Since the start of the AI boom in 2023, there have been several moments where the narrative around the buildout and long-term potential of this emerging technology has swung sharply from exuberance to doubt. Over the last few weeks, we appear to have entered another one of those doubt phases.
But it is important not to lose sight of how far this theme has come, not just over the last three years, but even over the last three months. In April, equities looked like they were entering a broader correction as geopolitical tensions flared and risk appetite deteriorated. Yet just weeks later, stocks found their footing and rallied aggressively.
That move was led by technology, AI-adjacent stocks, and, most notably, semiconductors. The SOXX semiconductor ETF more than doubled from those lows, while some of the biggest winners in the group, such as Micron Technology (MU), rallied more than 300% from depressed levels.
That kind of move naturally invites a reset.
Image Source: TradingView
The “Narrative Pendulum” is a concept I picked up from analyst Alex Barrow, and I think it is a useful framework for understanding this market (detailed here). The basic idea is that even when a powerful secular trend remains intact, the market’s perception of that trend can swing dramatically between extremes. In the case of AI, investors move from believing the opportunity is nearly unlimited to worrying that the entire buildout is excessive, wasteful, or unlikely to generate adequate returns.
That is where we are now. Concerns around overspending, capital misallocation, falling LLM costs, hyperscaler margins, and the ultimate return on invested capital are beginning to weigh on the AI trade. These concerns are not necessarily fatal to the long-term thesis. In fact, they are probably healthy. Periods of doubt help cool the kind of speculative enthusiasm that can drive prices almost straight higher and create a more durable base for the next leg of the cycle.
I continue to believe the AI boom has room to run, but a pause or correction here would not be surprising.
That brings me to Qualcomm (QCOM), a major player in the semiconductor industry that, until recently, has been best known as the dominant force in mobile chips. That remains a core business for the company, but smartphones are now a mature market. As a result, Qualcomm has increasingly been viewed as a slower-growth, more cyclical, and somewhat commoditized semiconductor company, not unlike how Micron was viewed in the memory space a little over a year ago.
That perception may now be changing.
A couple of weeks ago, at the company’s investor day event, Qualcomm management announced a significant pivot in the company’s strategic direction. While the company had been hinting at a larger role in AI over the last several months, the investor day made that shift far more explicit. Management unveiled a broader slate of AI-related business verticals, major hyperscaler relationships, and a much more ambitious vision for Qualcomm’s role in the AI infrastructure stack.
The key takeaway is that Qualcomm is not simply trying to enter the AI sector with one product. It is trying to position itself as a broader AI infrastructure platform.
That could include chips, connectivity, edge AI, inference capabilities, custom silicon opportunities, and data center acceleration. In other words, Qualcomm appears to be moving from being primarily viewed as a mobile-chip company to something closer to an “AI factory accelerator” — a company that helps hyperscalers and enterprise customers build, connect, optimize, and scale the infrastructure required for AI workloads.
I have many thoughts on this evolution, which I will detail more fully, but the timing of the announcement has been somewhat unfortunate in the short to medium term. Qualcomm unveiled this strategic pivot just as the semiconductor narrative began to swing from exuberance back toward skepticism. The stock initially reacted strongly to the news, but has since faded to multi-month lows.
In my view, that weakness has less to do with Qualcomm’s specific developments and more to do with the broader industry pullback. The market is currently questioning the entire AI infrastructure trade, and Qualcomm is being dragged into that reset despite potentially having just laid out one of the more important strategic transitions in its recent history.
If Qualcomm can successfully execute on this pivot, the stock may no longer deserve to trade primarily as a mature mobile-chip company. Instead, investors may begin to revalue it as a broader AI infrastructure beneficiary with exposure to hyperscalers, edge AI, data center acceleration, and next-generation compute demand.
The timing may be unfortunate, but the setup is becoming increasingly interesting.
Scope of Qualcomm’s Endeavors
The financial targets alone show how ambitious Qualcomm’s AI pivot has become. Management is targeting $5 billion in data center revenue by fiscal 2027 and $15 billion by fiscal 2029, with the early ramp expected to come largely from custom silicon and connectivity before the company’s accelerators and server CPUs become bigger contributors.
That is a major shift for a company still mostly viewed through the lens of smartphones.
At the center of the strategy is Qualcomm Dragonfly, the company’s new data center platform. Dragonfly is not one product, but a layered portfolio that includes connectivity silicon from the Alphawave acquisition, custom silicon for hyperscalers, AI inference accelerators, and eventually Oryon-based server CPUs. In the data center, Qualcomm expects the sequence to begin with connectivity, move into custom silicon in early fiscal 2027, then AI accelerators in the second half of fiscal 2027, followed by Oryon server CPUs in fiscal 2028.
