<p>在過去兩年中，沒有哪個頭條新聞比人工智能（AI）的崛起更能主導華爾街。賦予軟件和系統自主推理和行動的能力，使得這項技術似乎擁有無限的長期潛力。</p>
<p>儘管人工智能解決方案的採用率和實用性仍在變化，但普華永道的分析師認為，這項顛覆性技術將在 2030 年前徹底改變全球增長軌跡。到十年之交，普華永道的《Sizing the Prize》報告預計，人工智能將對國內生產總值產生 15.7 萬億美元的積極影響。</p>
<p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MTI3MDEvd29tYW4tYXJ0aWZpY2lhbC1pbnRlbGxpZ2VuY2UtYWktdmlydHVhbC1hZ2VudC1jb21wdXRlci1jbG91ZC1nZXR0eS5qcGcmdz03MDA?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-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MTI3MDEvd29tYW4tYXJ0aWZpY2lhbC1pbnRlbGxpZ2VuY2UtYWktdmlydHVhbC1hZ2VudC1jb21wdXRlci1jbG91ZC1nZXR0eS5qcGcmdz03MDA"/></p>
<p>圖片來源：Getty Images。</p>
<p>15.7 萬億美元的蛋糕足以容納一長串贏家。然而，沒有哪家公司比圖形處理單元（GPU）巨頭 <strong>Nvidia</strong>（NVDA -1.51%）更直接受益於人工智能革命。Nvidia 的業務從 2022 年底的 3600 億美元增長到兩年後接近 4 萬億美元的市值。</p>
<p>儘管 Nvidia 無疑是人工智能的成功故事，但這家領先企業很可能很快會成為自身成功的受害者。</p>
<h2>Nvidia 的崛起簡直是教科書般的</h2>
<p>在連接 Nvidia 完美運營擴展如何可能自我解構之前，瞭解它如何成為人工智能革命基礎的背景是很重要的。</p>
<p>Nvidia 成功的核心在於其 Hopper（H100）和下一代 Blackwell GPU。前者在過去兩年中有效地成為了人工智能加速數據中心的標準。同時，最近推出的 Blackwell GPU 架構顯著提高了生成性人工智能解決方案和量子計算的計算能力，同時比其前身芯片更具能效。</p>
<p>為了進一步説明這一點，目前沒有其他芯片製造商能夠接近 Nvidia 硬件的全面功能。雖然其他公司可能在某些任務上提供狹窄的競爭優勢，但沒有其他 GPU 在人工智能加速數據中心的計算性能上真正挑戰 Hopper 或 Blackwell。</p>
<p>更糟糕的是，Nvidia 恰好處於受益於人工智能 GPU 稀缺的有利位置。儘管全球領先的芯片製造公司 <strong>台積電</strong> 努力迅速擴大其芯片 - 晶圓 - 基板（CoWoS）能力——CoWoS 對於包裝高帶寬內存以滿足高計算數據中心的需求至關重要——但人工智能 GPU 的需求始終超過其供應。因此，Nvidia 對其人工智能 GPU 的定價溢價達到了 100%（或更高），相較於競爭對手。</p>
<p>Nvidia 的 CUDA 軟件平台將這一切完美結合。這個工具包被開發者用來最大化其 GPU 的性能，以及構建大型語言模型，對於保持客户在 Nvidia 產品和服務生態系統內至關重要。</p>
<h2>Nvidia 可能成為自身成功的受害者</h2>
<p>儘管華爾街分析師的 <strong>壓倒性</strong> 共識認為這家人工智能寵兒將繼續上漲——平均分析師目標價為 171 美元，意味着 50% 的上漲空間——但存在兩條可行路徑，可能導致 Nvidia 成為自身成功的受害者，其股價因此受損。</p>
<p>首先，Nvidia 的人工智能 GPU 在計算能力上被認為是優越的。儘管短期需求繼續壓倒人工智能 GPU 的供應，但 Hopper 和 Blackwell 的計算優勢可能會顯著減緩未來的升級/更換週期。</p>
<p>可以這樣想：當 <strong>蘋果</strong> 首次推出 iPhone 時，消費者和企業紛紛抓住機會升級到這款革命性設備。隨着時間的推移，下一代 iPhone 版本雖然進行了升級，但與 2007 年原始 iPhone 對消費者和企業的影響相比，變化遠不如前。由於 Nvidia 的硬件價格更高且速度更快，企業升級其人工智能數據中心硬件的動力可能會被推遲多年。這對華爾街期望其保持驚人增長率的公司來説並不是好消息。</p>
<p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MTI3MDEvZGF0YS1jZW50ZXItZW5naW5lZXItZ3B1LWFydGlmaWNpYWwtaW50ZWxsaWdlbmNlLWFpLWNsb3VkLXNlcnZlci1nZXR0eS5qcGcmdz03MDA?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-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MTI3MDEvZGF0YS1jZW50ZXItZW5naW5lZXItZ3B1LWFydGlmaWNpYWwtaW50ZWxsaWdlbmNlLWFpLWNsb3VkLXNlcnZlci1nZXR0eS5qcGcmdz03MDA"/></p>
<p>圖片來源：Getty Images。</p>
<p>Nvidia 的第二個潛在問題是，華爾街可能會因為在未來幾個季度放棄其在人工智能數據中心的部分壟斷市場份額而懲罰它。</p>
<p>根據半導體公司 TechInsights 的分析，Nvidia 在 2022 年和 2023 年佔據了企業數據中心出貨 GPU 的驚人 98%。即使像 <strong>超微</strong> 這樣的公司正在加大 Instinct 系列人工智能加速芯片的生產，Nvidia 仍然可能在去年佔據了人工智能數據中心 “頭腦” 的近乎壟斷的市場份額。</p>
