<p>在 2022 年 11 月底 ChatGPT 推出後不久，大型科技巨頭意識到被稱為 圖形處理單元（GPU）的芯片組可以用於開發下一代人工智能（AI）應用。當時，<strong>Nvidia</strong>（NVDA 1.42%）在 GPU 領域具有先發優勢。因此，該公司的收入迅速飆升至創紀錄的水平，估值也隨之大幅上漲。</p>
<p>在股價快速且持續上漲後，Nvidia 迅速進入了 萬億美元俱樂部。目前，Nvidia 的市值為 4.7 萬億美元，位居這一獨特名單的首位。</p>
<p>雖然 GPU 使 Nvidia 成為全球最有價值的公司，但我認為它下一個萬億美元的機會在其他地方。幸運的是，Nvidia 首席執行官黃仁勳給我們提供了一些線索。讓我們探討一下 Nvidia 最近的投資方向，並評估這些舉措對公司在 AI 基礎設施時代發展軌跡的意義。</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>圖片來源：Nvidia。</p>
<h2>理解 AI 網絡的關鍵角色</h2>
<p>AI 網絡是指連接數據中心內 GPU 集羣的專用互連。如果沒有極高的帶寬、超低的延遲和無損性能，模型訓練和推理部署之間的通信將會產生擁堵，導致 GPU 的利用率降低。</p>
<p>這會導致數十億美元硬件投資的回報減少。從這個意義上説，網絡的重要性與原始計算能力相當。換句話説，如果數據無法在集羣之間高效流動，最快的 GPU 最終也無法提供太大價值。隨着模型變得更大，應用變得更復雜，支撐 AI 開發的底層網絡將成為整體系統性能的限制因素。</p>
<p>展開</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 股票報價" 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>今日變動</p>
<p>(-1.42%) $-2.77</p>
<p>當前價格</p>
<p>$192.97</p>
<h3>關鍵數據點</h3>
<p>市值</p>
<p>$4.7T</p>
<p>日內區間</p>
<p>$191.22 - $195.54</p>
<p>52 周區間</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 網絡解決方案？</h2>
<p>雖然其 GPU 業務備受矚目，但 Nvidia 在多個 針對 AI 的網絡技術 方面悄然積累了深厚的專業知識。</p>
<p>該公司的 InfiniBand 平台提供超低延遲、網絡內計算和高帶寬連接，優化了高性能計算（HPC）和 AI 工作負載。Nvidia 還通過其 Spectrum 系列以太網交換機進行補充，特別是 AI 優化的 Spectrum-X 平台，具有自適應路由、擁堵控制和在標準以太網環境中的可預測行為。</p>
<p>此外，還有 BlueField 數據處理單元（DPU），它們將網絡、存儲和安全任務從 CPU 和 GPU 中卸載。綜合來看，Nvidia 的網絡套件形成了一個全面的高性能基礎，能夠支持超大規模的 AI 部署。</p>
<h2>Nvidia 與多家 AI 網絡領導者合作</h2>
<p>為了加速其在網絡領域的地位並確保額外的供應，Nvidia 進行了幾項有針對性的戰略投資。</p>
<p>在 3 月，Nvidia 分別向 <strong>Coherent</strong> 和 <strong>Lumentum</strong> 投資了 20 億美元。這些合作的理由是為了提升 Nvidia 在光互連和硅光子學方面的地位，這兩者是以較低的功耗在長距離內以高速度傳輸數據的兩個重要層面。Nvidia 又投資了 20 億美元 於 <strong>Marvell Technology</strong>，以加深其在定製 AI 加速器設計方面的影響力。</p>
<p>這些舉措絕非偶然。相反，每項投資都在悄然增強 Nvidia 構建端到端光電和電氣網絡組件的能力，這些組件將疊加在其 GPU 生態系統之上。</p>
<p>通過將其 GPU 架構與不斷擴展的網絡組合，Nvidia 正將自己定位為完整 AI 工廠的關鍵供應商——集成系統，能夠捕獲計算、高速互連、DPU、軟件和先進光子學。目標是使客户能夠部署能夠以空前的規模和效率訓練和運行最大 AI 模型的交鑰匙基礎設施。</p>
<p>從估值的角度來看，Nvidia 目前的 前瞻市盈率（P/E）約為 22。如圖所示，這一倍數遠低於 Nvidia 在 AI 革命初期 GPU 階段的高峰熱情時所達到的高水平。</p>
<p></p>
<p>NVDA 前瞻市盈率 數據來自 YCharts.</p>
<p>我認為市場尚未完全考慮 Nvidia 在網絡領域的擴展。由於所有跡象都指向未來幾年 AI 基礎設施支出的加速，Nvidia 的真實盈利能力可能遠超當前預期。鑑於這些動態，我認為 Nvidia 是一隻值得買入和持有的股票，因為超大規模企業將增加其資本支出預算，分配更多資金用於網絡設備以滿足其產能需求。</p>

