<p>预计到 2025 年，全球对人工智能（AI）的支出将保持强劲，市场研究公司 IDC 估计，组织可能会投入 3370 亿美元用于这一技术，因为它们将 AI 工具整合到业务运营中。</p>
<p>值得注意的是，IDC 估计，今年预计的 AI 支出中有 67% 将来自企业将 AI 能力嵌入其核心业务运营，超过对领先云和数字服务提供商的投资。同时，云计算巨头将继续在构建其 AI 基础设施上投入巨资。</p>
<p><strong>Nvidia</strong> 在 AI 热潮中处于前沿，其强大的图形处理单元（GPU）帮助主要科技巨头训练大型语言模型（LLM）。芯片制造商很可能会继续见证其收入和盈利的显著增长。</p>
<p>然而，我们将关注另外两个也有望从 AI 在不同领域的普及中受益的公司，看看它们为何可能成为 2025 年的顶级 AI 投资选择。</p>
<h2>博通：值得购买的稳健 AI 芯片股票</h2>
<p>虽然 Nvidia 在 AI GPU 市场占据主导地位，但 <strong>博通</strong>（AVGO 1.84%）在应用特定集成电路（ASIC）方面处于领先地位。这些是可以编程以执行特定任务的定制芯片，与可以执行多种任务并用于通用计算目的的 GPU 不同。</p>
<p>ASIC 的特定性质使其在能效和处理其设计任务方面更为出色。这解释了为什么云计算巨头正在使用博通的定制 AI 处理器，称为 XPU，来为其数据中心提供动力。</p>
<p>报告显示，博通可能正在为 <strong>Meta Platforms</strong>、<strong>Alphabet</strong> 的谷歌、TikTok 母公司字节跳动和 OpenAI 制造芯片。现在有消息称，<strong>苹果</strong> 也可能利用博通制造其 AI 服务器芯片。</p>
<p>博通管理层在 2024 年 12 月的财报电话会议上指出，它正在向三家云超大规模客户提供定制 AI 芯片。该公司表示，这些客户计划在未来三年内部署其定制 AI 芯片集群达 100 万个。此外，博通表示，它已被另外两家超大规模客户选中，[他们] 正在为自己的下一代 AI XPU 进行高级开发。</p>
<p>这一不断扩大的 AI 客户基础可能会在未来三年为博通带来卓越的财务增长。这是因为该芯片制造商预计其在 AI 领域的可寻址市场将在 2027 财年达到 600 亿到 900 亿美元。相比之下，博通在刚刚结束的 2024 财年的 AI 收入为 122 亿美元，较前一年大幅增长 220%。</p>
<p>可寻址市场的规模表明，AI 可能会在 2025 年及以后大幅推动博通的增长，尤其是考虑到根据 <strong>摩根大通</strong> 的估计，它控制了约 55% 到 60% 的 ASIC 市场。毫不奇怪，分析师们已提高了对当前及未来两年财年的收入预期。</p>
<p><img src="https://imageproxy.pbkrs.com/https://media.ycharts.com/charts/b73c01fa7f45a5dfee85a68a3ae7e994.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/b73c01fa7f45a5dfee85a68a3ae7e994.png"/></p>
<p>AVGO 当前财年收入预估数据来自 YCharts</p>
<p>然而，博通可能能够实现比分析师预期更强劲的增长，尤其是在潜在新增客户的加入。此外，根据 Yahoo! Finance 的数据，其市盈率增长比（PEG 比率）目前仅为 0.7，基于其未来五年的预计盈利增长。</p>
<p>PEG 比率低于 1 意味着一只股票相对于其预期的盈利增长被低估，这表明即使在过去一年实现了 108% 的显著涨幅，成长型投资者仍然可以考虑购买博通。</p>
<h2>Snowflake：将 AI 应用于企业数据将成为公司的助力</h2>
<p>虽然博通和 Nvidia 等公司生产的芯片使主要云计算公司能够训练 LLM，但 <strong>Snowflake</strong>（SNOW -0.47%）为客户提供了将数据整合到单一平台的途径。客户可以利用这些数据构建生成式 AI 应用，深入了解数据，或通过将 AI 应用于数据来解决与业务相关的挑战。</p>
<p>因此，考虑到 IDC 预计 AI 支出的很大一部分将用于将 AI 能力嵌入企业运营，Snowflake 在 2025 年有着巨大的可寻址市场。好消息是，Snowflake 的 AI 解决方案已经在客户中获得了良好的反响。</p>
<p>例如，该公司的 Cortex AI 平台 “显示出显著的采用率”，管理层表示。客户可以使用 Cortex AI 分析数据并构建生成式 AI 应用，查找文档中的特定信息，并访问流行的 LLM，如 Llama 和 Anthropic Claude。事实证明，Snowflake 的 3200 多名客户已经在使用其 AI 和机器学习（ML）功能。</p>
<p>这令人印象深刻，考虑到 Snowflake 在 2023 年中期开始推出其 AI 产品，并持续增加更多服务，以便为用户提供更广泛的工具。</p>
<p>该公司还通过 Snowflake Intelligence 平台进入了快速增长的 AI 代理市场，该平台于去年 11 月宣布。该平台将使企业能够创建 AI 代理，这些代理可以在安全环境中分析和总结数据，然后根据这些见解采取行动。</p>
<p>预计到 2025 年，代理 AI 市场的收入将达到 450 亿美元，Snowflake 的 AI 聚焦可寻址机会理应扩大。因此，该公司有很大机会在未来实现超出预期的增长。</p>
<p>毕竟，Snowflake 的剩余业绩义务（RPO）在 2025 财年第三季度（截至 2024 年 10 月 31 日）激增了令人印象深刻的 55%，达到了 57 亿美元。这一增幅显著高于其季度收入同比增长 29%，达到 9.003 亿美元。RPO 指的是公司尚未履行的合同，这些合同将在未来执行时被确认收入。因此，与 Snowflake 的收入增长相比，这一指标的更快增长预示着潜在的增长加速。</p>
