<p>在 2022 年 11 月底 ChatGPT 推出后不久，大型科技巨头意识到被称为 图形处理单元（GPU）的芯片组可以用于开发下一代人工智能（AI）应用。当时，<strong>Nvidia</strong>（NVDA 1.42%）在 GPU 领域具有先发优势。因此，该公司的收入迅速飙升至创纪录的水平，估值也随之大幅上涨。</p>
<p>在股价快速且持续上涨后，Nvidia 迅速进入了 万亿美元俱乐部。目前，Nvidia 的市值为 4.7 万亿美元，位居这一独特名单的首位。</p>
<p>虽然 GPU 使 Nvidia 成为全球最有价值的公司，但我认为它下一个万亿美元的机会在其他地方。幸运的是，Nvidia 首席执行官黄仁勋给我们提供了一些线索。让我们探讨一下 Nvidia 最近的投资方向，并评估这些举措对公司在 AI 基础设施时代发展轨迹的意义。</p>
<p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84NzYyNDgvbnZpZGlhLWhlYWRxdWFydGVycy13aXRoLW52aWRpYS1zaWduLWluLWZyb250LTMxLnBuZyZ3PTM4NDA?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" width="2000" height="1200" original-src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84NzYyNDgvbnZpZGlhLWhlYWRxdWFydGVycy13aXRoLW52aWRpYS1zaWduLWluLWZyb250LTMxLnBuZyZ3PTM4NDA"/></p>
<p>图片来源：Nvidia。</p>
<h2>理解 AI 网络的关键角色</h2>
<p>AI 网络是指连接数据中心内 GPU 集群的专用互连。如果没有极高的带宽、超低的延迟和无损性能，模型训练和推理部署之间的通信将会产生拥堵，导致 GPU 的利用率降低。</p>
<p>这会导致数十亿美元硬件投资的回报减少。从这个意义上说，网络的重要性与原始计算能力相当。换句话说，如果数据无法在集群之间高效流动，最快的 GPU 最终也无法提供太大价值。随着模型变得更大，应用变得更复杂，支撑 AI 开发的底层网络将成为整体系统性能的限制因素。</p>
<p>展开</p>
<p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vYXJ0L2NvbXBhbnlsb2dvcy9tYXJrL05WREEucG5nJnc9MTI4?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="Nvidia 股票报价" width="80" height="80" original-src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vYXJ0L2NvbXBhbnlsb2dvcy9tYXJrL05WREEucG5nJnc9MTI4"/></p>
<h2>NASDAQ: NVDA</h2>
<p>Nvidia</p>
<p>今日变动</p>
<p>(-1.42%) $-2.77</p>
<p>当前价格</p>
<p>$192.97</p>
<h3>关键数据点</h3>
<p>市值</p>
<p>$4.7T</p>
<p>日内区间</p>
<p>$191.22 - $195.54</p>
<p>52 周区间</p>
<p>$151.49 - $236.54</p>
<p>成交量</p>
<p>4.8M</p>
<p>平均成交量</p>
<p>161.6M</p>
<p>毛利率</p>
<p>74.15%</p>
<p>股息收益率</p>
<p>0.14%</p>
<h2>Nvidia 是否提供 AI 网络解决方案？</h2>
<p>虽然其 GPU 业务备受瞩目，但 Nvidia 在多个 针对 AI 的网络技术 方面悄然积累了深厚的专业知识。</p>
<p>该公司的 InfiniBand 平台提供超低延迟、网络内计算和高带宽连接，优化了高性能计算（HPC）和 AI 工作负载。Nvidia 还通过其 Spectrum 系列以太网交换机进行补充，特别是 AI 优化的 Spectrum-X 平台，具有自适应路由、拥堵控制和在标准以太网环境中的可预测行为。</p>
<p>此外，还有 BlueField 数据处理单元（DPU），它们将网络、存储和安全任务从 CPU 和 GPU 中卸载。综合来看，Nvidia 的网络套件形成了一个全面的高性能基础，能够支持超大规模的 AI 部署。</p>
<h2>Nvidia 与多家 AI 网络领导者合作</h2>
<p>为了加速其在网络领域的地位并确保额外的供应，Nvidia 进行了几项有针对性的战略投资。</p>
<p>在 3 月，Nvidia 分别向 <strong>Coherent</strong> 和 <strong>Lumentum</strong> 投资了 20 亿美元。这些合作的理由是为了提升 Nvidia 在光互连和硅光子学方面的地位，这两者是以较低的功耗在长距离内以高速度传输数据的两个重要层面。Nvidia 又投资了 20 亿美元 于 <strong>Marvell Technology</strong>，以加深其在定制 AI 加速器设计方面的影响力。</p>
<p>这些举措绝非偶然。相反，每项投资都在悄然增强 Nvidia 构建端到端光电和电气网络组件的能力，这些组件将叠加在其 GPU 生态系统之上。</p>
<p>通过将其 GPU 架构与不断扩展的网络组合，Nvidia 正将自己定位为完整 AI 工厂的关键供应商——集成系统，能够捕获计算、高速互连、DPU、软件和先进光子学。目标是使客户能够部署能够以空前的规模和效率训练和运行最大 AI 模型的交钥匙基础设施。</p>
