원문 보기
What if the market is not crashing because the bull market is ending? What if it is shaking because the biggest opportunity is about to begin? That is the message Tom Lee is giving investors right now. The S&P 500 has already crossed his 7,300 target earlier than expected. It is now hovering near 7,600 and suddenly fear is gripping Wall Street.
Has the market moved too far, too fast? Is the AI trade overheating? Are investors walking into another painful correction? Tom Lee understands that fear. In fact, he expects the next few weeks could get messy as markets deal with midterm election uncertainty and a wave of massive IPOs from names like SpaceX, OpenAI, and Anthropic.
That kind of setup could drain liquidity, shake confidence, and make even a healthy bull market look bearish. But Lee's bigger message is completely different. He believes this short-term fear could become a godsend buying opportunity. Beyond the noise, he still sees 7,700 on the table for the S&P 500 in the second half of 2026 and believes 2027 could be the best year of our lifetime.
So, if the market gives investors one more dip, the real question is simple. Which stocks should they buy before 2027 begins? Today, we are looking at three stocks that fit Tom Lee's playbook perfectly. Companies tied directly to the infrastructure behind the next phase of AI. . First on the list is Marvell Technology, ticker symbol MRVL, and this is exactly the type of company that fits Tom Lee's view of a scary-looking correction before a much bigger long-term opportunity.
Marvell is not trying to replace Nvidia as the face of AI. Its role is different and more hidden. It sits inside the plumbing of artificial intelligence. Custom chips, optical interconnects, switching, silicon photonics, retimers, CXL, and the high-speed networking that allows massive AI data centers to actually function.
That is why Marvell's latest quarter matters. The company reported record fiscal Q1 2027 revenue of $2.418 billion, up 28% year-over-year and 9% sequentially. Data center revenue alone reached $1.83 billion, representing 76% of total revenue. This shows Marvell is no longer just a broad semiconductor supplier.
It is becoming a focused AI infrastructure company. The outlook is even stronger. Management expects fiscal 2027 revenue to grow roughly 40% to nearly $11.5 billion and fiscal 2028 revenue to reach about $16.5 billion, roughly $1.5 billion higher than its outlook just one quarter earlier. Data center revenue is expected to grow about 50% this year and 55% next year, while the interconnect business is projected to grow more than 70% in fiscal 2027.
That matters because AI is moving toward reasoning models, mixture of expert systems, and agentic AI, where one user request can trigger multiple model calls, memory lookups, and massive data movement across clusters. In that world, networking, optics, switching, and interconnect become more critical with every generation.
Marvell also has major catalysts, an expanded Nvidia partnership in silicon photonics, NVLink Fusion, and AI RAN, a custom silicon roadmap that could exceed $10 billion by fiscal 2029, the Alphabet related chip design opportunity, the Polariton acquisition, and more than $600 million expected from scale-out switching this year.
If the market pulls back, Marvell may not be the loudest AI stock, but it could be one of the most important companies building the stage beneath the AI revolution. Now, moving from the chips and interconnects behind AI to the raw compute capacity that AI companies desperate need, the second stock is Nebius, ticker symbol NBIS.
And this is not a traditional cloud story. Nebius is a pure play AI infrastructure company selling one of the most valuable resources in the modern economy, high-performance AI compute, GPUs, AI cloud capacity, data centers, software tools, power access, and infrastructure built for training and running the next generation of AI models.
The latest numbers explain why this stock has captured so much attention. In Q1 2026, Nebius reported revenue of $399 million, up 684% year-over-year and 75% sequentially. Its Nebius AI cloud segment generated $389.7 million, meaning about 98% of total revenue is now directly tied to AI cloud demand.
That is extraordinary concentration in one of the fastest growing areas of infrastructure. Even more impressive, adjusted EBITDA turned positive at $100 29.5 million, while Nebius AI adjusted EBITDA margin reached 45%. That matters because it shows Nebius is not just buying growth at any cost. Its live capacity is already showing real operating leverage, but the bigger story is the ramp ahead.
Nebius ended the quarter with annualized run rate revenue near $1.9 billion, yet management is guiding for $7 billion to $9 billion by the end of 2026. Demand is not the problem. Capacity is the bottleneck. Management says everything it builds is being sold, which is why 2026 CapEx guidance jumped to 20 billion dollars to 25 billion dollars, mainly to support 2027 capacity.
That connects directly to Tom Lee's 2027 thesis. If AI infrastructure remains scarce, companies that can bring compute online quickly may become extremely valuable. Nvidia already has validation from Microsoft, Meta, Nvidia, and Leopold Ash and Brenner's fund, which disclosed a 5.6% stake. Meta agreements could be worth up to 27 billion dollars over 5 years.
Microsoft has signed a large AI infrastructure deal, and Nvidia invested about 2 billion dollars. With Eigen AI strengthening inference optimization, Bloom Energy targeting the power bottleneck, and contracted capacity above 3.5 gigawatts, Nvidia is not just riding the AI wave. It is trying to build the wave.
And after Marvell powers the data flow and Nvidia expands the compute cloud, the third stock brings the story into physical deployment. Dell Technologies, ticker symbol d e l l. This may be the most surprising name in the AI infrastructure trade. For years, Dell was seen as a mature PC and server company, useful, profitable, but not exciting.
Now that image is changing fast because AI is no longer just about GPUs. It is about turning those GPUs into complete AI factories with servers, networking, storage, liquid cooling, enterprise services, and real-world deployment at massive scale. Dell's latest quarter made that shift impossible to ignore.
Revenue reached 43.8 billion dollars, up 88% year-over-year, while adjusted EPS jumped 214% to 4.86 dollars. The Infrastructure Solutions Group surged 181% to $29 billion and AI server revenue hit $16.1 billion in just one quarter, up 757% from a year ago. Dell is not watching the AI revolution from the sidelines.
It is helping companies physically deploy it. The demand picture is even more powerful. Dell converted $24.4 billion of AI orders during the quarter and still exited with a massive $51.3 billion AI server backlog. Management raised full-year AI server revenue guidance from $50 billion to $60 billion and lifted total fiscal year revenue guidance to roughly $167 billion at the midpoint.
The message was clear. Demand is not the issue. Supply is with memory availability still the key constraint. That matters because a gentic AI needs more than accelerator chips. As AI moves from experiments into real workflows, companies need orchestration, CPU servers, storage, security, networking, and data systems around every GPU cluster.
Dell participates across that stack. Its PowerScale and ObjectScale storage products are gaining attention as unstructured data becomes a major bottleneck. While its PC business also benefits from enterprise refresh cycles with roughly 1/3 of the installed base still 4 years or older. If Tom Lee is right about a short rough patch before a stronger 2027, Dell represents a different kind of AI winner, not the flashiest name, but one of the most essential companies building, shipping, and scaling the AI infrastructure cycle.
So, if the next few weeks look bearish, investors should not confuse volatility with the end of the story. Tom Lee's message is that fear may arrive before opportunity. Marvell, Nebius, and Dell each attack a different AI bottleneck: connectivity, compute capacity, and physical deployment. And if 2027 becomes the year Lee expects, these are the kinds of infrastructure stories investors will be watching closely.
Comments 0