<p><strong>英伟达</strong>(NVDA -2.62%) 最近预测，人工智能 (AI) 基础设施支出在未来几年可能会增长五倍。虽然它肯定会受益，但并不是唯一的受益者。</p>
<p>让我们看看三只将从这波支出激增中受益的 AI 股票，以及它们为何可能是买入机会。</p>
<p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MzI5NzEvZ2V0dHlpbWFnZXMtMTQ5MjMxOTg4NC5qcGcmdz03MDA?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-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MzI5NzEvZ2V0dHlpbWFnZXMtMTQ5MjMxOTg4NC5qcGcmdz03MDA"/></p>
<p>图片来源：Getty Images。</p>
<h2>1. 博通</h2>
<p><strong>博通</strong>(AVGO -3.93%) 已成为人工智能领域最重要的参与者之一。随着大型超大规模数据中心的拥有者寻求替代英伟达的方案以降低成本和多样化其供应链，越来越多的公司开始转向博通，帮助设计用于其 AI 工作负载的定制应用特定集成电路 (ASIC)。定制芯片通常比图形处理单元 (GPU) 更高效且更便宜，这在推理需求激增的情况下至关重要。</p>
<p>博通在这一领域首次证明了自己的实力，当时它帮助字母表公司设计了其张量处理单元，这些单元现在是 Google Cloud 的关键组成部分。这一成功使其赢得了 <strong>Meta Platforms</strong> 和字节跳动等客户的设计订单。博通管理层表示，这三位客户在 2027 财年（截至 2027 年 10 月）代表着 600 亿到 900 亿美元的机会。</p>
<p>然而，博通并没有止步于这三位客户，最近透露第四位客户，广泛认为是 OpenAI，已下了 100 亿美元的定制芯片订单，预计明年交付。虽然两家公司合作的消息早已为人所知，但它能够如此迅速地开发出芯片并在明年准备就绪，令人感到非常惊讶。由于 OpenAI 和 <strong>甲骨文</strong>(NYSE: ORCL) 已达成协议，在未来五年内在数据中心投资 3000 亿美元，这将为博通带来巨大的机会。</p>
<h2>2. AMD</h2>
<p>虽然博通看起来将成为推理领域的巨大赢家，<strong>超威半导体</strong>(AMD -0.75%) 也将受益。构建定制 AI 芯片涉及大量前期成本并需要时间，因此并不是每家公司都能选择。与此同时，AMD 在推理领域凭借其 GPU 开辟了一个稳固的市场，并且随着市场朝这个方向发展，它也将受益。</p>
<p>AMD 已经在推理领域取得了一些进展。其最新版本的软件平台 ROCm 7 是专为推理设计的，尽管它不如英伟达的 CUDA，但通常被认为足够用于推理。最大的 AI 运营商之一已经在其推理流量中运行了相当大一部分在 AMD 的 GPU 上，而前十名 AI 玩家中有七家现在使用其硬件。</p>
<p>AMD 和博通也是 UALink 联盟的一部分，该联盟正在推动一种开放标准互连替代英伟达的 NVLink。NVLink 使英伟达获得了巨大的优势，因为它基本上允许其 GPU 作为一个单元运作。开放的互连标准将使客户更容易混合和匹配来自不同供应商的 AI 芯片。</p>
<p>鉴于 AMD 的 AI 收入基础远小于博通，随着市场向推理转变，它应该有巨大的机会。如果它能在这个快速增长的市场中占据一点份额，股票应该会有很大的上升空间。</p>
<h2>3. 台湾半导体制造公司</h2>
<p>随着 AI 基础设施支出在未来几年继续飙升，<strong>台湾半导体制造公司</strong>(TSM 0.24%) 是参与这一现象的最佳和最安全的方式。原因在于，无论谁在 AI 芯片竞赛中获得市场份额，它都将受益。</p>
<p>台积电是全球最大的半导体制造商，服务于所有顶级芯片设计公司。该公司在先进芯片方面取得了巨大的领先，这些芯片需要在更小的节点尺寸下制造，节点尺寸指的是可以在芯片上容纳多少个晶体管。由于竞争对手 <strong>英特尔</strong> 和 <strong>三星</strong> 在小节点上的良率受到挑战，台积电已成为芯片公司的重要合作伙伴。这使得台积电成为可靠生产高性能半导体的唯一真正选择。它还赋予了台积电强大的定价权力，有报道称它计划明年提高 10% 的价格。</p>
<p>台积电预计 AI 芯片需求将在 2028 年前以超过 40% 的年复合增长率 (CAGR) 增长。该公司已经在向 2nm 生产迈进，这将使芯片更加高效，并使其在竞争对手面前保持领先。除了 AI，自动驾驶、机器人技术和量子计算都是长期的顺风因素，因为它们都需要先进的芯片。</p>
<p>因此，台积电是投资持续的 AI 基础设施支出热潮的最佳方式之一。</p>

博通

<p>Nvidia 预测人工智能基础设施支出将增加五倍，为其他公司带来机会。博通处于有利位置，已从 OpenAI 获得 100 亿美元的订单，并有可能从主要客户那里获得 600 亿到 900 亿美元的订单。AMD 也将从向推理的转变中受益，利用其 GPU 并加入 UALink 联盟。无论哪家公司在人工智能芯片领域领先，台湾半导体制造公司（TSMC）都将获利，预计人工智能芯片需求年复合增长率将达到 40%，并计划进行先进的 2nm 生产</p>

<p><strong>Nvidia</strong>(NVDA -2.62%) recently predicted that artificial intelligence (AI) infrastructure spending could rise fivefold over the next several years. While it surely will benefit, it's not the only one.</p>