<p>从估值的角度来看，Nvidia 目前的 前瞻市盈率（P/E）约为 22。如图所示，这一倍数远低于 Nvidia 在 AI 革命初期 GPU 阶段的高峰热情时所达到的高水平。</p>
<p></p>
<p>NVDA 前瞻市盈率 数据来自 YCharts.</p>
<p>我认为市场尚未完全考虑 Nvidia 在网络领域的扩展。由于所有迹象都指向未来几年 AI 基础设施支出的加速，Nvidia 的真实盈利能力可能远超当前预期。鉴于这些动态，我认为 Nvidia 是一只值得买入和持有的股票，因为超大规模企业将增加其资本支出预算，分配更多资金用于网络设备以满足其产能需求。</p>

英伟达

COHR

Lumentum控股

<p>英伟达下一个万亿增长机会在于人工智能网络，而不仅仅是 GPU。该公司正在通过 InfiniBand、Spectrum-X 交换机和 BlueField DPU 扩展其产品组合，以优化数据中心性能。对 Coherent、Lumentum 和 Marvell 各投资 20 亿美元，增强了其端到端光电网络能力。这使得英伟达成为完整人工智能基础设施解决方案的提供商，可能推动收益超出当前市场预期</p>

<p>Shortly after the launch of ChatGPT in late November 2022, big tech hyperscalers realized that chipsets known as graphics processing units (GPUs) could be used to develop next-generation applications in artificial intelligence (AI). At the time, <strong>Nvidia</strong>( NVDA 1.42%) had a first-mover advantage in the GPU landscape. As a result, the company's revenue skyrocketed to record levels seemingly overnight -- as did its valuation.</p>
<div class="lb-trans"><p>在 2022 年 11 月底 ChatGPT 推出后不久，大型科技巨头意识到被称为 图形处理单元（GPU）的芯片组可以用于开发下一代人工智能（AI）应用。当时，<strong>Nvidia</strong>（NVDA 1.42%）在 GPU 领域具有先发优势。因此，该公司的收入迅速飙升至创纪录的水平，估值也随之大幅上涨。</p>
</div><p>After rapid and sustained share price appreciation, Nvidia quickly entered the trillion-dollar club. With a market capitalization of $4.7 trillion, Nvidia now sits at the top of this exclusive roster.</p>
<div class="lb-trans"><p>在股价快速且持续上涨后，Nvidia 迅速进入了 万亿美元俱乐部。目前，Nvidia 的市值为 4.7 万亿美元，位居这一独特名单的首位。</p>
</div><p>While GPUs have ushered in Nvidia's status as the world's most valuable company, I think its next trillion-dollar opportunity lies elsewhere. Luckily, Nvidia CEO Jensen Huang has given us some clues. Let's explore where Nvidia has been investing lately, and assess what these moves could mean for the company's trajectory as the AI infrastructure era takes shape.</p>
<div class="lb-trans"><p>虽然 GPU 使 Nvidia 成为全球最有价值的公司，但我认为它下一个万亿美元的机会在其他地方。幸运的是，Nvidia 首席执行官黄仁勋给我们提供了一些线索。让我们探讨一下 Nvidia 最近的投资方向，并评估这些举措对公司在 AI 基础设施时代发展轨迹的意义。</p>
</div><p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84NzYyNDgvbnZpZGlhLWhlYWRxdWFydGVycy13aXRoLW52aWRpYS1zaWduLWluLWZyb250LTMxLnBuZyZ3PTM4NDA" alt="Nvidia headquarters with company logo on a sign out front." original-src="https://g.foolcdn.com/image/?url=https%3A%2F%2Fg.foolcdn.com%2Feditorial%2Fimages%2F876248%2Fnvidia-headquarters-with-nvidia-sign-in-front-31.png&amp;w=3840&amp;op=resize"/></p>
<p>Image source: Nvidia.</p>
<div class="lb-trans"><p>图片来源：Nvidia。</p>
