作为一名逆向科技投资者，我会在股票失宠时寻找机会。眼下，Cerebras Systems 引起了我的注意。这家总部位于加利福尼亚州森尼韦尔的AI芯片制造商于5月上市时，我喜欢它的技术，但认为该股估值严重过高。股价从185美元的IPO发行价飙升至最高386美元，但我在那些价位上对该股嗤之以鼻，甚至发表了一份报告，警告投资者不要买入。
As a contrarian technology investor, I look for opportunities when stocks fall out of favor. Right now, Cerebras Systems has my attention. When the Sunnyvale, California-based AI chipmaker went public in May , I liked its technology but thought the stock was grossly overvalued. Shares surged from their $185 IPO price to as high as $386, but I was pooh-poohing the stock at those levels and even published a report warning investors against buying.
如今，该股较峰值下跌约72%，我看到了一个截然不同的机会。要点 Cerebras创新的晶圆级引擎芯片设计可能使其在AI计算领域占据竞争优势。该公司的技术非常适合AI推理，这是该行业最大的增长机会之一。客户关系和可观的积压订单可能推动营收快速增长。Cerebras的股价受到了半导体板块整体抛售的拖累，但也有一些公司特有的担忧在压制该股。
Now with the stock down roughly 72% from its peak, I see a very different opportunity. Key points Cerebras' innovative Wafer-Scale Engine chip design could give it a competitive edge in AI computing. Company's technology is well-suited for AI inference, one of the industry's biggest growth opportunities. Customer relationships and a substantial backlog could fuel rapid revenue growth. Cerebras shares have been caught up in the broader semiconductor sell-off, but there are also company-specific concerns weighing on the stock.
其中最大的担忧之一涉及OpenAI，后者约占Cerebras合同业务的200亿美元。有关OpenAI在一个新AI模型中使用英伟达芯片而非Cerebras硬件的猜测，引发了外界对这段合作关系的质疑。我认为这些担忧被夸大了。Cerebras仍在建设其计算能力，OpenAI在此期间完全可能需要依赖英伟达。这并不改变我对Cerebras合作关系长期价值的看法。
One of the biggest involves OpenAI, which accounts for approximately $20 billion of Cerebras' contracted business. Speculation that OpenAI used Nvidia chips rather than Cerebras hardware for a new AI model have raised questions about the relationship. I think those fears are overblown. Cerebras is still building out its computing capacity, and it's entirely possible OpenAI needed to rely on Nvidia in the meantime. That doesn't change my belief in the long-term value of the Cerebras relationship.
我为何买入 一项可能具有突破性的芯片设计 我长期关注半导体技术，而Cerebras的晶圆级引擎是我多年来见过的最令人兴奋的创新之一。与传统的处理器不同，Cerebras的芯片将巨大的计算能力和带宽集成在单块硅片上。这使其能够快速处理海量信息，而无需在独立的芯片之间不断搬运数据。这种设计赋予了Cerebras真正的竞争优势。
Why I'm buying A potentially groundbreaking chip design I've been following semiconductor technology for a long time, and Cerebras' Wafer-Scale Engine is one of the most exciting innovations I've seen in years. Unlike traditional processors, Cerebras' chip packs enormous computing power and bandwidth onto a single piece of silicon. That allows it to process vast amounts of information quickly without constantly moving data between separate chips. This design gives Cerebras a real competitive advantage.
AI推理正是机会转移的方向。多年来，AI竞赛在很大程度上是围绕训练更大、更复杂的模型展开的。现在我认为真正的机会在于推理：让这些模型投入工作，生成内容并执行任务。这正是Cerebras的强项。其技术非常适合快速推理，这使它有机会挑战英伟达的GPU。CBRS年初至今走势图：Cerebras，年初至今。我还认为，Cerebras有机会与英伟达的技术形成互补，而不仅仅是与之竞争。
AI inference is where the opportunity is moving For years, the AI race was largely about training bigger and more sophisticated models. Now I think the real opportunity is inference: putting those models to work generating content and performing tasks. That's where Cerebras shines. Its technology is well suited for fast inference, giving it an opportunity to challenge Nvidia's GPUs. CBRS YTD mountain Cerebras, YTD I also see an opportunity for Cerebras to complement Nvidia's technology rather than simply compete with it.
通过与亚马逊云科技（AWS）的合作，Cerebras正在开发一种名为“推理解耦”的方法，将工作拆分给两类芯片。AWS的Trainium芯片负责处理请求的初始阶段，即“预填充”，而Cerebras负责“解码”，即生成响应。我认为这可能为Cerebras打开一个大得多的市场。重要客户，巨大的增长潜力。Cerebras已与一些科技界最知名的公司建立了合作关系，包括OpenAI、AMD、AWS和Flex。
Through a partnership with Amazon Web Services , Cerebras is developing an approach called inference disaggregation, which splits the work between two types of chips. AWS's Trainium chips handle the initial processing of a request, known as "prefill," while Cerebras handles "decode," or generating the response. I think this could open up a much bigger market for Cerebras. Major customers, massive growth potential Cerebras has built relationships with some of the biggest names in technology, including OpenAI, AMD, AWS and Flex.