<div class="lb-trans"><p><strong>英伟达</strong>(NVDA -2.62%) 最近预测，人工智能 (AI) 基础设施支出在未来几年可能会增长五倍。虽然它肯定会受益，但并不是唯一的受益者。</p>
</div><p>Let's look at three AI stocks set to benefit from this surge in spending and why they might be buying opportunities.</p>
<div class="lb-trans"><p>让我们看看三只将从这波支出激增中受益的 AI 股票，以及它们为何可能是买入机会。</p>
</div><p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MzI5NzEvZ2V0dHlpbWFnZXMtMTQ5MjMxOTg4NC5qcGcmdz03MDA" alt="" original-src="https://g.foolcdn.com/image/?url=https%3A%2F%2Fg.foolcdn.com%2Feditorial%2Fimages%2F832971%2Fgettyimages-1492319884.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>1. Broadcom</h2>
<div class="lb-trans"><h2>1. 博通</h2>
</div><p><strong>Broadcom</strong>(AVGO -3.93%) has emerged as one of the most important players in AI. With large hyperscalers (owners of massive data centers) looking for alternatives to Nvidia to help reduce costs and diversify their supply chains, more and more have been turning to Broadcom to help design custom application-specific integrated circuits, or ASICs, for their AI workloads. Custom chips are generally more efficient and cheaper to operate than graphics processing units (GPUs), which is critical as inference demand takes off.</p>
<div class="lb-trans"><p><strong>博通</strong>(AVGO -3.93%) 已成为人工智能领域最重要的参与者之一。随着大型超大规模数据中心的拥有者寻求替代英伟达的方案以降低成本和多样化其供应链，越来越多的公司开始转向博通，帮助设计用于其 AI 工作负载的定制应用特定集成电路 (ASIC)。定制芯片通常比图形处理单元 (GPU) 更高效且更便宜，这在推理需求激增的情况下至关重要。</p>
</div><p>Broadcom first proved itself in this area when it helped Alphabet design its tensor processing units, which are now a key part of Google Cloud. That success led to design wins with customers like <strong>Meta Platforms</strong> and ByteDance. Broadcom management has said these three customers represent a $60 billion to $90 billion opportunity by fiscal 2027 (ending October 2027).</p>
<div class="lb-trans"><p>博通在这一领域首次证明了自己的实力，当时它帮助字母表公司设计了其张量处理单元，这些单元现在是 Google Cloud 的关键组成部分。这一成功使其赢得了 <strong>Meta Platforms</strong> 和字节跳动等客户的设计订单。博通管理层表示，这三位客户在 2027 财年（截至 2027 年 10 月）代表着 600 亿到 900 亿美元的机会。</p>
</div><p>However, Broadcom has not stopped with those three customers, and it recently revealed that a fourth customer, widely believed to be OpenAI, has placed a $10 billion order for custom chips that will be delivered next year. While it was well known that the two companies were working together, how quickly it was able to develop the chips and have them ready for next year was a huge surprise. With OpenAI and <strong>Oracle</strong>(NYSE: ORCL) having an agreement in place to spend $300 billion on data centers over the next five years, this will be a massive opportunity for Broadcom.</p>
