<p>有人说数据是新的石油。这在人工智能（AI）领域尤其真实，因为不断增加的计算成本是增长的主要瓶颈。正如许多读者所知，<strong>Nvidia</strong>（NVDA 2.51%）从数据中心计算日益重要中受益匪浅。这家计算机芯片设计公司现在是全球市值最大的公司，上个季度其收入达到了 467 亿美元，同比增长 56%。</p>
<p>大型数据中心建设者如 <strong>Meta Platforms</strong>（META -0.12%）每年在 Nvidia 的芯片上花费数百亿美元。为了降低这些成本，Meta 刚刚收购了一家名为 Rivos 的 AI 计算机芯片初创公司。这是否意味着 Nvidia 的一个主要客户瞬间变成了竞争对手？答案比看起来要复杂得多。</p>
<h2>定制半导体以降低成本</h2>
<p>Meta 计划在今年花费约 700 亿美元用于资本支出，并在 2026 年投入更多资金，Meta 正在投入大量资金以满足其对 Nvidia 芯片的巨大需求，以训练和支持 AI 系统。Meta 早已与 Rivos 就使用 RISC-V 架构（一个开源芯片设计平台）构建芯片进行谈判。现在，在收购之后，Rivos 将专注于为 Meta 内部使用构建 AI 芯片。</p>
<p>这可能会影响 Nvidia 的主导地位。不仅可能减少其出货量，还可能削弱公司的持续定价能力，这使得其毛利率在过去 12 个月中达到了 70%，息税前利润（EBIT）率达到了 58%。定价能力是 Nvidia 现在拥有全球最大市值的一个重要原因。这一趋势的逆转可能会给其股东带来痛苦。</p>
<p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MzY4MjgvbnZpZGlhLWhlYWRxdWFydGVycy13aXRoLW52aWRpYS1zaWduLWluLWZyb250LnBuZyZ3PTcwMA?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-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MzY4MjgvbnZpZGlhLWhlYWRxdWFydGVycy13aXRoLW52aWRpYS1zaWduLWluLWZyb250LnBuZyZ3PTcwMA"/></p>
<p>图片来源：Nvidia</p>
<h2>Nvidia 不会在一夜之间被颠覆</h2>
<p>Nvidia 的股东应密切关注客户试图自行构建 AI 芯片以减轻其在 AI 数据中心市场的控制力所带来的竞争加剧。相反，Meta 的投资者应密切关注其设计自有计算机芯片并投入生产的进展，因为这可能在长期内为业务带来巨大的成本节省。</p>
<p>无论发生什么，Nvidia 不会在一夜之间被颠覆。设计和验证一款芯片可能需要整整一年或更长时间，而后还需在代工厂排队生产。即便如此，芯片设计者的首个版本通常也比更成熟竞争对手的最新产品落后数年。Nvidia 不会坐以待毙，任由竞争对手追赶。</p>
<p><img src="https://imageproxy.pbkrs.com/https://media.ycharts.com/charts/38a2e3e1c0dce84039e100890a7b37d3.png?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://media.ycharts.com/charts/38a2e3e1c0dce84039e100890a7b37d3.png"/></p>
<p>NVDA 毛利率数据来源于 YCharts。</p>
<h2>一家 AI 公司比 Meta 更早考虑</h2>
<p>Meta 开始设计自己的芯片是明智的。Nvidia 的税负很大，并且短期内不会消失。即使 Meta 需要数年和数十亿美元的投资，通过垂直整合的芯片设计每年节省数十亿美元的成本将是非常值得的。</p>
<p>一家多年前就学到了这个教训的公司是 <strong>Alphabet</strong>。作为 Google 和 YouTube 的母公司，Alphabet 在 2015 年初开始使用其首个张量处理单元（TPU），并自那时以来逐步扩大这些计算机芯片在其数据中心的使用。这是 Nvidia 客户实际生产出与其产品在成本效益上相当的计算机芯片的唯一例子。与其他 AI 参与者相比，Alphabet 现在因其 TPU 而拥有竞争优势。</p>
<p>即便如此，10 年后 Alphabet 仍然是 Nvidia 的主要客户。以 TPU 为案例，Meta 收购 Rivos 不太可能会杀死 Nvidia 的业务。不要以此作为卖出股票的理由。</p>