</div><h2>Understanding the critical role of AI networking</h2>
<div class="lb-trans"><h2>理解 AI 网络的关键角色</h2>
</div><p>AI networking refers to the specialized interconnects that link GPU clusters inside data centers. Without extremely high bandwidth, ultra-low latency, and lossless performance, communication among model training and inference deployments creates congestion that leaves GPUs underutilized.</p>
<div class="lb-trans"><p>AI 网络是指连接数据中心内 GPU 集群的专用互连。如果没有极高的带宽、超低的延迟和无损性能，模型训练和推理部署之间的通信将会产生拥堵，导致 GPU 的利用率降低。</p>
</div><p>This results in diminished returns on multi-billion-dollar hardware investments. In this sense, networking can be seen as being as important as raw compute. In other words, the fastest GPUs ultimately deliver little value if data cannot flow efficiently between clusters. As models grow larger and applications become more complex, the underlying network stitching AI development together becomes the limiting factor for overall system performance.</p>
<div class="lb-trans"><p>这会导致数十亿美元硬件投资的回报减少。从这个意义上说，网络的重要性与原始计算能力相当。换句话说，如果数据无法在集群之间高效流动，最快的 GPU 最终也无法提供太大价值。随着模型变得更大，应用变得更复杂，支撑 AI 开发的底层网络将成为整体系统性能的限制因素。</p>
</div><p>Expand</p>
<div class="lb-trans"><p>展开</p>
</div><p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vYXJ0L2NvbXBhbnlsb2dvcy9tYXJrL05WREEucG5nJnc9MTI4" alt="Nvidia Stock Quote" original-src="https://g.foolcdn.com/image/?url=https%3A%2F%2Fg.foolcdn.com%2Fart%2Fcompanylogos%2Fmark%2FNVDA.png&amp;w=128&amp;op=resize"/></p>
<h2>NASDAQ: NVDA</h2>
<div class="lb-trans"><h2>NASDAQ: NVDA</h2>
</div><p>Nvidia</p>
<div class="lb-trans"><p>Nvidia</p>
</div><p>Today's Change</p>
<div class="lb-trans"><p>今日变动</p>
</div><p>(-1.42%) $-2.77</p>
<div class="lb-trans"><p>(-1.42%) $-2.77</p>
</div><p>Current Price</p>
<div class="lb-trans"><p>当前价格</p>
</div><p>$192.97</p>
<div class="lb-trans"><p>$192.97</p>
</div><h3>Key Data Points</h3>
<div class="lb-trans"><h3>关键数据点</h3>
</div><p>Market Cap</p>
<div class="lb-trans"><p>市值</p>
</div><p>$4.7T</p>
<div class="lb-trans"><p>$4.7T</p>
</div><p>Day's Range</p>
<div class="lb-trans"><p>日内区间</p>
</div><p>$191.22 - $195.54</p>
<div class="lb-trans"><p>$191.22 - $195.54</p>
</div><p>52wk Range</p>
<div class="lb-trans"><p>52 周区间</p>
</div><p>$151.49 - $236.54</p>
<div class="lb-trans"><p>$151.49 - $236.54</p>
</div><p>Volume</p>
<div class="lb-trans"><p>成交量</p>
</div><p>4.8M</p>
<div class="lb-trans"><p>4.8M</p>
</div><p>Avg Vol</p>
<div class="lb-trans"><p>平均成交量</p>
</div><p>161.6M</p>
<div class="lb-trans"><p>161.6M</p>