<div class="lb-trans"><p>然而，博通并没有止步于这三位客户，最近透露第四位客户，广泛认为是 OpenAI，已下了 100 亿美元的定制芯片订单，预计明年交付。虽然两家公司合作的消息早已为人所知，但它能够如此迅速地开发出芯片并在明年准备就绪，令人感到非常惊讶。由于 OpenAI 和 <strong>甲骨文</strong>(NYSE: ORCL) 已达成协议，在未来五年内在数据中心投资 3000 亿美元，这将为博通带来巨大的机会。</p>
</div><h2>2. AMD</h2>
<div class="lb-trans"><h2>2. AMD</h2>
</div><p>While Broadcom looks poised to be a huge inference winner, <strong>Advanced Micro Devices</strong>(AMD -0.75%) is also set to benefit. Building custom AI chips involves a lot of upfront costs and takes time, and as such, they are not an option for every company. AMD, meanwhile, has carved a solid niche in inference with its GPUs, and it should also benefit as the market heads in this direction.</p>
<div class="lb-trans"><p>虽然博通看起来将成为推理领域的巨大赢家，<strong>超威半导体</strong>(AMD -0.75%) 也将受益。构建定制 AI 芯片涉及大量前期成本并需要时间，因此并不是每家公司都能选择。与此同时，AMD 在推理领域凭借其 GPU 开辟了一个稳固的市场，并且随着市场朝这个方向发展，它也将受益。</p>
</div><p>AMD has already been making inroads in inference. The newest version of its software platform, ROCm 7, was designed with inference in mind, and while it is not as good as Nvidia's CUDA, it is generally viewed as good enough for inference. One of the largest AI operators is already running a big portion of its inference traffic on AMD's GPUs, and seven of the top-10 AI players now use some of its hardware.</p>
<div class="lb-trans"><p>AMD 已经在推理领域取得了一些进展。其最新版本的软件平台 ROCm 7 是专为推理设计的，尽管它不如英伟达的 CUDA，但通常被认为足够用于推理。最大的 AI 运营商之一已经在其推理流量中运行了相当大一部分在 AMD 的 GPU 上，而前十名 AI 玩家中有七家现在使用其硬件。</p>
</div><p>AMD and Broadcom are also a part of the UALink Consortium, which is pushing an open standard interconnect alternative to Nvidia's NVLink. NVLink has given Nvidia a big advantage, as it essentially allows its GPUs to act as a single unit. An open interconnect standard would allow customers to more easily mix and match AI chips from different vendors.</p>
<div class="lb-trans"><p>AMD 和博通也是 UALink 联盟的一部分，该联盟正在推动一种开放标准互连替代英伟达的 NVLink。NVLink 使英伟达获得了巨大的优势，因为它基本上允许其 GPU 作为一个单元运作。开放的互连标准将使客户更容易混合和匹配来自不同供应商的 AI 芯片。</p>
</div><p>Given AMD's much smaller AI revenue base, it should have a huge opportunity as the market shifts toward inference. If it can just take a little share in this fast-growing market, the stock should have plenty of upside from here.</p>