The strategic logic is built around a major shift in AI workloads. The first phase of the AI boom was dominated by training large models, where Nvidia’s GPUs and CUDA software stack remain the standard. But the next phase may be increasingly driven by inference, especially as agentic AI systems begin chaining together dozens of model calls to complete more complex tasks. That dramatically increases the number of inference requests and makes power efficiency, memory bandwidth and cost per token far more important.
This is where Qualcomm believes it has an opening.
The company’s most important technical announcement was High-Bandwidth Compute, or HBC. Rather than relying on the traditional model of pairing accelerators with stacks of high-bandwidth memory, Qualcomm is pursuing a “memory first” architecture that places compute more directly beneath the memory stack. The goal is to reduce the distance data has to travel, improve efficiency, lower power consumption and address one of the biggest bottlenecks in AI inference.
Just as important is the software announcement. Qualcomm’s acquisition of Modular may be the key to making the whole strategy work. Hardware adoption in AI is heavily dependent on the developer ecosystem, and Nvidia’s CUDA moat has made it difficult for competitors to gain meaningful share. Cristiano Amon has framed the Modular acquisition as a potential Android or Linux moment for AI infrastructure, where a more open, hardware-agnostic software layer could reduce dependence on any single vendor.
That is a powerful idea. Rather than asking customers to abandon Nvidia overnight, Qualcomm can offer a software platform that runs across Nvidia, AMD and Qualcomm silicon, while still creating a natural path toward its own accelerators over time. If it works, Modular gives Qualcomm a much more credible way to enter the AI infrastructure market than hardware alone.
The company also added customer validation to the roadmap. Microsoft is expected to deploy Qualcomm’s HBC technology in Azure, while Meta has committed to a multigenerational agreement for Qualcomm CPUs in its data centers. Qualcomm also reinforced the software story through a partnership with Hugging Face, giving developers a path to deploy open models across Qualcomm platforms.
Finally, Qualcomm’s connectivity expertise may be one of its most underappreciated advantages. AI data centers are increasingly constrained not only by compute and memory, but by the ability to move massive amounts of data across racks and clusters. Through Alphawave, Qualcomm now has high-speed connectivity assets that are already generating revenue, giving Dragonfly a current revenue stream while the broader AI platform develops.
Execution risk remains significant. Qualcomm is entering a crowded market with powerful incumbents, and several of the most important products will not reach commercial scale until fiscal 2027 or fiscal 2028. But the scope of the announcement is hard to dismiss. Qualcomm is not simply adding AI exposure. It is attempting to build a full data center platform around the economics of inference, where power efficiency, memory bandwidth, custom silicon, software openness and connectivity may become increasingly important competitive advantages.
Image Source: Qualcomm
Qualcomm’s Auto Execution Extrapolated
Full disclosure, going into Qualcomm’s Investor Day, I had my doubts about the company’s foray into the AI data center buildout.
The technical capability was never really the question. Qualcomm has long been one of the most sophisticated chip designers in the world, with deep expertise in power efficiency, connectivity, system integration and edge computing. The bigger question was whether the company was simply too late. In a market already dominated by Nvidia, increasingly targeted by AMD and aggressively pursued by hyperscalers’ own internal silicon teams, it was fair to wonder whether Qualcomm could carve out a meaningful position.
But the more I look at the strategy, the more compelling it becomes.
Qualcomm is not making a single bet on one AI chip. It is taking a multi-pronged approach across connectivity, custom silicon, AI inference accelerators, server CPUs and software. That gives the company multiple ways to win. Some pieces of the portfolio may lag expectations, and that would not be surprising given the scale of the undertaking. But if even one or two segments meaningfully outperform, the overall opportunity could still become material.
I view the entire project almost as a strategic experiment. Qualcomm is putting several products into the market, testing where hyperscaler demand is strongest, and positioning itself around the areas where AI infrastructure is most likely to evolve next. Management may not describe it that way explicitly, but I think it is the right approach. The AI data center market is still young, and the economics are changing quickly. Rather than trying to predict the entire future with one product, Qualcomm is building a platform broad enough to adapt as the market develops.
That approach becomes more credible when viewed through the lens of Qualcomm’s recent success in automotive.
News from the automotive segment can get lost when management is announcing something as exciting as AI data center infrastructure, but the execution there may be the best model for what Qualcomm is trying to do now. The automotive business did not emerge overnight. Qualcomm entered through connectivity, expanded into the digital cockpit, and then moved deeper into advanced driver assistance and broader vehicle compute.
That layered strategy has worked. Automotive has quickly grown into one of Qualcomm’s most important non-handset businesses, crossing a $5 billion annualized revenue run rate in fiscal Q2 2026, with management expecting to exit fiscal 2026 above a $6 billion run rate. That is no longer a side project. It is becoming a real business line and a meaningful proof point for Qualcomm’s diversification strategy.