META

<p>Meta Platforms 已收购人工智能芯片初创公司 Rivos，以减少对 Nvidia 芯片的依赖，这可能将一个主要客户转变为竞争对手。这一举措可能会影响 Nvidia 的定价能力和市场主导地位，因为 Meta 计划在定制半导体上进行大量投资。然而，过渡将需要时间，Nvidia 不太可能立即受到冲击。Alphabet 的例子表明，尽管其在仍依赖 Nvidia 的同时开发了自己的芯片，但 Meta 的战略可能会带来成本节约，但并不会消除 Nvidia 的业务</p>

<p>Some say data is the new oil. This is especially true in the artificial intelligence (AI) space, where increased computing costs are the primary bottleneck on growth. As many readers will know, <strong>Nvidia</strong>(NVDA 2.51%) has benefited tremendously from the growing importance of data center computing. The computer chip designer is now the largest company in the world by market capitalization, and the $46.7 billion in revenue it posted last quarter was up 56% year over year.</p>
<div class="lb-trans"><p>有人说数据是新的石油。这在人工智能（AI）领域尤其真实，因为不断增加的计算成本是增长的主要瓶颈。正如许多读者所知，<strong>Nvidia</strong>（NVDA 2.51%）从数据中心计算日益重要中受益匪浅。这家计算机芯片设计公司现在是全球市值最大的公司，上个季度其收入达到了 467 亿美元，同比增长 56%。</p>
</div><p>Large data center builders such as <strong>Meta Platforms</strong>(META -0.12%) are now spending tens of billions of dollars on Nvidia's chips every year. In a bid to cut down on these costs, Meta just acquired an AI computer chip start-up called Rivos. Did this just turn one of Nvidia's top customers into a competitor overnight? The answer is more complicated than it seems.</p>
<div class="lb-trans"><p>大型数据中心建设者如 <strong>Meta Platforms</strong>（META -0.12%）每年在 Nvidia 的芯片上花费数百亿美元。为了降低这些成本，Meta 刚刚收购了一家名为 Rivos 的 AI 计算机芯片初创公司。这是否意味着 Nvidia 的一个主要客户瞬间变成了竞争对手？答案比看起来要复杂得多。</p>
</div><h2>Custom semiconductors to decrease costs</h2>
<div class="lb-trans"><h2>定制半导体以降低成本</h2>
</div><p>With plans to spend around $70 billion on capital expenditures this year and even more in 2026, Meta is putting considerable sums toward feeding its insatiable need for Nvidia chips to train and power AI systems. Meta had already been in talks with Rivos about building chips using the RISC-V architecture, an open-source chip design platform. Now, after its acquisition, Rivos will focus on building AI chips for Meta's internal use.</p>
<div class="lb-trans"><p>Meta 计划在今年花费约 700 亿美元用于资本支出，并在 2026 年投入更多资金，Meta 正在投入大量资金以满足其对 Nvidia 芯片的巨大需求，以训练和支持 AI 系统。Meta 早已与 Rivos 就使用 RISC-V 架构（一个开源芯片设计平台）构建芯片进行谈判。现在，在收购之后，Rivos 将专注于为 Meta 内部使用构建 AI 芯片。</p>
</div><p>This could impact Nvidia's dominance. Not only could it decrease its unit volume, but it could eliminate the company's sustained pricing power, which has sent gross profit margins up to 70% and EBIT (earnings before interest and taxes) margins up to 58% over the last 12 months. Pricing power is a huge reason why Nvidia now has the largest market cap in the world. A reversal of this trend could mean pain for its shareholders.</p>
<div class="lb-trans"><p>这可能会影响 Nvidia 的主导地位。不仅可能减少其出货量，还可能削弱公司的持续定价能力，这使得其毛利率在过去 12 个月中达到了 70%，息税前利润（EBIT）率达到了 58%。定价能力是 Nvidia 现在拥有全球最大市值的一个重要原因。这一趋势的逆转可能会给其股东带来痛苦。</p>
</div><p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MzY4MjgvbnZpZGlhLWhlYWRxdWFydGVycy13aXRoLW52aWRpYS1zaWduLWluLWZyb250LnBuZyZ3PTcwMA" alt="" original-src="https://g.foolcdn.com/image/?url=https%3A%2F%2Fg.foolcdn.com%2Feditorial%2Fimages%2F836828%2Fnvidia-headquarters-with-nvidia-sign-in-front.png&amp;op=resize&amp;w=700"/></p>
<p>Image source: Nvidia</p>
<div class="lb-trans"><p>图片来源：Nvidia</p>
</div><h2>Nvidia will not get disrupted overnight</h2>
<div class="lb-trans"><h2>Nvidia 不会在一夜之间被颠覆</h2>