<p>现有客户在其平台上的支出增加，季度净收入留存率达到了 127%。该指标比较了 Snowflake 客户在一个季度的支出与这些客户在前一年同期的支出，因此超过 100% 的读数表明现有客户对其服务的采用有所增长。</p>
<p>随着越来越多的 Snowflake 客户开始使用其人工智能产品，这一趋势可能会持续。到上一个季度末，Snowflake 拥有超过 10,600 名客户，其中 30% 正在使用其人工智能/机器学习解决方案。因此，随着公司向更多客户推销其人工智能产品，可能会在收入和盈利方面看到显著改善。</p>
<p>所有这些都解释了为什么 Snowflake 预计在未来几个财年将实现超过 40% 的盈利增长。</p>
<p><img src="https://imageproxy.pbkrs.com/https://media.ycharts.com/charts/c4ee0190d540cc9c800804749fcadef2.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/c4ee0190d540cc9c800804749fcadef2.png"/></p>
<p>SNOW 当前财年每股收益预估数据来自 YCharts</p>
<p>因此，考虑到 Snowflake 的可服务市场预计将在 2025 年因其数据云平台中增加以人工智能为中心的工具而得到良好扩展，Snowflake 看起来是一个值得购买的顶级人工智能股票，这将为未来的卓越盈利增长奠定基础。</p>

博通

SNOW

<p>全球人工智能支出预计将在 2025 年达到 3370 亿美元，其中 67% 来自企业将人工智能整合到运营中。虽然 Nvidia 在人工智能 GPU 领域处于领先地位，但博通在 ASIC 方面表现出色，为 Meta 和 Google 等主要公司提供定制的人工智能芯片，预计到 2027 年市场规模将增长至 900 亿美元。Snowflake 也处于良好位置，提供用于数据整合和应用开发的人工智能解决方案。预计这两家公司将在未来几年中从人工智能的繁荣中显著受益</p>

<p>Global spending on artificial intelligence (AI) is expected to remain solid in 2025, with market research firm IDC estimating that organizations are likely to pour $337 billion into this technology as they integrate AI tools into their business operations.</p>
<div class="lb-trans"><p>预计到 2025 年，全球对人工智能（AI）的支出将保持强劲，市场研究公司 IDC 估计，组织可能会投入 3370 亿美元用于这一技术，因为它们将 AI 工具整合到业务运营中。</p>
</div><p>Notably, IDC estimates that 67% of the projected AI spending this year &#34;will come from enterprises embedding AI capabilities into their core business operations, surpassing investments in leading cloud and digital service providers.&#34; At the same time, cloud computing giants are set to continue spending big money on building out their AI infrastructure.</p>
<div class="lb-trans"><p>值得注意的是，IDC 估计，今年预计的 AI 支出中有 67% 将来自企业将 AI 能力嵌入其核心业务运营，超过对领先云和数字服务提供商的投资。同时，云计算巨头将继续在构建其 AI 基础设施上投入巨资。</p>
</div><p><strong>Nvidia</strong> has been at the forefront of the AI boom with its powerful graphics processing units (GPUs) that have helped major tech giants train large language models (LLMs). And there is a good chance that the chipmaker will continue to witness impressive growth in its revenue and earnings.</p>
<div class="lb-trans"><p><strong>Nvidia</strong> 在 AI 热潮中处于前沿，其强大的图形处理单元（GPU）帮助主要科技巨头训练大型语言模型（LLM）。芯片制造商很可能会继续见证其收入和盈利的显著增长。</p>
</div><p>However, we will take a look at two other names that are also on track to benefit from the proliferation of AI in different verticals and see why they can turn out to be top AI picks for 2025.</p>
<div class="lb-trans"><p>然而，我们将关注另外两个也有望从 AI 在不同领域的普及中受益的公司，看看它们为何可能成为 2025 年的顶级 AI 投资选择。</p>
</div><h2>Broadcom: A solid AI chip stock to buy</h2>
