<p>在过去两年中，没有哪个头条新闻比人工智能（AI）的崛起更能主导华尔街。赋予软件和系统自主推理和行动的能力，使得这项技术似乎拥有无限的长期潜力。</p>
<p>尽管人工智能解决方案的采用率和实用性仍在变化，但普华永道的分析师认为，这项颠覆性技术将在 2030 年前彻底改变全球增长轨迹。到十年之交，普华永道的《Sizing the Prize》报告预计，人工智能将对国内生产总值产生 15.7 万亿美元的积极影响。</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>图片来源：Getty Images。</p>
<p>15.7 万亿美元的蛋糕足以容纳一长串赢家。然而，没有哪家公司比图形处理单元（GPU）巨头 <strong>Nvidia</strong>（NVDA -1.51%）更直接受益于人工智能革命。Nvidia 的业务从 2022 年底的 3600 亿美元增长到两年后接近 4 万亿美元的市值。</p>
<p>尽管 Nvidia 无疑是人工智能的成功故事，但这家领先企业很可能很快会成为自身成功的受害者。</p>
<h2>Nvidia 的崛起简直是教科书般的</h2>
<p>在连接 Nvidia 完美运营扩展如何可能自我解构之前，了解它如何成为人工智能革命基础的背景是很重要的。</p>
<p>Nvidia 成功的核心在于其 Hopper（H100）和下一代 Blackwell GPU。前者在过去两年中有效地成为了人工智能加速数据中心的标准。同时，最近推出的 Blackwell GPU 架构显著提高了生成性人工智能解决方案和量子计算的计算能力，同时比其前身芯片更具能效。</p>
<p>为了进一步说明这一点，目前没有其他芯片制造商能够接近 Nvidia 硬件的全面功能。虽然其他公司可能在某些任务上提供狭窄的竞争优势，但没有其他 GPU 在人工智能加速数据中心的计算性能上真正挑战 Hopper 或 Blackwell。</p>
<p>更糟糕的是，Nvidia 恰好处于受益于人工智能 GPU 稀缺的有利位置。尽管全球领先的芯片制造公司 <strong>台积电</strong> 努力迅速扩大其芯片 - 晶圆 - 基板（CoWoS）能力——CoWoS 对于包装高带宽内存以满足高计算数据中心的需求至关重要——但人工智能 GPU 的需求始终超过其供应。因此，Nvidia 对其人工智能 GPU 的定价溢价达到了 100%（或更高），相较于竞争对手。</p>
<p>Nvidia 的 CUDA 软件平台将这一切完美结合。这个工具包被开发者用来最大化其 GPU 的性能，以及构建大型语言模型，对于保持客户在 Nvidia 产品和服务生态系统内至关重要。</p>
<h2>Nvidia 可能成为自身成功的受害者</h2>
<p>尽管华尔街分析师的 <strong>压倒性</strong> 共识认为这家人工智能宠儿将继续上涨——平均分析师目标价为 171 美元，意味着 50% 的上涨空间——但存在两条可行路径，可能导致 Nvidia 成为自身成功的受害者，其股价因此受损。</p>
<p>首先，Nvidia 的人工智能 GPU 在计算能力上被认为是优越的。尽管短期需求继续压倒人工智能 GPU 的供应，但 Hopper 和 Blackwell 的计算优势可能会显著减缓未来的升级/更换周期。</p>
<p>可以这样想：当 <strong>苹果</strong> 首次推出 iPhone 时，消费者和企业纷纷抓住机会升级到这款革命性设备。随着时间的推移，下一代 iPhone 版本虽然进行了升级，但与 2007 年原始 iPhone 对消费者和企业的影响相比，变化远不如前。由于 Nvidia 的硬件价格更高且速度更快，企业升级其人工智能数据中心硬件的动力可能会被推迟多年。这对华尔街期望其保持惊人增长率的公司来说并不是好消息。</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>图片来源：Getty Images。</p>
<p>Nvidia 的第二个潜在问题是，华尔街可能会因为在未来几个季度放弃其在人工智能数据中心的部分垄断市场份额而惩罚它。</p>
<p>根据半导体公司 TechInsights 的分析，Nvidia 在 2022 年和 2023 年占据了企业数据中心出货 GPU 的惊人 98%。即使像 <strong>超微</strong> 这样的公司正在加大 Instinct 系列人工智能加速芯片的生产，Nvidia 仍然可能在去年占据了人工智能数据中心 “头脑” 的近乎垄断的市场份额。</p>
