The artificial intelligence (AI) revolution has delivered outsized gains to a concentrated set of market participants over the last few years. In particular, semiconductor companies supplying advanced processors and high-performance memory, hyperscalers building out cloud data center fleets, and select infrastructure providers enabling connectivity and custom silicon have all captured enormous value.
The reason is straightforward: Training and running frontier AI models requires enormous volumes of specialized compute, memory, and networking equipment. Hyperscalers and neoclouds have responded with record levels of capital expenditures, creating powerful demand for all the underlying hardware that supports this technology.
The question smart investors are asking is which stocks remain the best opportunities as "chipflation" ripples through the tech sector and beyond.
Image source: Getty Images.
What is "chipflation" and how is it impacting AI investors?
Morgan Stanley's research team describes "chipflation" as a phenomenon where the prices of memory chips rise sharply over time and stay lofty as demand persistently exceeds supply and many companies can't buy the volumes they need. For those that can get supply, the question becomes how much of their higher expenses to pass on to customers, and how much to accept reduced profit margins.
Demand for GPUs, high-bandwidth memory (HBM), and advanced DRAM should continue to outstrip supply because model sizes keep growing and inference workloads are expanding rapidly as agentic applications move into wider use. As hyperscalers continue to accumulate more compute capacity, Morgan Stanley suggests that the market needs to adjust to the idea of a "durable supply-demand reset," as it takes years to get new chip manufacturing foundries built and operating at full capacity.
This is far from a bearish stance for many of the companies in the sector. Morgan Stanley views the current environment as a transition from one characterized by hardware deployment occurring at a torrid pace to one where the focus shifts toward utilization rates, token economics, and returns on capital invested in infrastructure.
Against this backdrop, the long-term thesis supporting AI infrastructure remains intact: Data center build-outs and the development of AI-enhanced devices are unlikely to stop anytime soon, though the pace may begin to plateau. Investors who recognize this distinction can benefit by taking any near-term weakness in share prices as an opportunity to buy the dip on high-quality AI enablers rather than as a cue to exit the theme entirely.
Micron can still capture value in a constrained memory market
Micron Technology (MU 1.05%) remains exceptionally well positioned to navigate the chipflation environment. As a leading producer of HBM and DRAM, the company sits at the center of AI chip supply chains.
It's important to understand that even if the average selling prices for memory were to moderate from their recent peak levels, a structural shortage in HBM will persist, because each new generation of generative models and broader inference deployment will require greater amounts of specialized memory. For now, the hyperscalers are continuing to expand their capex budgets to build the massive cloud infrastructure that will be needed to support agentic AI when it goes mainstream.
This incremental spending growth will keep translating directly into higher unit volumes for Micron. In an environment where minor earnings disappointments can trigger sharp selling pressure in the AI chip trade, Micron's competitive position and long-term supply agreements provide a buffer.
As big tech focuses on scaling adoption rates and delivering measurable returns on AI infrastructure investments, the sheer scale of new model deployments favors suppliers with proven HBM capabilities. For this reason, Micron is positioned to turn its clients' continued capex expansions into durable earnings growth.
Image source: The Motley Fool.
Broadcom benefits from optimizing the network layer
Broadcom (AVGO 0.31%) stock offers a complementary way to participate in the same dynamics that are fueling Micron's ascent. While much of the discussion of the tight hardware supply issues in the AI sector revolves around memory and GPUs, the broader AI build-out cannot scale without proportional advances in high-speed networking, switches, and custom silicon. Broadcom's Tomahawk and Jericho switch families, optical connectivity solutions, and custom application-specific integrated circuit (ASIC) designs are becoming more essential as GPU clusters grow.
NASDAQ: AVGO
Key Data Points
Hyperscalers are increasingly looking for additional flexibility to accelerate spending on the supporting infrastructure layer within AI data centers. Broadcom's custom silicon work with Alphabet, Apple, and Meta Platforms is helping big tech to optimize model performance and reduce the total cost of ownership.
Since networking demand scales with the number of accelerators deployed on servers rather than with raw chip prices, Broadcom is somewhat insulated from the volatility seen in other areas of the AI chip supply chains. Moreover, the company's diversified portfolio across semiconductors and infrastructure software adds further resilience during periods when investors' focus shifts from pure hardware momentum to the magnitudes of hyperscalers' capex budgets.
Taken together, these dynamics illustrate why the current environment represents a new acceleration phase rather than a moderation of the AI infrastructure theme. Companies with durable competitive advantages that position them to be direct beneficiaries of hyperscalers' expanding capex budgets should deliver attractive returns as the market shifts from explosive deployments of cloud capacity into a more durable, predictable phase.