<p>Nvidia 面臨的問題是，其許多按淨銷售額計算的頂級客户，主要包括 “七大巨頭” 的成員，正在內部開發自己的人工智能 GPU 和人工智能解決方案，以便在其數據中心中使用。儘管這些內部開發的芯片對 Hopper 或 Blackwell 的計算速度並不構成威脅，並且沒有計劃對外銷售，但它們可能會導致 Nvidia 在未來的數據中心市場中失去機會。這些芯片的價格更便宜且更易獲得。</p>
<p>如果未來的升級週期變得更加漫長，或者華爾街的人工智能寵兒失去了寶貴的數據中心市場份額給其頂級客户，那麼它最大的競爭優勢——人工智能 GPU 的稀缺性——將會消失。這意味着定價能力下降，未來的利潤率也會減弱。</p>
<p>英偉達的歷史性上升已沒有犯錯的餘地。如果公司的增長率放緩，甚至如果其毛利率繼續從 78.4% 的高峰下降，它可能很快就會成為自身成功的犧牲品。</p>

英偉達

<p>英偉達作為領先的 GPU 製造商，因人工智能（AI）的崛起而顯著受益，市值從 3600 億美元增長到近 4 萬億美元，僅在兩年多的時間裏。然而，分析師警告稱，英偉達可能會因自身的成功而面臨挑戰。其 AI-GPU 的優勢可能會減緩未來的升級週期，類似於 iPhone 的演變，而潛在的市場份額損失可能導致華爾街的負面反應。儘管前景樂觀，英偉達的增長可能會受到這些因素的制約</p>

<p>For more than two years, no headlines have dominated Wall Street more than those dealing with the rise of artificial intelligence (AI). Empowering software and systems with the capacity to reason and act on their own gives this technology a seemingly limitless long-term ceiling.</p>
<div class="lb-trans"><p>在過去兩年中，沒有哪個頭條新聞比人工智能（AI）的崛起更能主導華爾街。賦予軟件和系統自主推理和行動的能力，使得這項技術似乎擁有無限的長期潛力。</p>
</div><p>While the adoption rate and utility of AI solutions remains fluid, analysts at PwC see this game-changing technology completely altering the global growth trajectory by 2030. Come the turn of the decade, PwC's <em>Sizing the Prize</em> report estimates a $15.7 trillion positive impact on gross domestic product, all thanks to AI.</p>
<div class="lb-trans"><p>儘管人工智能解決方案的採用率和實用性仍在變化，但普華永道的分析師認為，這項顛覆性技術將在 2030 年前徹底改變全球增長軌跡。到十年之交，普華永道的《Sizing the Prize》報告預計，人工智能將對國內生產總值產生 15.7 萬億美元的積極影響。</p>
</div><p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MTI3MDEvd29tYW4tYXJ0aWZpY2lhbC1pbnRlbGxpZ2VuY2UtYWktdmlydHVhbC1hZ2VudC1jb21wdXRlci1jbG91ZC1nZXR0eS5qcGcmdz03MDA" alt="" original-src="https://g.foolcdn.com/image/?url=https%3A%2F%2Fg.foolcdn.com%2Feditorial%2Fimages%2F812701%2Fwoman-artificial-intelligence-ai-virtual-agent-computer-cloud-getty.jpg&amp;op=resize&amp;w=700"/></p>
<p>Image source: Getty Images.</p>
<div class="lb-trans"><p>圖片來源：Getty Images。</p>
</div><p>A $15.7 trillion pie is big enough to accommodate a long list of winners. However, no company has been a more direct beneficiary of the AI revolution than graphics processing unit (GPU) giant <strong>Nvidia</strong>(NVDA -1.51%). Nvidia grew from a $360 billion business at the end of 2022 to one that came within a stone's throw of reaching a $4 trillion market cap on an intra-day basis just over two years later.</p>
<div class="lb-trans"><p>15.7 萬億美元的蛋糕足以容納一長串贏家。然而，沒有哪家公司比圖形處理單元（GPU）巨頭 <strong>Nvidia</strong>（NVDA -1.51%）更直接受益於人工智能革命。Nvidia 的業務從 2022 年底的 3600 億美元增長到兩年後接近 4 萬億美元的市值。</p>
</div><p>Although there's no denying Nvidia has been an AI success story, there's a real possibility this leading business could soon become a victim of its own success.</p>