英偉達

COHR

Lumentum控股

<p>英偉達下一個萬億增長機會在於人工智能網絡，而不僅僅是 GPU。該公司正在通過 InfiniBand、Spectrum-X 交換機和 BlueField DPU 擴展其產品組合，以優化數據中心性能。對 Coherent、Lumentum 和 Marvell 各投資 20 億美元，增強了其端到端光電網絡能力。這使得英偉達成為完整人工智能基礎設施解決方案的提供商，可能推動收益超出當前市場預期</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 推出後不久，大型科技巨頭意識到被稱為 圖形處理單元（GPU）的芯片組可以用於開發下一代人工智能（AI）應用。當時，<strong>Nvidia</strong>（NVDA 1.42%）在 GPU 領域具有先發優勢。因此，該公司的收入迅速飆升至創紀錄的水平，估值也隨之大幅上漲。</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>在股價快速且持續上漲後，Nvidia 迅速進入了 萬億美元俱樂部。目前，Nvidia 的市值為 4.7 萬億美元，位居這一獨特名單的首位。</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>雖然 GPU 使 Nvidia 成為全球最有價值的公司，但我認為它下一個萬億美元的機會在其他地方。幸運的是，Nvidia 首席執行官黃仁勳給我們提供了一些線索。讓我們探討一下 Nvidia 最近的投資方向，並評估這些舉措對公司在 AI 基礎設施時代發展軌跡的意義。</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>圖片來源：Nvidia。</p>
</div><h2>Understanding the critical role of AI networking</h2>
<div class="lb-trans"><h2>理解 AI 網絡的關鍵角色</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 網絡是指連接數據中心內 GPU 集羣的專用互連。如果沒有極高的帶寬、超低的延遲和無損性能，模型訓練和推理部署之間的通信將會產生擁堵，導致 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>這會導致數十億美元硬件投資的回報減少。從這個意義上説，網絡的重要性與原始計算能力相當。換句話説，如果數據無法在集羣之間高效流動，最快的 GPU 最終也無法提供太大價值。隨着模型變得更大，應用變得更復雜，支撐 AI 開發的底層網絡將成為整體系統性能的限制因素。</p>
</div><p>Expand</p>
<div class="lb-trans"><p>展開</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>今日變動</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>當前價格</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>關鍵數據點</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>日內區間</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 周區間</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 網絡解決方案？</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>雖然其 GPU 業務備受矚目，但 Nvidia 在多個 針對 AI 的網絡技術 方面悄然積累了深厚的專業知識。</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>該公司的 InfiniBand 平台提供超低延遲、網絡內計算和高帶寬連接，優化了高性能計算（HPC）和 AI 工作負載。Nvidia 還通過其 Spectrum 系列以太網交換機進行補充，特別是 AI 優化的 Spectrum-X 平台，具有自適應路由、擁堵控制和在標準以太網環境中的可預測行為。</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，還有 BlueField 數據處理單元（DPU），它們將網絡、存儲和安全任務從 CPU 和 GPU 中卸載。綜合來看，Nvidia 的網絡套件形成了一個全面的高性能基礎，能夠支持超大規模的 AI 部署。</p>
</div><h2>Nvidia is partnering with several AI networking leaders</h2>
<div class="lb-trans"><h2>Nvidia 與多家 AI 網絡領導者合作</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>為了加速其在網絡領域的地位並確保額外的供應，Nvidia 進行了幾項有針對性的戰略投資。</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 分別向 <strong>Coherent</strong> 和 <strong>Lumentum</strong> 投資了 20 億美元。這些合作的理由是為了提升 Nvidia 在光互連和硅光子學方面的地位，這兩者是以較低的功耗在長距離內以高速度傳輸數據的兩個重要層面。Nvidia 又投資了 20 億美元 於 <strong>Marvell Technology</strong>，以加深其在定製 AI 加速器設計方面的影響力。</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>這些舉措絕非偶然。相反，每項投資都在悄然增強 Nvidia 構建端到端光電和電氣網絡組件的能力，這些組件將疊加在其 GPU 生態系統之上。</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>通過將其 GPU 架構與不斷擴展的網絡組合，Nvidia 正將自己定位為完整 AI 工廠的關鍵供應商——集成系統，能夠捕獲計算、高速互連、DPU、軟件和先進光子學。目標是使客户能夠部署能夠以空前的規模和效率訓練和運行最大 AI 模型的交鑰匙基礎設施。</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>從估值的角度來看，Nvidia 目前的 前瞻市盈率（P/E）約為 22。如圖所示，這一倍數遠低於 Nvidia 在 AI 革命初期 GPU 階段的高峰熱情時所達到的高水平。</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 數據來自 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>我認為市場尚未完全考慮 Nvidia 在網絡領域的擴展。由於所有跡象都指向未來幾年 AI 基礎設施支出的加速，Nvidia 的真實盈利能力可能遠超當前預期。鑑於這些動態，我認為 Nvidia 是一隻值得買入和持有的股票，因為超大規模企業將增加其資本支出預算，分配更多資金用於網絡設備以滿足其產能需求。</p>
</div>

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

- 英偉達因 GPU 需求激增而市值達到 4.7 萬億美元，迅速進入萬億俱樂部。  
- 公司在 AI 網絡和光互連技術上持續投資，以提升其整體 AI 基礎設施能力。  
- 隨着 AI 基礎設施支出加速，市場尚未完全反映英偉達的未來收益潛力。

英偉達下一個萬億美元的機會可能與 GPU 毫無關係