<p>Nvidia 面临的问题是，其许多按净销售额计算的顶级客户，主要包括 “七大巨头” 的成员，正在内部开发自己的人工智能 GPU 和人工智能解决方案，以便在其数据中心中使用。尽管这些内部开发的芯片对 Hopper 或 Blackwell 的计算速度并不构成威胁，并且没有计划对外销售，但它们可能会导致 Nvidia 在未来的数据中心市场中失去机会。这些芯片的价格更便宜且更易获得。</p>
<p>如果未来的升级周期变得更加漫长，或者华尔街的人工智能宠儿失去了宝贵的数据中心市场份额给其顶级客户，那么它最大的竞争优势——人工智能 GPU 的稀缺性——将会消失。这意味着定价能力下降，未来的利润率也会减弱。</p>
<p>英伟达的历史性上升已没有犯错的余地。如果公司的增长率放缓，甚至如果其毛利率继续从 78.4% 的高峰下降，它可能很快就会成为自身成功的牺牲品。</p>

英伟达

<p>英伟达作为领先的 GPU 制造商，因人工智能（AI）的崛起而显著受益，市值从 3600 亿美元增长到近 4 万亿美元，仅在两年多的时间里。然而，分析师警告称，英伟达可能会因自身的成功而面临挑战。其 AI-GPU 的优势可能会减缓未来的升级周期，类似于 iPhone 的演变，而潜在的市场份额损失可能导致华尔街的负面反应。尽管前景乐观，英伟达的增长可能会受到这些因素的制约</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>在过去两年中，没有哪个头条新闻比人工智能（AI）的崛起更能主导华尔街。赋予软件和系统自主推理和行动的能力，使得这项技术似乎拥有无限的长期潜力。</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>尽管人工智能解决方案的采用率和实用性仍在变化，但普华永道的分析师认为，这项颠覆性技术将在 2030 年前彻底改变全球增长轨迹。到十年之交，普华永道的《Sizing the Prize》报告预计，人工智能将对国内生产总值产生 15.7 万亿美元的积极影响。</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>图片来源：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 万亿美元的蛋糕足以容纳一长串赢家。然而，没有哪家公司比图形处理单元（GPU）巨头 <strong>Nvidia</strong>（NVDA -1.51%）更直接受益于人工智能革命。Nvidia 的业务从 2022 年底的 3600 亿美元增长到两年后接近 4 万亿美元的市值。</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>尽管 Nvidia 无疑是人工智能的成功故事，但这家领先企业很可能很快会成为自身成功的受害者。</p>
</div><h2>Nvidia's ascent has been nothing short of textbook</h2>
<div class="lb-trans"><h2>Nvidia 的崛起简直是教科书般的</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>在连接 Nvidia 完美运营扩展如何可能自我解构之前，了解它如何成为人工智能革命基础的背景是很重要的。</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 成功的核心在于其 Hopper（H100）和下一代 Blackwell GPU。前者在过去两年中有效地成为了人工智能加速数据中心的标准。同时，最近推出的 Blackwell GPU 架构显著提高了生成性人工智能解决方案和量子计算的计算能力，同时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>为了进一步说明这一点，目前没有其他芯片制造商能够接近 Nvidia 硬件的全面功能。虽然其他公司可能在某些任务上提供狭窄的竞争优势，但没有其他 GPU 在人工智能加速数据中心的计算性能上真正挑战 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 恰好处于受益于人工智能 GPU 稀缺的有利位置。尽管全球领先的芯片制造公司 <strong>台积电</strong> 努力迅速扩大其芯片 - 晶圆 - 基板（CoWoS）能力——CoWoS 对于包装高带宽内存以满足高计算数据中心的需求至关重要——但人工智能 GPU 的需求始终超过其供应。因此，Nvidia 对其人工智能 GPU 的定价溢价达到了 100%（或更高），相较于竞争对手。</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 软件平台将这一切完美结合。这个工具包被开发者用来最大化其 GPU 的性能，以及构建大型语言模型，对于保持客户在 Nvidia 产品和服务生态系统内至关重要。</p>
