其快速的处理能力，它们已成为支持 AI 工作负载的主要芯片。</p>
<p>英伟达巨大市场份额的秘密在于其 CUDA 软件平台。它是第一个允许开发者将 GPU 编程用于超出其原始用途的任务的软件平台，使其成为开发者学习编程 GPU 的标准。从那时起，该公司在 CUDA 之上构建了一系列库和工具，称为 CUDA X，帮助提高 AI 模型训练的效率和可扩展性，并改善推理时间。</p>
<p>另一方面，博通帮助其客户开发称为应用特定集成电路（ASIC）的定制 AI 芯片。ASIC 是为非常特定的任务设计的。因此，这些芯片的性能更好，功耗低于大众市场的 GPU。缺点是开发这些定制芯片需要相当长的时间，并且前期成本也很高。它们也没有 GPU 那种可以编程用于多种任务的灵活性。</p>
<p>毫不奇怪，在 AI 基础设施建设的初期阶段，英伟达的 GPU 是超大规模数据中心运营商（hyperscalers）用于训练 AI 模型和构建数据中心基础设施以满足强劲需求的最常用芯片。毕竟，大众市场的 GPU 更容易获得，而英伟达的 CUDA 软件平台使它们能够轻松编程以完成手头的任务。</p>
<p>然而，随着英伟达 GPU 价格因其主导市场地位而飙升，越来越多的公司开始在博通的帮助下设计自己的 AI 定制芯片。<strong>字母表</strong>（Alphabet）是该公司的第一个定制 AI 芯片客户，博通帮助其开发了用于 Google Cloud 业务的张量处理单元（TPU）。这些芯片专门设计用于优化 Google Cloud 的 TensorFlow 框架中的 AI 工作负载。字母表已将其归功于提高性能，并因其较低的功耗而降低了运营成本。</p>
<p>博通逐渐获得了更多的定制 AI 芯片客户。<strong>Meta Platforms</strong> 和 OpenAI 据报道是与博通合作开发 AI 芯片的下一家公司。博通预测，这三位客户在 2027 财年（截至 2027 年 10 月）将有价值在 600 亿美元到 900 亿美元之间的可服务市场机会。目前，博通的年度 AI 收入运行率刚刚超过 160 亿美元，显示出显著的增长潜力。</p>
<p>与此同时，该公司在最初的三位客户之外增加了额外的定制 AI 芯片客户。<strong>苹果</strong>（Apple）和抖音母公司字节跳动（ByteDance）被广泛认为是其开发进展最为顺利的下两位客户。字母表从设计到生产其芯片大约花费了 15 个月，这为这些客户成为重要收入贡献者提供了一个大致的时间框架。博通还暗示了另外两位定制 AI 芯片客户。</p>
<p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MTUwNjEvZ2V0dHlpbWFnZXMtMTM2NDg1OTcyMi5qcGcmdz03MDA?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MTUwNjEvZ2V0dHlpbWFnZXMtMTM2NDg1OTcyMi5qcGcmdz03MDA"/></p>
<p>图片来源：Getty Images。</p>
<h2>哪只股票更值得购买？</h2>
<p>从估值的角度来看，这两家公司在未来的市盈率（P/E）倍数上相似，英伟达为 21.5 倍，博通为 23 倍。虽然英伟达面前仍有强大的机会，但博通似乎在未来的增长前景更强。</p>
<p>部分原因是大数法则，因为英伟达在过去两年中已经将其收入增长了超过 380%。与此同时，博通在 AI 芯片领域也取得了强劲的进展。考虑到设计定制芯片所需的时间和前期成本，博通的客户将会在大型项目中使用这些芯片。</p>
<p>我认为这两家公司都将继续赢得其在 AI 芯片市场的合理份额，并且都将看到稳健的增长。然而，考虑到博通面前的巨大机会，我在当前水平上更偏好博通的股票。当然，在现实世界中，您应该能够长期持有这两者。</p>

英伟达

博通

<p>英伟达和博通是人工智能芯片市场的领先者，各自采取不同的策略。英伟达在 GPU 市场占据超过 80% 的市场份额，利用其 CUDA 软件平台来处理 AI 工作负载。相比之下，博通专注于为特定任务定制的人工智能芯片（ASIC），并在 Alphabet、Meta 和 OpenAI 等客户中获得了关注。虽然两家公司有类似的市盈率，但博通因其不断扩大的客户基础和显著的市场机会，被认为具有更强的增长前景。作者更看好博通的股票，认为其潜力巨大，尽管预计两家公司在人工智能芯片领域都将蓬勃发展</p>

