The Internet Is Changing Fast

以下是内容的中文翻译： 这期由乔恩·夸斯特主持，嘉宾为马特·弗兰克尔和泰勒·克劳的“莫特利·富尔隐藏宝石投资”播客，深入探讨了三个不同但同样引人入胜的金融话题：Cloudflare大胆的AI流量预测、美国政府对采矿业的投资以及英特尔的大规模融资。 **Cloudflare的AI代理流量预测** 这一部分以Cloudflare首席执行官马修·普林斯披露的消息开场，他表示5月份其网络上的AI代理活动已提前超越人类活动。普林斯预测，五年内AI代理流量将比人类流量大1000倍，乔恩强调了这一声明所蕴含的深远增长影响。 马特·弗兰克尔承认其基本逻辑，指出代理流量在过去一年增长了18倍，而五年内增长1000倍实际上是这种增速的*放缓*（大约需要每年翻四倍）。他指出，人类互联网流量存在自然上限，而AI代理则随着令牌成本下降和使用量上升而没有上限。 泰勒·克劳则采取了“悲观者”的立场，表示怀疑，暗示鉴于Cloudflare可能从指数级增长的网络流量中受益，普林斯可能是在“为自家公司站台”。他将这种指数级增长比作一分钱不可持续的翻倍，强调了投资领域的S型曲线以及AI代理最终需要证明投资回报率，即使计算成本正在下降。 如果普林斯的预测方向是正确的，互联网可能会经历巨大的变革。马特推测，以广告为中心的互联网商业模式可能会终结，因为如果人类的观看量变得微不足道，广告商将需要新的方式来触达客户。泰勒将其与向移动友好型网页的转变以及“SEO军备竞赛”相提并论，预测未来将出现“AI代理优化型网页”，届时人类可读性将变得次要。 对于投资者而言，马特关注支持互联网流量增长的基础设施公司，例如为数据中心提供电力和冷却服务的公司，但他警告当前的估值过高（援引思科互联网泡沫作为警示）。泰勒虽然承认不确定性，但也承认这支持了人工智能基础设施建设的叙事。他思考了那些能够区分人类和AI流量的公司，甚至那些可以通过新形式的“广告”影响AI代理的公司所蕴含的机会。 **美国政府对采矿业的投资** 讨论随后转向特朗普政府宣布的30亿美元采矿业投资计划，其中包括1亿美元用于教育，旨在减少美国对中国矿产的依赖。乔恩指出，中国每年有3000名采矿工程师毕业，而美国只有170名，并且美国一半的采矿劳动力将在三年内退休。 马特·弗兰克尔对这项教育计划的短期影响表示怀疑。他解释了采矿业的周期性、偏远的工作地点以及过去的动荡（例如2015-2020年），这些因素导致工程专业入学人数大幅下降。此外，四年的学士学位也意味着毕业生无法立即填补眼前的空缺。 泰勒·克劳表达了进一步的疑虑，强调了“邻避效应”（NIMBY，即“不要在我家后院”）现象，他举例说缅因州拥有大量锂矿，但州法律阻止其开采。他透露，30亿美元投资中的14亿美元实际上是美国能源部向一家*私人*锂离子电池初创公司提供的贷款，而非直接的采矿公司，该初创公司专注于硅阳极——一种相对易于开采的材料，而非经常被讨论的稀土。泰勒警告称，那些靠“希望和祈祷”维系的矿业公司将不可避免地出现，他建议投资者避开这些“诱惑”，转而在周期性低迷时期等待成熟的公司。 **英特尔150亿美元融资** 最后，专家小组讨论了英特尔通过出售股权融资150亿美元的决定，此举导致其股价小幅下跌。英特尔给出的理由是资本支出、一般营运资金以及“在新兴领域的进展，包括实体AI……专用芯片、先进封装和外部晶圆”。 马特·弗兰克尔赞扬了这一举动，称其为在股价“昂贵”时融资的明智决定（类似于特斯拉的策略）。他指出，150亿美元的融资稀释了不到3%的股权，并避免了增加更多债务。关键的是，提及“外部晶圆”表明其第三方代工业务正在加速，这是英特尔投资论点的关键所在。鉴于英特尔过去一年股价飙升了395%，且资本支出计划不断升级（今年200亿美元，2027年更高），马特认为现在从强势而非绝望的境地融资更好。 乔恩指出，英特尔当前的市销率（8倍）是其10年平均水平（3倍）的两倍多，这支持了“估值过高”的观点。他还提到标普500指数的股息收益率较低（1.04%），暗示整体市场估值偏高。 马特和泰勒都认为其他公司很可能会进行更多的股权融资。马特引用了最近的例子，如Alphabet的450亿美元股权融资以及美国整体股权融资的增加，这些都受到利用高估值和AI基础设施建设巨大资金需求的双重驱动。泰勒补充说，当前现金流不足以支持支出计划，导致公司不得不举债和进行表外融资。他认为这一趋势将持续下去，直到市场“不堪重负”，这可能由信用评级下调等事件引发，从而迫使公司重新评估其资本支出计划。

This Motley Fool Hidden Gems Investing podcast, hosted by Jon Quast with guests Matt Frankel and Tyler Crowe, delved into three distinct but equally compelling financial topics: Cloudflare's bold AI traffic predictions, the U.S. government's investment in mining, and Intel's significant capital raise. **Cloudflare's AI Agent Traffic Prediction** The segment kicked off with Cloudflare CEO Matthew Prince's revelation that AI agents surpassed human activity on their network in May, ahead of schedule. Prince projected AI agent traffic to be 1,000 times larger than human traffic within five years, a statement Jon highlighted for its profound growth implications. Matt Frankel acknowledged the underlying logic, noting that agentic traffic grew 18x in the past year, and 1,000x in five years would actually be a *deceleration* of that pace (requiring roughly quadrupling annually). He pointed out that human internet traffic has a natural ceiling, while AI agents, with falling token costs and rising usage, do not. Tyler Crowe, adopting a "curmudgeon" stance, expressed skepticism, suggesting Prince might be "talking his book" given Cloudflare's potential benefit from exponential web traffic. He compared the exponential growth to the unsustainable doubling of a penny, emphasizing the S-curve of investing and the eventual need for AI agents to demonstrate ROI despite declining compute costs. If Prince's prediction is directionally correct, the internet could undergo massive changes. Matt speculated on the demise of the advertising-centric internet business model, as advertisers would need new ways to reach customers if human views become negligible. Tyler drew parallels to the shift to mobile-friendly web pages and the "SEO nuclear arms race," predicting a future of "AI agent-optimized