<div class="lb-trans"><p>儘管 Nvidia 無疑是人工智能的成功故事，但這家領先企業很可能很快會成為自身成功的受害者。</p>
</div><h2>Nvidia's ascent has been nothing short of textbook</h2>
<div class="lb-trans"><h2>Nvidia 的崛起簡直是教科書般的</h2>
</div><p>Before connecting the dots of how Nvidia's perfect operating expansion can potentially unravel on itself, it's important to understand the background of how it became the foundation of the AI revolution.</p>
<div class="lb-trans"><p>在連接 Nvidia 完美運營擴展如何可能自我解構之前，瞭解它如何成為人工智能革命基礎的背景是很重要的。</p>
</div><p>At the heart of Nvidia's success is its Hopper (H100) and next-generation Blackwell GPUs. The former effectively became the standard in AI-accelerated data centers over the previous two years. Meanwhile, the recently rolled-out Blackwell GPU architecture meaningfully improves computational capabilities for generative AI solutions and quantum computing, all while being considerably more energy-efficient than its predecessor chip.</p>
<div class="lb-trans"><p>Nvidia 成功的核心在於其 Hopper（H100）和下一代 Blackwell GPU。前者在過去兩年中有效地成為了人工智能加速數據中心的標準。同時，最近推出的 Blackwell GPU 架構顯著提高了生成性人工智能解決方案和量子計算的計算能力，同時比其前身芯片更具能效。</p>
</div><p>To build on this point, no other chipmakers have come particularly close, as of now, to challenging the all-around functionality of Nvidia's hardware. While others might offer narrow competitive advantages in certain tasks, no other GPUs have truly challenged the Hopper or Blackwell on the basis of computing performance within AI-accelerated data centers.</p>
<div class="lb-trans"><p>為了進一步説明這一點，目前沒有其他芯片製造商能夠接近 Nvidia 硬件的全面功能。雖然其他公司可能在某些任務上提供狹窄的競爭優勢，但沒有其他 GPU 在人工智能加速數據中心的計算性能上真正挑戰 Hopper 或 Blackwell。</p>
</div><p>Adding fuel to the fire, Nvidia has been perfectly positioned to benefit from AI-GPU scarcity. Despite the best efforts of world-leading chip fabrication company <strong>Taiwan Semiconductor Manufacturing</strong> to rapidly expand its chip-on-wafer-on-substrate capacity (CoWoS) -- CoWoS is essential for packaging the high-bandwidth memory necessary for high-compute data centers -- demand for AI-GPUs has persistently overwhelmed their supply. As a result, Nvidia has been commanding a 100% (or greater) price premium for its AI-GPUs, relative to its competition.</p>
<div class="lb-trans"><p>更糟糕的是，Nvidia 恰好處於受益於人工智能 GPU 稀缺的有利位置。儘管全球領先的芯片製造公司 <strong>台積電</strong> 努力迅速擴大其芯片 - 晶圓 - 基板（CoWoS）能力——CoWoS 對於包裝高帶寬內存以滿足高計算數據中心的需求至關重要——但人工智能 GPU 的需求始終超過其供應。因此，Nvidia 對其人工智能 GPU 的定價溢價達到了 100%（或更高），相較於競爭對手。</p>
</div><p>Nvidia's CUDA software platform ties things into a nice bow. This toolkit, which is used by developers to maximize the performance of their GPUs, as well as to build large language models, has been pivotal in keeping clients within Nvidia's ecosystem of products and services.</p>