引人注目的是其合同业务规模相对于当前收入的比例。该公司拥有约250亿美元的剩余履约义务，其中约200亿美元与OpenAI相关。这与该公司上一季度约2.1亿美元的核心收入相比，反差鲜明。Cerebras仍需扩大其计算能力以履行这些承诺，但我预计到2027年其收入将增长两倍以上，达到约28亿美元，公司同时实现现金流转正。
What stands out is the size of its contracted business relative to its current revenue. The company has approximately $25 billion in remaining performance obligations, including about $20 billion tied to OpenAI. That's striking compared with the roughly $210 million in core revenue it generated last quarter. Cerebras still needs to expand its computing capacity to fulfill those commitments, but I expect revenue to more than triple to approximately $2.8 billion in 2027, with the company also turning cash-flow positive.
为什么是现在？该股目前交易价约为107美元。我认为这轮抛售是一个有吸引力的入场点，可以买入一家拥有创新技术和巨大增长潜力的公司。我在Freedom Capital Markets给出的目标价为209美元，意味着较当前水平约有91%的上涨空间。总结：风险是存在的。
Why now? Shares are trading around $107 now. I see the sell-off as an attractive entry point into a company with innovative technology and substantial growth potential. My price target at Freedom Capital Markets is $209, implying roughly 91% upside from current levels. Bottom line There are risks.
Cerebras仍需兑现其重大客户承诺，并证明其技术能够在英伟达主导的市场中成功竞争。但市场过于关注短期不确定性，而忽视了Cerebras未来可能成为什么样的公司。当这只股票首次公开发行、股价飙升时，我曾劝告投资者不要购买。如今股价已大幅下跌，我站到了交易的另一边。这就是逆向投资者的妙处：在他人眼中是风险的地方发现机会。
Cerebras still needs to execute on its large customer commitments and prove that its technology can compete successfully in a market dominated by Nvidia. But the market is focusing too much on the near-term uncertainty and not enough on what Cerebras could become. I was warning investors against buying this stock when it first went public and shares were soaring. Now that shares have plunged, I'm on the other side of the trade. That's the beauty of being a contrarian investor: finding opportunity where others see risk.
Paul Meeks是Freedom Capital Markets的董事总经理兼技术研究主管。披露：Meeks本人或通过公司均未持有该股票。[CNBC Pro撰稿人]表达的所有观点仅为其个人观点，不代表CNBC或其母公司及附属公司的观点，且可能之前已通过电视、广播、互联网或其他媒介由其传播。本内容作为我们的编辑产出提供，仅供参考，不构成财务、投资、税务或法律建议，也不构成购买任何证券或其他金融资产的推荐。
Paul Meeks is the Managing Director and Head of Technology Research of Freedom Capital Markets . Disclosures: Meeks does not own shares personally or through the firm. All opinions expressed by the [CNBC Pro contributors] are solely their opinions and do not reflect the opinions of CNBC, or its parent company or affiliates, and may have been previously disseminated by them on television, radio, internet or another medium. This content is provided as part of our editorial output for informational purposes only and does not constitute financial, investment, tax or legal advice or a recommendation to buy any security or other financial asset.
本内容具有普遍性，不反映任何个人的独特个人情况。上述内容可能不适合您的特定情况。在做出任何财务决策之前，您应强烈考虑寻求您自己的财务或投资顾问的建议。本内容仅供参考，不构成财务、投资、税务或法律建议，也不构成购买任何证券或其他金融资产的推荐。本内容具有普遍性，不反映任何个人的独特个人情况。
The content is general in nature and does not reflect any individual's unique personal circumstances. The above content might not be suitable for your particular circumstances. Before making any financial decisions, you should strongly consider seeking advice from your own financial or investment advisor. THIS CONTENT IS PROVIDED FOR INFORMATIONAL PURPOSES ONLY AND DOES NOT CONSTITUTE FINANCIAL, INVESTMENT, TAX OR LEGAL ADVICE OR A RECOMMENDATION TO BUY ANY SECURITY OR OTHER FINANCIAL ASSET. THE CONTENT IS GENERAL IN NATURE AND DOES NOT REFLECT ANY INDIVIDUAL'S UNIQUE PERSONAL CIRCUMSTANCES.
上述内容可能不适合您的特定情况。在做出任何财务决策之前，您应强烈考虑寻求您自己的财务或投资顾问的建议。
THE ABOVE CONTENT MIGHT NOT BE SUITABLE FOR YOUR PARTICULAR CIRCUMSTANCES. BEFORE MAKING ANY FINANCIAL DECISIONS, YOU SHOULD STRONGLY CONSIDER SEEKING ADVICE FROM YOUR OWN FINANCIAL OR INVESTMENT ADVISOR.