<div class="lb-trans"><h2>博通：值得购买的稳健 AI 芯片股票</h2>
</div><p>While Nvidia is dominating the market for AI GPUs, <strong>Broadcom</strong>(AVGO 1.84%) is leading the charge in application-specific integrated circuits (ASICs). These are custom chips that can be programmed to perform specific tasks, unlike GPUs that can perform multiple tasks and are used for general computing purposes.</p>
<div class="lb-trans"><p>虽然 Nvidia 在 AI GPU 市场占据主导地位，但 <strong>博通</strong>（AVGO 1.84%）在应用特定集成电路（ASIC）方面处于领先地位。这些是可以编程以执行特定任务的定制芯片，与可以执行多种任务并用于通用计算目的的 GPU 不同。</p>
</div><p>The specific nature of ASICs makes them more energy-efficient and adept at tackling the tasks they are designed to do. This explains why cloud computing giants are using Broadcom's custom AI processors, known as XPUs, to power their data centers.</p>
<div class="lb-trans"><p>ASIC 的特定性质使其在能效和处理其设计任务方面更为出色。这解释了为什么云计算巨头正在使用博通的定制 AI 处理器，称为 XPU，来为其数据中心提供动力。</p>
</div><p>Reports suggest that Broadcom could be manufacturing chips for <strong>Meta Platforms</strong>, <strong>Alphabet</strong>'s Google, TikTok parent ByteDance, and OpenAI. It has now emerged that even <strong>Apple</strong> could tap Broadcom to manufacture its AI server chip.</p>
<div class="lb-trans"><p>报告显示，博通可能正在为 <strong>Meta Platforms</strong>、<strong>Alphabet</strong> 的谷歌、TikTok 母公司字节跳动和 OpenAI 制造芯片。现在有消息称，<strong>苹果</strong> 也可能利用博通制造其 AI 服务器芯片。</p>
</div><p>Broadcom management pointed out on the company's December 2024 earnings conference call that it is supplying its custom AI chips to three cloud hyperscale customers. The company notes that these customers are on track to deploy one million of its custom AI chip clusters over the next three years. Additionally, Broadcom says it has &#34;been selected by two additional hyperscalers and [they] are in advanced development for their own next-generation AI XPUs.&#34;</p>
<div class="lb-trans"><p>博通管理层在 2024 年 12 月的财报电话会议上指出，它正在向三家云超大规模客户提供定制 AI 芯片。该公司表示，这些客户计划在未来三年内部署其定制 AI 芯片集群达 100 万个。此外，博通表示，它已被另外两家超大规模客户选中，[他们] 正在为自己的下一代 AI XPU 进行高级开发。</p>
</div><p>This expanding AI customer base could lead to outstanding financial growth for Broadcom over the next three years. That's because the chipmaker sees its addressable market in AI reaching a range of $60 billion to $90 billion in fiscal 2027. For comparison, Broadcom's AI revenue stood at $12.2 billion in the recently concluded fiscal 2024, which was a massive increase of 220% from the preceding year.</p>
<div class="lb-trans"><p>这一不断扩大的 AI 客户基础可能会在未来三年为博通带来卓越的财务增长。这是因为该芯片制造商预计其在 AI 领域的可寻址市场将在 2027 财年达到 600 亿到 900 亿美元。相比之下，博通在刚刚结束的 2024 财年的 AI 收入为 122 亿美元，较前一年大幅增长 220%。</p>