<p>When it comes to designing semiconductor chips used to power artificial intelligence (AI) workloads, two companies stand head and shoulders above the rest: <strong>Nvidia</strong>(NVDA -4.36%) and <strong>Broadcom</strong>(AVGO -2.53%). Nvidia has been the biggest early winner in AI chips, but Broadcom has been gaining momentum.</p>
<div class="lb-trans"><p>在设计用于支持人工智能（AI）工作负载的半导体芯片时，有两家公司远远领先于其他公司：<strong>英伟达</strong>（Nvidia）（NVDA -4.36%）和 <strong>博通</strong>（Broadcom）（AVGO -2.53%）。英伟达在 AI 芯片领域是最大的早期赢家，但博通正在逐渐获得动力。</p>
</div><p>Let's see which stock investors should look to buy during this current market sell-off.</p>
<div class="lb-trans"><p>让我们看看在当前市场抛售期间，投资者应该关注哪只股票。</p>
</div><h2>Two different approaches</h2>
<div class="lb-trans"><h2>两种不同的策略</h2>
</div><p>Nvidia and Broadcom are taking two very different approaches to the AI chip market. Nvidia is the clear market leader in graphics processing units (GPUs), where it holds a more than 80% market share. GPUs were originally designed to speed up graphics rendering in video games, but due to their fast processing power, they have become the main chips used to power AI workloads.</p>
<div class="lb-trans"><p>英伟达和博通在 AI 芯片市场采取了两种截然不同的策略。英伟达在图形处理单元（GPU）方面是明显的市场领导者，市场份额超过 80%。GPU 最初是为了加速视频游戏中的图形渲染而设计的，但由于其快速的处理能力，它们已成为支持 AI 工作负载的主要芯片。</p>
</div><p>The secret sauce behind Nvidia's huge market share is its CUDA software platform. It was the first software platform to allow developers to program GPUs for tasks beyond their original purpose, making it the standard on which developers learned to program GPUs. Since then, the company has built a collection of libraries and tools on top of CUDA, called CUDA X, that help enhance the efficiency and scalability of training AI models and improve inference times.</p>
<div class="lb-trans"><p>英伟达巨大市场份额的秘密在于其 CUDA 软件平台。它是第一个允许开发者将 GPU 编程用于超出其原始用途的任务的软件平台，使其成为开发者学习编程 GPU 的标准。从那时起，该公司在 CUDA 之上构建了一系列库和工具，称为 CUDA X，帮助提高 AI 模型训练的效率和可扩展性，并改善推理时间。</p>
</div><p>Broadcom, on the other hand, helps its customers develop custom AI chips called application-specific integrated circuits (ASICs). ASICs are designed for very specific tasks. As such, these chips have better performance and consume less power than mass-market GPUs. The downside is that it takes considerable time to develop these custom chips and there are a lot of upfront costs, as well. They also don't provide that same type of flexibility as GPUs, which can be programmed for a variety of tasks.</p>
<div class="lb-trans"><p>另一方面，博通帮助其客户开发称为应用特定集成电路（ASIC）的定制 AI 芯片。ASIC 是为非常特定的任务设计的。因此，这些芯片的性能更好，功耗低于大众市场的 GPU。缺点是开发这些定制芯片需要相当长的时间，并且前期成本也很高。它们也没有 GPU 那种可以编程用于多种任务的灵活性。</p>
</div><p>Not surprisingly, in the initial surge of the AI infrastructure buildout, Nvidia's GPUs were the most common chips that hyperscalers (operators of large data centers) used to train AI models and build out data center infrastructure to meet strong demand. After all, mass-market GPUs were much more readily available and Nvidia's CUDA software platform made them easily programmable for the tasks at hand.</p>
<div class="lb-trans"><p>毫不奇怪，在 AI 基础设施建设的初期阶段，英伟达的 GPU 是超大规模数据中心运营商（hyperscalers）用于训练 AI 模型和构建数据中心基础设施以满足强劲需求的最常用芯片。毕竟，大众市场的 GPU 更容易获得，而英伟达的 CUDA 软件平台使它们能够轻松编程以完成手头的任务。</p>
</div><p>However, as Nvidia's GPU prices have skyrocketed due to its dominant market position, more and more companies have started to design their own AI custom chips with the help of Broadcom. <strong>Alphabet</strong> was the company's first custom AI chip customer, with Broadcom helping it develop its Tensor Processing Unit (TPU) for its Google Cloud business. These chips are designed specifically to optimize AI workloads within Google Cloud's TensorFlow framework. Alphabet has credited them for improving performance, as well as reducing its operating costs due to their lower power consumption.</p>