<div class="lb-trans"><p>Nvidia 的 CUDA 軟件平台將這一切完美結合。這個工具包被開發者用來最大化其 GPU 的性能，以及構建大型語言模型，對於保持客户在 Nvidia 產品和服務生態系統內至關重要。</p>
</div><h2>Nvidia can be the victim of its own success</h2>
<div class="lb-trans"><h2>Nvidia 可能成為自身成功的受害者</h2>
</div><p>Although the <em>overwhelming</em> consensus among Wall Street analysts is that this artificial intelligence darling is headed higher -- the average analyst price target of $171 points to 50% upside -- two viable paths exist where Nvidia becomes a victim of its own success, and its share price suffers for it.</p>
<div class="lb-trans"><p>儘管華爾街分析師的 <strong>壓倒性</strong> 共識認為這家人工智能寵兒將繼續上漲——平均分析師目標價為 171 美元，意味着 50% 的上漲空間——但存在兩條可行路徑，可能導致 Nvidia 成為自身成功的受害者，其股價因此受損。</p>
</div><p>To begin with, Nvidia's AI-GPUs are considered superior, in terms of their computing capabilities. While short-term demand continues to overwhelm supply for AI-GPUs, the computing superiority of Hopper and Blackwell might dramatically slow down any future upgrade/replacement cycles.</p>
<div class="lb-trans"><p>首先，Nvidia 的人工智能 GPU 在計算能力上被認為是優越的。儘管短期需求繼續壓倒人工智能 GPU 的供應，但 Hopper 和 Blackwell 的計算優勢可能會顯著減緩未來的升級/更換週期。</p>
</div><p>Think of it this way: When <strong>Apple</strong> first came out with the iPhone, consumers and businesses jumped at the chance to upgrade to this revolutionary device. Over time, next-generation versions of the iPhone have featured upgrades, but they've been far less dramatic or impressive, compared to what the original iPhone did for consumers and businesses in 2007. With Nvidia's hardware being pricier and faster than its peers, the impetus for companies to upgrade their AI-data center hardware may be pushed out for years. That's not good news for a company that Wall Street expects to sustain an eye-popping growth rate.</p>
<div class="lb-trans"><p>可以這樣想：當 <strong>蘋果</strong> 首次推出 iPhone 時，消費者和企業紛紛抓住機會升級到這款革命性設備。隨着時間的推移，下一代 iPhone 版本雖然進行了升級，但與 2007 年原始 iPhone 對消費者和企業的影響相比，變化遠不如前。由於 Nvidia 的硬件價格更高且速度更快，企業升級其人工智能數據中心硬件的動力可能會被推遲多年。這對華爾街期望其保持驚人增長率的公司來説並不是好消息。</p>
</div><p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MTI3MDEvZGF0YS1jZW50ZXItZW5naW5lZXItZ3B1LWFydGlmaWNpYWwtaW50ZWxsaWdlbmNlLWFpLWNsb3VkLXNlcnZlci1nZXR0eS5qcGcmdz03MDA" alt="" original-src="https://g.foolcdn.com/image/?url=https%3A%2F%2Fg.foolcdn.com%2Feditorial%2Fimages%2F812701%2Fdata-center-engineer-gpu-artificial-intelligence-ai-cloud-server-getty.jpg&amp;op=resize&amp;w=700"/></p>
<p>Image source: Getty Images.</p>
<div class="lb-trans"><p>圖片來源：Getty Images。</p>