</div><p>The size of the addressable market suggests that AI is likely to supercharge Broadcom's growth in 2025 and beyond, especially considering that it controls an estimated 55% to 60% of the ASIC market, as per <strong>JPMorgan</strong>. Not surprisingly, analysts have raised their revenue estimates for the current and the next two fiscal years.</p>
<div class="lb-trans"><p>可寻址市场的规模表明，AI 可能会在 2025 年及以后大幅推动博通的增长，尤其是考虑到根据 <strong>摩根大通</strong> 的估计，它控制了约 55% 到 60% 的 ASIC 市场。毫不奇怪，分析师们已提高了对当前及未来两年财年的收入预期。</p>
</div><p><img src="https://imageproxy.pbkrs.com/https://media.ycharts.com/charts/b73c01fa7f45a5dfee85a68a3ae7e994.png" alt="" original-src="https://media.ycharts.com/charts/b73c01fa7f45a5dfee85a68a3ae7e994.png"/></p>
<p>AVGO Revenue Estimates for Current Fiscal Year data by YCharts</p>
<div class="lb-trans"><p>AVGO 当前财年收入预估数据来自 YCharts</p>
</div><p>However, Broadcom may be able to deliver stronger growth than analysts expect following the potential addition of new customers. In addition, its price/earnings-to-growth ratio (PEG ratio) stands at just 0.7 right now based on its estimated earnings growth for the next five years, as per Yahoo! Finance.</p>
<div class="lb-trans"><p>然而，博通可能能够实现比分析师预期更强劲的增长，尤其是在潜在新增客户的加入。此外，根据 Yahoo! Finance 的数据，其市盈率增长比（PEG 比率）目前仅为 0.7，基于其未来五年的预计盈利增长。</p>
</div><p>A PEG ratio of less than 1 means that a stock is undervalued with respect to the earnings growth that it is expected to deliver, suggesting that growth investors can still consider buying Broadcom even after the impressive 108% gains that it has delivered in the past year.</p>
<div class="lb-trans"><p>PEG 比率低于 1 意味着一只股票相对于其预期的盈利增长被低估，这表明即使在过去一年实现了 108% 的显著涨幅，成长型投资者仍然可以考虑购买博通。</p>
</div><h2>Snowflake: The application of AI to enterprise data is going to be a tailwind for the company</h2>
<div class="lb-trans"><h2>Snowflake：将 AI 应用于企业数据将成为公司的助力</h2>
</div><p>While the likes of Broadcom and Nvidia produce chips that allow major cloud computing companies to train LLMs, <strong>Snowflake</strong>(SNOW -0.47%) gives customers an avenue to consolidate their data into a single platform. Its customers can then use this data to build generative AI apps, gain insights into their data, or solve business-related challenges by applying AI to that data.</p>
<div class="lb-trans"><p>虽然博通和 Nvidia 等公司生产的芯片使主要云计算公司能够训练 LLM，但 <strong>Snowflake</strong>（SNOW -0.47%）为客户提供了将数据整合到单一平台的途径。客户可以利用这些数据构建生成式 AI 应用，深入了解数据，或通过将 AI 应用于数据来解决与业务相关的挑战。</p>