web pages" where human readability becomes secondary. For investors, Matt focused on infrastructure companies that support increased internet traffic, such as those providing power and cooling for data centers, though he cautioned on current high valuations (citing Cisco's dot-com bubble as a cautionary tale). Tyler, while admitting uncertainty, conceded it supports the AI infrastructure build-out narrative. He pondered opportunities for companies that can differentiate between human and AI traffic, or even those that can influence AI agents through new forms of "advertising." **U.S. Government Investment in Mining** The discussion then shifted to the Trump administration's $3 billion announcement concerning mining, including $100 million for education, aimed at reducing U.S. reliance on China for minerals. Jon noted China graduates 3,000 mining engineers annually compared to the U.S.'s 170, and half of the U.S. mining workforce is set to retire in three years. Matt Frankel was skeptical of the educational initiative's immediate impact. He explained mining's cyclical nature, remote job locations, and past instability (e.g., 2015-2020), which caused engineering enrollments to drop significantly. A four-year degree also means graduates won't fill immediate gaps. Tyler Crowe expressed further doubts, highlighting the "NIMBY" (Not In My Backyard) phenomenon, citing Maine's large lithium deposits that state laws prevent from being mined. He revealed that $1.4 billion of the $3 billion investment was actually a Department of Energy loan to a *private* lithium-ion battery startup, not a direct mining company, focusing on silicon anodes—a relatively easy-to-extract material, not the rare earths often discussed. Tyler warned against "hopes and prayers mining companies" that will inevitably emerge, advising investors to avoid these "siren songs" and instead wait for established companies during cyclical lulls. **Intel's $15 Billion Capital Raise** Finally, the panel discussed Intel's decision to raise $15 billion by selling equity, causing a slight dip in its stock. Intel cited CapEx, general working capital, and "progress in emerging areas, including physical AI... purpose-built silicon, advanced packaging, and external wafers" as reasons. Matt Frankel lauded the move, calling it a smart decision to raise capital when the stock is "expensive" (similar to Tesla's strategy). He noted the $15 billion represents less than 3% dilution and avoids taking on more debt. Crucially, the mention of "external wafers" suggests their third-party foundry business is accelerating, which is key to Intel's investment thesis. Given Intel's 395% stock jump in the past year and escalating CapEx plans ($20 billion this year, higher for 2027), Matt argued it's better to raise capital now from a position of strength rather than desperation later. Jon pointed out Intel's current price-to-sales ratio (8x) is more than double its 10-year average (3x), supporting the "expensive" valuation. He also noted the S&P 500's low dividend yield (1.04%), suggesting a highly valued market overall. Both Matt and Tyler agreed more equity raises from other companies are likely. Matt cited recent examples like Alphabet's $45 billion equity raise and increased U.S. equity raises in general, driven by both opportunistic valuations and the massive capital needs for AI infrastructure build-out. Tyler added that current cash flows aren't supporting spending plans, leading to debt and off-balance sheet financing. He believes this trend will continue until the market "cries uncle," potentially triggered by events like a credit downgrade, which would force companies to re-evaluate their capital spending plans.