</div><h2>Nvidia can be the victim of its own success</h2>
<div class="lb-trans"><h2>Nvidia 可能成为自身成功的受害者</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>尽管华尔街分析师的 <strong>压倒性</strong> 共识认为这家人工智能宠儿将继续上涨——平均分析师目标价为 171 美元，意味着 50% 的上涨空间——但存在两条可行路径，可能导致 Nvidia 成为自身成功的受害者，其股价因此受损。</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 在计算能力上被认为是优越的。尽管短期需求继续压倒人工智能 GPU 的供应，但 Hopper 和 Blackwell 的计算优势可能会显著减缓未来的升级/更换周期。</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>可以这样想：当 <strong>苹果</strong> 首次推出 iPhone 时，消费者和企业纷纷抓住机会升级到这款革命性设备。随着时间的推移，下一代 iPhone 版本虽然进行了升级，但与 2007 年原始 iPhone 对消费者和企业的影响相比，变化远不如前。由于 Nvidia 的硬件价格更高且速度更快，企业升级其人工智能数据中心硬件的动力可能会被推迟多年。这对华尔街期望其保持惊人增长率的公司来说并不是好消息。</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>图片来源：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 的第二个潜在问题是，华尔街可能会因为在未来几个季度放弃其在人工智能数据中心的部分垄断市场份额而惩罚它。</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>根据半导体公司 TechInsights 的分析，Nvidia 在 2022 年和 2023 年占据了企业数据中心出货 GPU 的惊人 98%。即使像 <strong>超微</strong> 这样的公司正在加大 Instinct 系列人工智能加速芯片的生产，Nvidia 仍然可能在去年占据了人工智能数据中心 “头脑” 的近乎垄断的市场份额。</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 面临的问题是，其许多按净销售额计算的顶级客户，主要包括 “七大巨头” 的成员，正在内部开发自己的人工智能 GPU 和人工智能解决方案，以便在其数据中心中使用。尽管这些内部开发的芯片对 Hopper 或 Blackwell 的计算速度并不构成威胁，并且没有计划对外销售，但它们可能会导致 Nvidia 在未来的数据中心市场中失去机会。这些芯片的价格更便宜且更易获得。</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>如果未来的升级周期变得更加漫长，或者华尔街的人工智能宠儿失去了宝贵的数据中心市场份额给其顶级客户，那么它最大的竞争优势——人工智能 GPU 的稀缺性——将会消失。这意味着定价能力下降，未来的利润率也会减弱。</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>英伟达的历史性上升已没有犯错的余地。如果公司的增长率放缓，甚至如果其毛利率继续从 78.4% 的高峰下降，它可能很快就会成为自身成功的牺牲品。</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 年前改变全球经济增长轨迹，影响达 15.7 万亿美元。  
- Nvidia 在 AI-GPU 市场占据主导地位，但面临升级周期延长和市场份额下降的风险。  
- 如果增长放缓或毛利率下降，Nvidia 可能会成为自身成功的牺牲品。  

人工智能（AI）GPU 巨头英伟达是否会因自身的成功而受害？