</div><p>So, Snowflake has a huge addressable market to tap in 2025, considering that IDC is expecting a big chunk of AI spending to go toward embedding AI capabilities in businesses' operations. The good part is that Snowflake's AI-focused solutions are already gaining healthy traction among customers.</p>
<div class="lb-trans"><p>因此，考虑到 IDC 预计 AI 支出的很大一部分将用于将 AI 能力嵌入企业运营，Snowflake 在 2025 年有着巨大的可寻址市场。好消息是，Snowflake 的 AI 解决方案已经在客户中获得了良好的反响。</p>
</div><p>The company's Cortex AI platform, for instance, is &#34;showing significant adoption,&#34; as per management. Customers can use Cortex AI to analyze their data and build generative AI apps, find specific information within documents, and get access to popular LLMs such as Llama and Anthropic Claude. As it turns out, over 3,200 of Snowflake's customers are already using its AI and machine learning (ML) features.</p>
<div class="lb-trans"><p>例如，该公司的 Cortex AI 平台 “显示出显著的采用率”，管理层表示。客户可以使用 Cortex AI 分析数据并构建生成式 AI 应用，查找文档中的特定信息，并访问流行的 LLM，如 Llama 和 Anthropic Claude。事实证明，Snowflake 的 3200 多名客户已经在使用其 AI 和机器学习（ML）功能。</p>
</div><p>That's impressive, considering that Snowflake started rolling out its AI offerings in mid-2023, and it has continued to add more services so it can give users access to a wider range of tools.</p>
<div class="lb-trans"><p>这令人印象深刻，考虑到 Snowflake 在 2023 年中期开始推出其 AI 产品，并持续增加更多服务，以便为用户提供更广泛的工具。</p>
</div><p>The company has also made a move into the fast-growing market for AI agents as well with the Snowflake Intelligence platform, which was announced in November last year. This platform will enable enterprises to create AI agents, which can analyze and summarize data in a secure environment and then take action using those insights.</p>
<div class="lb-trans"><p>该公司还通过 Snowflake Intelligence 平台进入了快速增长的 AI 代理市场，该平台于去年 11 月宣布。该平台将使企业能够创建 AI 代理，这些代理可以在安全环境中分析和总结数据，然后根据这些见解采取行动。</p>
</div><p>With the agentic AI market expected to clock an estimated revenue of $45 billion in 2025, Snowflake's AI-focused addressable opportunity should ideally expand. As a result, there is a solid chance that the company will be able to clock better-than-expected growth going forward.</p>
<div class="lb-trans"><p>预计到 2025 年，代理 AI 市场的收入将达到 450 亿美元，Snowflake 的 AI 聚焦可寻址机会理应扩大。因此，该公司有很大机会在未来实现超出预期的增长。</p>
</div><p>After all, Snowflake's remaining performance obligations (RPO) shot up an impressive 55% in the third quarter of fiscal 2025 (which ended on Oct. 31, 2024) to $5.7 billion. That was significantly higher than the 29% year-over-year increase in its quarterly revenue to $900.3 million. RPO refers to the unfulfilled contracts of a company that will be recognized as revenue in the future when those contracts are executed. So, faster growth in this metric, as compared to Snowflake's top line, points toward a potential acceleration in growth.</p>