<div class="lb-trans"><p>鉴于 AMD 的 AI 收入基础远小于博通，随着市场向推理转变，它应该有巨大的机会。如果它能在这个快速增长的市场中占据一点份额，股票应该会有很大的上升空间。</p>
</div><h2>3. Taiwan Semiconductor Manufacturing</h2>
<div class="lb-trans"><h2>3. 台湾半导体制造公司</h2>
</div><p>With AI infrastructure spending set to continue to soar in the coming years, <strong>Taiwan Semiconductor</strong> <strong>Manufacturing</strong>(TSM 0.24%) is one of the best and safest ways to play this phenomenon. The reason is that it doesn't matter who takes share in the AI chip race -- it wins regardless.</p>
<div class="lb-trans"><p>随着 AI 基础设施支出在未来几年继续飙升，<strong>台湾半导体制造公司</strong>(TSM 0.24%) 是参与这一现象的最佳和最安全的方式。原因在于，无论谁在 AI 芯片竞赛中获得市场份额，它都将受益。</p>
</div><p>TSMC is the largest manufacturer of semiconductors in the world, serving all the top chip designers. The company has taken a huge lead with advanced chips, which need to be manufactured at smaller node sizes, referring to how many transistors can fit onto a chip. The company has become an important partner to chip companies, as rivals <strong>Intel</strong> and <strong>Samsung</strong> have struggled with yields at smaller nodes. That leaves TSMC as the only real option for producing high-performance semiconductors reliably. It also gives it strong pricing power, with reports saying it plans to raise prices by 10% next year.</p>
<div class="lb-trans"><p>台积电是全球最大的半导体制造商，服务于所有顶级芯片设计公司。该公司在先进芯片方面取得了巨大的领先，这些芯片需要在更小的节点尺寸下制造，节点尺寸指的是可以在芯片上容纳多少个晶体管。由于竞争对手 <strong>英特尔</strong> 和 <strong>三星</strong> 在小节点上的良率受到挑战，台积电已成为芯片公司的重要合作伙伴。这使得台积电成为可靠生产高性能半导体的唯一真正选择。它还赋予了台积电强大的定价权力，有报道称它计划明年提高 10% 的价格。</p>
</div><p>TSMC expects AI chip demand to grow at an over 40% compounded annual growth rate (CAGR) through 2028. The company is already moving toward 2nm production, which will make chips even more efficient and keep it ahead of rivals. Beyond AI, autonomous driving, robotics, and quantum computing are all long-term tailwinds since they will all need advanced chips.</p>
<div class="lb-trans"><p>台积电预计 AI 芯片需求将在 2028 年前以超过 40% 的年复合增长率 (CAGR) 增长。该公司已经在向 2nm 生产迈进，这将使芯片更加高效，并使其在竞争对手面前保持领先。除了 AI，自动驾驶、机器人技术和量子计算都是长期的顺风因素，因为它们都需要先进的芯片。</p>
</div><p>As such, TSMC is one of the best ways to invest in the continuing AI infrastructure spending boom.</p>
<div class="lb-trans"><p>因此，台积电是投资持续的 AI 基础设施支出热潮的最佳方式之一。</p>
</div>

AI Spending Could Soar 500%: 3 Brilliant AI Stocks to Buy (Hint: Not Nvidia)

- Nvidia 预测人工智能基础设施支出将在未来几年增长五倍。  
- Broadcom、AMD 和台积电是受益于这一趋势的三大公司。  
- 预计台积电的 AI 芯片需求将以超过 40% 的年复合增长率增长。

人工智能支出可能会飙升 500%：3 只值得购买的优秀 AI 股票（提示：不是英伟达）