</div><p>Gross Margin</p>
<div class="lb-trans"><p>毛利率</p>
</div><p>74.15%</p>
<div class="lb-trans"><p>74.15%</p>
</div><p>Dividend Yield</p>
<div class="lb-trans"><p>股息收益率</p>
</div><p>0.14%</p>
<div class="lb-trans"><p>0.14%</p>
</div><h2>Does Nvidia even offer AI networking solutions?</h2>
<div class="lb-trans"><h2>Nvidia 是否提供 AI 网络解决方案？</h2>
</div><p>While its GPU business takes the spotlight, Nvidia has quietly assembled deep expertise across multiple networking technologies tailored for AI.</p>
<div class="lb-trans"><p>虽然其 GPU 业务备受瞩目，但 Nvidia 在多个 针对 AI 的网络技术 方面悄然积累了深厚的专业知识。</p>
</div><p>The company's InfiniBand platform offers ultra-low latency, in-network computing, and high-bandwidth connectivity optimized for high-performance computing (HPC) and AI workloads. Nvidia complements this with its Spectrum family of Ethernet switches -- particularly the AI-optimized Spectrum-X platform, which features adaptive routing, congestion control, and predictable behavior within standard Ethernet environments.</p>
<div class="lb-trans"><p>该公司的 InfiniBand 平台提供超低延迟、网络内计算和高带宽连接，优化了高性能计算（HPC）和 AI 工作负载。Nvidia 还通过其 Spectrum 系列以太网交换机进行补充，特别是 AI 优化的 Spectrum-X 平台，具有自适应路由、拥堵控制和在标准以太网环境中的可预测行为。</p>
</div><p>Rounding out the offering are the BlueField data processing units (DPUs), which offload networking, storage, and security tasks from CPUs and GPUs. Taken together, Nvidia's networking suite forms a comprehensive, high-performance foundation capable of powering hyperscale AI deployments.</p>
<div class="lb-trans"><p>此外，还有 BlueField 数据处理单元（DPU），它们将网络、存储和安全任务从 CPU 和 GPU 中卸载。综合来看，Nvidia 的网络套件形成了一个全面的高性能基础，能够支持超大规模的 AI 部署。</p>
</div><h2>Nvidia is partnering with several AI networking leaders</h2>
<div class="lb-trans"><h2>Nvidia 与多家 AI 网络领导者合作</h2>
</div><p>To accelerate its position and secure additional supply in the networking domain, Nvidia has made several targeted strategic investments.</p>
<div class="lb-trans"><p>为了加速其在网络领域的地位并确保额外的供应，Nvidia 进行了几项有针对性的战略投资。</p>
</div><p>In March, Nvidia invested $2 billion each in <strong>Coherent</strong> and <strong>Lumentum</strong>. The rationale behind these partnerships is to advance Nvidia's position in optical interconnects and silicon photonics, two essential layers for transmitting data at high speeds over long distances with lower power consumption. Nvidia invested another $2 billion in <strong>Marvell Technology</strong> to deepen its reach in designing custom AI accelerators.</p>