<div class="lb-trans"><p>然而，随着英伟达 GPU 价格因其主导市场地位而飙升，越来越多的公司开始在博通的帮助下设计自己的 AI 定制芯片。<strong>字母表</strong>（Alphabet）是该公司的第一个定制 AI 芯片客户，博通帮助其开发了用于 Google Cloud 业务的张量处理单元（TPU）。这些芯片专门设计用于优化 Google Cloud 的 TensorFlow 框架中的 AI 工作负载。字母表已将其归功于提高性能，并因其较低的功耗而降低了运营成本。</p>
</div><p>Broadcom has progressively acquired more custom AI chip customers. <strong>Meta Platforms</strong> and OpenAI are reportedly the next companies to collaborate with Broadcom in developing AI chips. Broadcom has forecast that these three customers will have a serviceable market opportunity valued between $60 billion and $90 billion for its fiscal year 2027 (ending in October 2027). Currently, Broadcom's annual AI revenue run rate stands at just over $16 billion, indicating a significant growth potential.</p>
<div class="lb-trans"><p>博通逐渐获得了更多的定制 AI 芯片客户。<strong>Meta Platforms</strong> 和 OpenAI 据报道是与博通合作开发 AI 芯片的下一家公司。博通预测，这三位客户在 2027 财年（截至 2027 年 10 月）将有价值在 600 亿美元到 900 亿美元之间的可服务市场机会。目前，博通的年度 AI 收入运行率刚刚超过 160 亿美元，显示出显著的增长潜力。</p>
</div><p>Meanwhile, the company has added additional custom AI chip customers beyond its initial three. <strong>Apple</strong> and TikTok owner ByteDance are widely believed to be its next two customers that are furthest along in their development. It took about 15 months for Alphabet to go from design to production for its chips, which gives a general time frame of how long it could take before these customers can become meaningful revenue contributors. Broadcom has also hinted at two additional custom AI chip customers.</p>
<div class="lb-trans"><p>与此同时，该公司在最初的三位客户之外增加了额外的定制 AI 芯片客户。<strong>苹果</strong>（Apple）和抖音母公司字节跳动（ByteDance）被广泛认为是其开发进展最为顺利的下两位客户。字母表从设计到生产其芯片大约花费了 15 个月，这为这些客户成为重要收入贡献者提供了一个大致的时间框架。博通还暗示了另外两位定制 AI 芯片客户。</p>
</div><p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MTUwNjEvZ2V0dHlpbWFnZXMtMTM2NDg1OTcyMi5qcGcmdz03MDA" alt="" original-src="https://g.foolcdn.com/image/?url=https%3A%2F%2Fg.foolcdn.com%2Feditorial%2Fimages%2F815061%2Fgettyimages-1364859722.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>Which is the better stock to buy?</h2>
<div class="lb-trans"><h2>哪只股票更值得购买？</h2>
</div><p>From a valuation perspective, the companies trade at similar forward price-to-earnings (P/E) multiples of 21.5 times for Nvidia and 23 times for Broadcom. While Nvidia still has a strong opportunity in front of it, Broadcom appears to have the stronger growth prospects going forward.</p>
<div class="lb-trans"><p>从估值的角度来看，这两家公司在未来的市盈率（P/E）倍数上相似，英伟达为 21.5 倍，博通为 23 倍。虽然英伟达面前仍有强大的机会，但博通似乎在未来的增长前景更强。</p>
</div><p>Part of this is due to the law of large numbers, as Nvidia has already grown its revenue by more than 380% over the past two years. Meanwhile, Broadcom has been making strong inroads into the AI chip space. Given the time and upfront costs involved with designing custom chips, Broadcom's customers will be using these chips for large projects.</p>
<div class="lb-trans"><p>部分原因是大数法则，因为英伟达在过去两年中已经将其收入增长了超过 380%。与此同时，博通在 AI 芯片领域也取得了强劲的进展。考虑到设计定制芯片所需的时间和前期成本，博通的客户将会在大型项目中使用这些芯片。</p>
</div><p>I think both companies should continue to win their fair share of the AI chip market and both will see solid growth. However, I prefer Broadcom's stock a bit more at current levels given the huge opportunity in front of it. Of course in the real world, you should be able to own both over the long run.</p>
<div class="lb-trans"><p>我认为这两家公司都将继续赢得其在 AI 芯片市场的合理份额，并且都将看到稳健的增长。然而，考虑到博通面前的巨大机会，我在当前水平上更偏好博通的股票。当然，在现实世界中，您应该能够长期持有这两者。</p>
</div>

Better AI Chip Stock: Nvidia vs. Broadcom

- Nvidia 和 Broadcom 在 AI 芯片市场中占据重要地位。  
- Nvidia 以 GPU 为主导，市场份额超过 80%。  
- Broadcom 则专注于定制 AI 芯片，预计到 2027 财年市场机会达 600 亿至 900 亿美元。

哪只 AI 芯片股票更好：英伟达与博通