</div><p>Nvidia shareholders should keep a close eye on increased competition from customers trying to build their own AI chips to ease its choke hold on the AI data center market. Conversely, investors in Meta should keep close track of its progress in designing its own computer chips and bringing them into production, as that could result in huge cost savings for the business over the long term.</p>
<div class="lb-trans"><p>Nvidia 的股东应密切关注客户试图自行构建 AI 芯片以减轻其在 AI 数据中心市场的控制力所带来的竞争加剧。相反，Meta 的投资者应密切关注其设计自有计算机芯片并投入生产的进展，因为这可能在长期内为业务带来巨大的成本节省。</p>
</div><p>Regardless of what happens, Nvidia will not get disrupted overnight. It can take a full year or longer to design and validate a chip, which then has to wait in line at a foundry to get manufactured. Even then, first iterations for chip designers are usually years behind the latest offerings from more-established competitors. It is not like Nvidia is going to sit tight and just wait for the competition to catch up.</p>
<div class="lb-trans"><p>无论发生什么，Nvidia 不会在一夜之间被颠覆。设计和验证一款芯片可能需要整整一年或更长时间，而后还需在代工厂排队生产。即便如此，芯片设计者的首个版本通常也比更成熟竞争对手的最新产品落后数年。Nvidia 不会坐以待毙，任由竞争对手追赶。</p>
</div><p><img src="https://imageproxy.pbkrs.com/https://media.ycharts.com/charts/38a2e3e1c0dce84039e100890a7b37d3.png" alt="" original-src="https://media.ycharts.com/charts/38a2e3e1c0dce84039e100890a7b37d3.png"/></p>
<p>NVDA Gross Profit Margin data by YCharts.</p>
<div class="lb-trans"><p>NVDA 毛利率数据来源于 YCharts。</p>
</div><h2>One AI company thought far ahead of Meta</h2>
<div class="lb-trans"><h2>一家 AI 公司比 Meta 更早考虑</h2>
</div><p>Meta is smart to start designing its own chips. The Nvidia tax is large and not going away anytime soon. Even if it takes Meta years and billions of dollars in investment, the tens of billions of dollars it could eventually save every year through vertically integrated chip design will make it well worth it.</p>
<div class="lb-trans"><p>Meta 开始设计自己的芯片是明智的。Nvidia 的税负很大，并且短期内不会消失。即使 Meta 需要数年和数十亿美元的投资，通过垂直整合的芯片设计每年节省数十亿美元的成本将是非常值得的。</p>
</div><p>One company learned this lesson many years ago: <strong>Alphabet</strong>. The parent of Google and YouTube started using its first tensor processing units (TPUs) in early 2015, and has slowly been scaling up the usage of these computer chips in its own data centers ever since. This is the only example of an Nvidia customer actually producing computer chips that are close to equivalent to its wares in terms of cost-effectiveness. Compared to every other AI player, Alphabet now has a competitive advantage because of its TPUs.</p>
<div class="lb-trans"><p>一家多年前就学到了这个教训的公司是 <strong>Alphabet</strong>。作为 Google 和 YouTube 的母公司，Alphabet 在 2015 年初开始使用其首个张量处理单元（TPU），并自那时以来逐步扩大这些计算机芯片在其数据中心的使用。这是 Nvidia 客户实际生产出与其产品在成本效益上相当的计算机芯片的唯一例子。与其他 AI 参与者相比，Alphabet 现在因其 TPU 而拥有竞争优势。</p>
</div><p>Even so, Alphabet remains a major customer of Nvidia 10 years later. Taking the TPU as a case study, it is highly unlikely that Meta's acquisition of Rivos is going to kill Nvidia's business. Don't use it as a reason to sell the stock.</p>
<div class="lb-trans"><p>即便如此，10 年后 Alphabet 仍然是 Nvidia 的主要客户。以 TPU 为案例，Meta 收购 Rivos 不太可能会杀死 Nvidia 的业务。不要以此作为卖出股票的理由。</p>
</div>

Did Meta Platforms Just Say Checkmate to Nvidia?

- 数据在人工智能领域的重要性日益增加，Nvidia 受益显著。  
- Meta 收购 Rivos 以降低对 Nvidia 芯片的依赖，可能影响 Nvidia 的市场主导地位。  
- 尽管竞争加剧，Nvidia 短期内不会被颠覆，仍需关注其市场表现。

Meta Platforms 刚刚对 Nvidia 宣布胜利了吗？