<div class="lb-trans"><p>在 3 月，Nvidia 分别向 <strong>Coherent</strong> 和 <strong>Lumentum</strong> 投资了 20 亿美元。这些合作的理由是为了提升 Nvidia 在光互连和硅光子学方面的地位，这两者是以较低的功耗在长距离内以高速度传输数据的两个重要层面。Nvidia 又投资了 20 亿美元 于 <strong>Marvell Technology</strong>，以加深其在定制 AI 加速器设计方面的影响力。</p>
</div><p>These moves are far from random. Rather, each investment quietly strengthens Nvidia's ability to build end-to-end optical and electrical networking components layered atop its GPU ecosystem.</p>
<div class="lb-trans"><p>这些举措绝非偶然。相反，每项投资都在悄然增强 Nvidia 构建端到端光电和电气网络组件的能力，这些组件将叠加在其 GPU 生态系统之上。</p>
</div><p>By combining its GPUs' architectures with an expanding networking portfolio, Nvidia is positioning itself as the key provider of complete AI factories -- integrated systems that capture compute, high-speed interconnects, DPUs, software, and advanced photonics. The goal is to enable customers to deploy turnkey infrastructure capable of training and running the largest AI models at unprecedented scale and efficiency.</p>
<div class="lb-trans"><p>通过将其 GPU 架构与不断扩展的网络组合，Nvidia 正将自己定位为完整 AI 工厂的关键供应商——集成系统，能够捕获计算、高速互连、DPU、软件和先进光子学。目标是使客户能够部署能够以空前的规模和效率训练和运行最大 AI 模型的交钥匙基础设施。</p>
</div><p>From a valuation standpoint, Nvidia currently trades at a forward price-to-earnings (P/E) ratio of roughly 22. As the chart illustrates, this multiple sits well below the elevated levels Nvidia reached during the peak enthusiasm of the AI revolution's initial GPU phase.</p>
<div class="lb-trans"><p>从估值的角度来看，Nvidia 目前的 前瞻市盈率（P/E）约为 22。如图所示，这一倍数远低于 Nvidia 在 AI 革命初期 GPU 阶段的高峰热情时所达到的高水平。</p>
</div><p></p>
<div class="lb-trans"><p></p>
</div><p>NVDA PE Ratio (Forward) data by YCharts.</p>
<div class="lb-trans"><p>NVDA 前瞻市盈率 数据来自 YCharts.</p>
</div><p>I don't think the market has fully incorporated Nvidia's expansion into networking. Since all signs point to an acceleration in AI infrastructure spending over the next several years, Nvidia's true earnings power could prove considerably larger than current expectations. Given these dynamics, I see Nvidia as a no-brainer stock to buy and hold, as hyperscalers bolster their capex budgets, allocating more to networking gear to meet their capacity needs.</p>
<div class="lb-trans"><p>我认为市场尚未完全考虑 Nvidia 在网络领域的扩展。由于所有迹象都指向未来几年 AI 基础设施支出的加速，Nvidia 的真实盈利能力可能远超当前预期。鉴于这些动态，我认为 Nvidia 是一只值得买入和持有的股票，因为超大规模企业将增加其资本支出预算，分配更多资金用于网络设备以满足其产能需求。</p>
</div>

Nvidia's Next $1 Trillion Opportunity May Have Nothing to Do With GPUs

- 英伟达因 GPU 需求激增而市值达到 4.7 万亿美元，迅速进入万亿俱乐部。  
- 公司在 AI 网络和光互连技术上持续投资，以提升其整体 AI 基础设施能力。  
- 随着 AI 基础设施支出加速，市场尚未完全反映英伟达的未来收益潜力。

英伟达下一个万亿美元的机会可能与 GPU 毫无关系