Dan Ives Just Named The 3 Biggest AI Bottlenecks - These Stocks Are Still Dirt-Cheap - Get In Now

Memory is the biggest constraint that we see and you probably don't have equilibrium till 2028 2029. Now you could argue different on in terms of the stocks or what are they reflecting is it cyclical? Has this almost changed the cyclicality? Power is clearly the biggest constraint. And I think the biggest constraint is just data centers need to get built in the US for this all to happen. That's there's no gray. I think the biggest risk relative to US China is a data center that because there's no debate where we're winning on chips, models hyperscalers IP China winning on robotics and energy. But if data centers are not built here, that's that'd be like, you know, a hospital not having, you know, surgeons or not having, you know, the technology to do what they do. So that that continues to really be the bottom >> If you were waiting for the most sane reality-grounded voice on what comes next in the AI trade as recent market jitters send edge investors scrambling, Wall Street veteran Dan Ives just delivered a crucial reality check. While sensationalist headlines warn of an impending AI bubble burst driven by Chinese open-weight models eroding pricing power, the underlying math tells a vastly different story. Big Tech's hyperscale capital expenditure isn't slowing down. It's accelerating toward a projected $600 billion plus in 2026 with over 75% explicitly funneled into physical AI infrastructure. We aren't at the end of the line. We are only in the third inning of a decades-long technological transformation. Frontier closed-source labs like OpenAI and Anthropic are building massive enterprise moats around proprietary data that open-source commoditization simply cannot touch. The real risk to the AI trade isn't a lack of software demand or a drop in model margins. The real risk lies in the massive supply-side bottlenecks choking the entire ecosystem. High-bandwidth memory deficits, a severe power grid crisis, >> [music] >> and a lack of physical data center capacity. To win this market shift, you have to follow the physical bottlenecks. Here is the complete breakdown of the top stocks positioning themselves to solve these three critical infrastructure deficits. Let's begin with Micron Technology, ticker MU, the cornerstone of American memory manufacturing and a premier play on the high-bandwidth hardware deficit. Micron stands as a prime beneficiary of the global high-bandwidth memory shortage, positioning it directly at the heart of the AI buildout. What makes the current moment so compelling is a sharp disconnect between short-term market sentiment and pure operational performance. Following a recent 26% decline driven largely by broader market volatility and leveraged liquidations in overseas memory indices, Micron's fundamental engine hasn't missed a single beat. The company is riding high on unprecedented demand for HBM3E, which is essential for powering next-generation AI accelerators like Nvidia's Vera Rubin and AMD's Helios architectures. Modern AI models require massively expanding context windows, meaning hardware architectures demand more memory bandwidth, not less. To meet this massive appetite, hyperscalers are entering long-term contractual commitments. Micron has locked in 16 strategic customer agreements, 14 of which represent a minimum floor of $100 billion in remaining performance obligations over the next 5 years. Crucially, these floor prices guarantee profit margins that exceed peak quarterly levels seen in any past memory cycle. Looking forward, the financial picture is staggering. Following a quarter that delivered $41 billion in revenue at 68% net margins, Micron's near-term guidance projects revenue pushing toward $50 billion. Beyond individual contract visibility, domestic US hyperscalers are prioritizing supply security, giving Micron a distinct structural advantage alongside continuous support from US Chips Act incentives. Even as memory price increases naturally normalize over time, Micron's cash flow generation remains insulated by taker pay agreements and $22 billion in committed customer financing. With industry peers like Samsung indicating that the global memory supply shortfall could persist straight through 2028, Micron finds itself operating at a historic inflection point. The business is converting massive infrastructure demand into high-margin contracted cash flows, solidifying its place as an indispensable player in the hardware landscape. Pivoting from memory chips straight into physical connectivity constraints, our next selection is Marvell Technology, ticker M RVL, a high-speed data center infrastructure powerhouse experiencing a structural transformation. Marvell sits directly at the intersection of AI cluster scaling, high-bandwidth optical interconnects, and custom silicon development, making it essential as massive data centers push thermal and bandwidth boundaries. While the stock recently underwent a steep 50% correction from its June highs, down to around 46 times forward earnings, dip buyers have begun re-emerging near key technical support as long-term fundamentals continue to solidify. The fundamental validation for Marvell came via a major $2 billion strategic investment and [music] partnership from Nvidia. Integrating Marvell's custom XPUs and optical solutions directly into the NVLink Fusion platform and Spectrum 6 Ethernet ecosystem. To sharpen its focus on these hyper-growth opportunities, Marvell streamlined its operations by offloading its non-core automotive assets for $2.5 billion in cash, redeploying capital straight into high-return AI chip development. Marvell's competitive moat rests on three massive catalysts. First is its custom ASIC business, where hyperscaler programs, including AWS Trainium 2 and 3, Google Axion ARM CPUs, and Meta DPUs are projected to drive custom silicon revenue from $1.5 billion up to over $4 billion by fiscal year 2028. Second is its interconnect leadership, where mass deployment of 1.6 terabit optical DSPs is expected to fuel more than 70% revenue growth in its interconnect segment. Third is its pioneering position in the Compute Express Link or CXL market, which directly addresses memory supply bottlenecks by allowing servers to pool DRAM efficiently. Marvell projects its CXL revenue alone to reach $2 billion by fiscal 2028. As agentic AI models scale, triggering thousands of autonomous model queries across distributed clusters, the demand for Marvell's re-timers, network interface cards, and optical switches expands [music] exponentially. With management setting revenue targets of $11.5 billion for fiscal 2027 and $16.5 billion for fiscal 2028, Marvell represents a vital infrastructure player, turning physical connectivity constraints into multi-year revenue inflection. Crossing over from data center hardware into the raw grid power required to keep these facilities running, our final pick is Constellation Energy, ticker CEG. Addressing the ultimate bottleneck of the entire AI revolution, power generation. As hyperscale data centers scale from tens of megawatts to gigawatt class campuses, access to reliable 24/7 carbon-free energy has become the single biggest gatekeeper to AI expansion. Constellation Energy holds an unmatched competitive position as the largest producer of carbon-free electricity in the United States, controlling more than half of the nation's merchant nuclear fleet. While critics often mistake Constellation for a traditional utility, the fundamental reality hinges on pure structural scarcity. US data center power demand is projected to surge by up to 90 gigawatts by 2030, but the total deliverable >> Mhm.