The parallel to AI infrastructure is important. In automotive, Qualcomm did not need to own the entire car to create value. It needed to identify the parts of the vehicle where compute, connectivity and software were becoming more important, then expand its content over time. In data centers, the same logic may apply. Qualcomm does not need to displace Nvidia across the full AI stack to succeed. It needs to find the areas where its advantages matter most.
That is why the inference-first focus is so important. Qualcomm is not trying to win yesterday’s AI infrastructure battle. It is trying to position itself for the next phase of the market, where power efficiency, memory bandwidth, connectivity and cost per token become more important as AI workloads scale from training into large-scale inference. Those are exactly the types of engineering problems Qualcomm has spent decades solving.
This does not eliminate execution risk. The data center market is larger, faster moving and more competitive than automotive. Nvidia’s ecosystem is entrenched, hyperscalers are increasingly building their own chips, and Qualcomm still has to prove that its roadmap can translate into commercial deployments at scale.
But automotive shows that Qualcomm can execute this type of transition. It can move beyond handsets, build a platform in an adjacent market, expand its content over time and convert long design cycles into meaningful revenue. That does not guarantee success in AI infrastructure, but it makes the plan far easier to take seriously.
For investors, that may be the key point. Qualcomm’s AI data center strategy should not be judged only as a late attempt to chase Nvidia. It should be viewed as the next test of the same diversification playbook that is already working in automotive. If the company can repeat even part of that success, the market may be underestimating how different Qualcomm’s business could look over the next several years.
Qualcomm Stock Breaks Down
The technical picture in QCOM stock offers a more tactical view of the setup.
Back in May, the stock rerated significantly higher after the company teased a major hyperscaler deal. From there, it built out a broad consolidation pattern, but since the full announcement, the stock has traded lower. Over the last week, QCOM broke below a key level of support, mirroring the broader weakness across the semiconductor sector.
Technical analysis does not provide reliable forecasting ability on its own, but it can show where large orders have left footprints. That is essentially what a “level” represents: an area where a meaningful amount of shares have changed hands and where buyers or sellers have previously shown up.
For now, QCOM remains below that breakdown level, and the near-term downtrend appears intact. That makes the stock more difficult for traders looking for a clean short-term entry. But at roughly 17x forward earnings, and with a potentially much larger long-term AI infrastructure opportunity beginning to take shape, the setup may be more attractive for investors looking for a bigger multi-year win rather than traders trying to capture the next short-term move.
The earnings revision picture may also supports a more patient view. Qualcomm currently has a Zacks Rank #3 (Hold), reflecting earnings estimates that have been relatively flat. That means analysts are not aggressively raising expectations yet, but they also are not cutting estimates in a meaningful way. In the context of a major strategic pivot, that leaves room for upside if management begins converting these announcements into visible revenue opportunities.
If revisions start to move higher, that could become an important bullish catalyst. A pickup in estimate momentum would signal that analysts are beginning to underwrite the AI data center opportunity more directly into their models, rather than treating it as a longer-dated optionality story.
Ultimately, the next major move in QCOM stock appears heavily tied to the broader semiconductor cycle. There may still be downside ahead over the next month if the group continues to unwind. But when the narrative pendulum finally bottoms and the market begins to lift the AI infrastructure theme again, Qualcomm could emerge with a much stronger story than it had in prior cycles.
The stock has broken down technically, but the business may be breaking out strategically.
Image Source: TradingView
Bottom Line on Qualcomm Stock
Qualcomm’s AI data center strategy is still early, and execution risk remains high. The company is entering a crowded market, several key products are still years from scale, and the stock remains caught in the broader semiconductor pullback.
But the announcement changes the long-term story. Qualcomm is no longer just a mature mobile-chip company looking for incremental growth. It is attempting to build a broader AI infrastructure platform across inference, connectivity, custom silicon, software and power-efficient compute.
For now, the technical setup is weak and earnings revisions remain flat, which supports the Zacks Rank #3 (Hold). But that also leaves room for upside if analysts begin raising estimates as AI data center revenue becomes more visible.
In the near term, QCOM may still trade with the broader semiconductor group. Over the next several years, however, the bigger question is whether Qualcomm can turn this roadmap into a real second growth engine.
The stock is not without risk, but the setup is becoming much more interesting.
Beyond Nvidia: AI's Second Wave Is Here
The AI revolution has already minted millionaires. But the stocks everyone knows about aren't likely to keep delivering the biggest profits. AI’s second wave is moving from infrastructure to implementation and these companies are at the forefront of this transition, positioned to become what Amazon and Google were to the internet era.See Stocks Now >>
Want the latest recommendations from Zacks Investment Research? Today, you can download 7 Best Stocks for the Next 30 Days. Click to get this free report
QUALCOMM Incorporated (QCOM): Free Stock Analysis Report
Micron Technology, Inc. (MU): Free Stock Analysis Report
This article originally published on Zacks Investment Research (zacks.com).
Zacks Investment Research