<div class="lb-trans"><p>毕竟，Snowflake 的剩余业绩义务（RPO）在 2025 财年第三季度（截至 2024 年 10 月 31 日）激增了令人印象深刻的 55%，达到了 57 亿美元。这一增幅显著高于其季度收入同比增长 29%，达到 9.003 亿美元。RPO 指的是公司尚未履行的合同，这些合同将在未来执行时被确认收入。因此，与 Snowflake 的收入增长相比，这一指标的更快增长预示着潜在的增长加速。</p>
</div><p>Existing customers spent more on its platform, as evidenced by a net revenue retention rate of 127% during the quarter. The metric compares the spending by Snowflake's customers in a quarter to the spending by those same customers in the prior-year period, so a reading of more than 100% indicates that the adoption of its services by existing customers grew.</p>
<div class="lb-trans"><p>现有客户在其平台上的支出增加，季度净收入留存率达到了 127%。该指标比较了 Snowflake 客户在一个季度的支出与这些客户在前一年同期的支出，因此超过 100% 的读数表明现有客户对其服务的采用有所增长。</p>
</div><p>This is a trend that's likely to continue as more Snowflake customers start using its AI offerings. Snowflake had just over 10,600 customers at the end of the previous quarter, and 30% of them are using its AI/ML solutions. So, as the company upsells its AI products to more customers, it could witness a nice improvement in both revenue and earnings.</p>
<div class="lb-trans"><p>随着越来越多的 Snowflake 客户开始使用其人工智能产品，这一趋势可能会持续。到上一个季度末，Snowflake 拥有超过 10,600 名客户，其中 30% 正在使用其人工智能/机器学习解决方案。因此，随着公司向更多客户推销其人工智能产品，可能会在收入和盈利方面看到显著改善。</p>
</div><p>All this explains why Snowflake is expected to deliver 40%-plus earnings growth for the next couple of fiscal years.</p>
<div class="lb-trans"><p>所有这些都解释了为什么 Snowflake 预计在未来几个财年将实现超过 40% 的盈利增长。</p>
</div><p><img src="https://imageproxy.pbkrs.com/https://media.ycharts.com/charts/c4ee0190d540cc9c800804749fcadef2.png" alt="" original-src="https://media.ycharts.com/charts/c4ee0190d540cc9c800804749fcadef2.png"/></p>
<p>SNOW EPS Estimates for Current Fiscal Year data by YCharts</p>
<div class="lb-trans"><p>SNOW 当前财年每股收益预估数据来自 YCharts</p>
</div><p>As such, Snowflake looks like a top AI stock to buy, considering that its addressable market is likely to expand nicely in 2025 thanks to the addition of AI-centric tools in its data cloud platform, which should set the stage for outstanding earnings growth going forward.</p>
<div class="lb-trans"><p>因此，考虑到 Snowflake 的可服务市场预计将在 2025 年因其数据云平台中增加以人工智能为中心的工具而得到良好扩展，Snowflake 看起来是一个值得购买的顶级人工智能股票，这将为未来的卓越盈利增长奠定基础。</p>
</div>

My Top 2 Artificial Intelligence (AI) Stocks for 2025 (Hint: Nvidia Is Not One of Them)

- 预计 2025 年全球人工智能支出将达 3370 亿美元，67% 来自企业核心业务。  
- Broadcom 和 Snowflake 在 AI 领域有显著增长潜力，分别专注于定制芯片和数据整合。  
- Broadcom 的 AI 市场预计到 2027 年可达 600 亿至 900 亿美元，Snowflake 则有望实现 40% 以上的盈利增长。

我在 2025 年最看好的两只人工智能（AI）股票（提示：不包括 Nvidia）