</div><p>The second potential problem for Nvidia is that Wall Street might punish it for simply ceding some of its monopoly like market share in AI-data centers in the coming quarters.</p>
<div class="lb-trans"><p>Nvidia 的第二個潛在問題是，華爾街可能會因為在未來幾個季度放棄其在人工智能數據中心的部分壟斷市場份額而懲罰它。</p>
</div><p>Based on an analysis from semiconductor firm TechInsights, Nvidia accounted for an astounding 98% of the GPUs shipped to enterprise data centers in 2022 <em>and</em> 2023. Even with the likes of <strong>Advanced Micro Devices</strong> ramping up production of its Instinct series AI-accelerating chips, Nvidia is still likely to have commanded a near-monopoly like share of the &#34;brains&#34; of AI data centers last year.</p>
<div class="lb-trans"><p>根據半導體公司 TechInsights 的分析，Nvidia 在 2022 年和 2023 年佔據了企業數據中心出貨 GPU 的驚人 98%。即使像 <strong>超微</strong> 這樣的公司正在加大 Instinct 系列人工智能加速芯片的生產，Nvidia 仍然可能在去年佔據了人工智能數據中心 “頭腦” 的近乎壟斷的市場份額。</p>
</div><p>The issue for Nvidia is that many of its top customers by net sales, which primarily includes members of the &#34;Magnificent Seven,&#34; are internally developing AI-GPUs and AI solutions of their own that can be used in their data centers. Even though these internally developed chips aren't a threat to the computing speed of Hopper or Blackwell, and there are no plans for them to be sold externally, they could easily cause Nvidia to lose out on future data center real estate. These chips are both cheaper and more readily available than Nvidia's hardware.</p>
<div class="lb-trans"><p>Nvidia 面臨的問題是，其許多按淨銷售額計算的頂級客户，主要包括 “七大巨頭” 的成員，正在內部開發自己的人工智能 GPU 和人工智能解決方案，以便在其數據中心中使用。儘管這些內部開發的芯片對 Hopper 或 Blackwell 的計算速度並不構成威脅，並且沒有計劃對外銷售，但它們可能會導致 Nvidia 在未來的數據中心市場中失去機會。這些芯片的價格更便宜且更易獲得。</p>
</div><p>If future upgrade cycles become more drawn out and/or Wall Street's AI darling loses precious data center real estate to its own top customers, what's arguably its biggest competitive advantage -- AI-GPU scarcity -- will disappear. This means less in the way of pricing power and weaker future margins.</p>
<div class="lb-trans"><p>如果未來的升級週期變得更加漫長，或者華爾街的人工智能寵兒失去了寶貴的數據中心市場份額給其頂級客户，那麼它最大的競爭優勢——人工智能 GPU 的稀缺性——將會消失。這意味着定價能力下降，未來的利潤率也會減弱。</p>
</div><p>Nvidia's historic ascent has left no room for error. If the company's growth rate were to slow, or even if its gross margin continues to decline from a peak of 78.4%, it could quickly become a victim of its own success.</p>
<div class="lb-trans"><p>英偉達的歷史性上升已沒有犯錯的餘地。如果公司的增長率放緩，甚至如果其毛利率繼續從 78.4% 的高峰下降，它可能很快就會成為自身成功的犧牲品。</p>
</div>

Is Artificial Intelligence (AI) GPU Giant Nvidia Poised to Be a Victim of Its Own Success?

- 人工智能的崛起可能在 2030 年前改變全球經濟增長軌跡，影響達 15.7 萬億美元。  
- Nvidia 在 AI-GPU 市場佔據主導地位，但面臨升級週期延長和市場份額下降的風險。  
- 如果增長放緩或毛利率下降，Nvidia 可能會成為自身成功的犧牲品。  

人工智能（AI）GPU 巨頭英偉達是否會因自身的成功而受害？