Investors searching for the next leg of the artificial intelligence trade are increasingly looking past the highest-profile chip names and toward companies that supply the plumbing of the AI buildout. Broadcom, Marvell Technology and SanDisk are emerging as three of the more closely watched beneficiaries of that shift, with market participants betting that sustained demand for networking, custom silicon, storage and high-performance data-center components could support a broader advance in their shares.
The argument is not that these companies are suddenly new to the AI story. Rather, it is that their businesses sit at critical points in the infrastructure chain, where spending remains elevated even as the market becomes more selective about valuation. Broadcom has become a central name in AI networking and custom accelerator-related demand, Marvell is tied to data-center connectivity and specialized chips, and SanDisk is positioned to benefit from storage demand as AI workloads continue to expand the need for fast, reliable memory and flash capacity.
Infrastructure Over Hype
What makes the current setup notable is the way it differs from the first wave of AI enthusiasm, which concentrated heavily on a small group of mega-cap leaders. That initial phase rewarded companies with direct exposure to model training, cloud compute and semiconductor fabrication. The next phase may be more about the less glamorous but essential layers that allow AI systems to scale: switching, interconnects, storage, and custom silicon designed for specific workloads.
Broadcom has been one of the clearest examples of this transition. The company's mix of semiconductor and infrastructure software exposure gives it a diversified earnings base, while its role in AI-related networking and custom chip programs has sharpened investor focus on the growth potential of its data-center franchise. Marvell, meanwhile, has been building a case around its position in high-speed connectivity and custom silicon for cloud customers. Its fortunes are increasingly linked to how aggressively hyperscale operators continue to spend on AI-capable infrastructure.
SanDisk, though often discussed less prominently than the larger chipmakers, also fits the theme. As AI systems generate and process more data, storage demand becomes a more important part of the investment case. Flash memory and related storage products are not the flashiest part of the AI ecosystem, but they are indispensable to the economics of large-scale computing. That makes SanDisk a potential beneficiary if the current infrastructure cycle broadens beyond compute alone.
Earnings Expectations Matter
The market's interest in these names is being reinforced by a familiar but powerful catalyst: rising earnings expectations. In equity markets, upward revisions to profit forecasts often matter as much as, or more than, the headline growth narrative. When analysts begin to lift estimates, it can signal that demand is proving more durable than expected, margins are holding up, or management commentary suggests a stronger second half than the market had priced in.
That dynamic is especially important in a sector where valuations can move quickly on sentiment. If investors believe the AI buildout is still in its early or middle stages, companies with credible exposure to the underlying infrastructure may attract fresh capital even after substantial gains. The result can be a self-reinforcing cycle: stronger earnings expectations support higher share prices, which in turn draw more attention from momentum and growth investors.
Still, the setup is not without risk. The AI trade has already become crowded in parts of the market, and any sign of slower capital expenditure from cloud providers could pressure the group. Competition is also intense, and investors will be watching for evidence that demand is translating into sustained revenue growth rather than one-time order strength. For Broadcom and Marvell, execution on product cycles and customer concentration will remain central. For SanDisk, the key question is whether storage demand can keep pace with the broader AI infrastructure expansion.
Rotation Could Broaden
The broader market implication is that leadership in AI may be widening. If the first winners were the obvious compute and platform names, the next winners may be the suppliers that make the entire system work at scale. That kind of rotation often matters for the durability of a bull market, because it allows capital to spread across more companies and more parts of the supply chain.
For now, Broadcom, Marvell and SanDisk are not being presented as speculative stories detached from fundamentals. They are being viewed as companies with tangible exposure to a secular spending cycle that remains intact. If AI infrastructure investment continues to accelerate, and if earnings revisions keep moving higher, these stocks could remain among the more compelling ways to participate in the theme without relying solely on the market's most crowded names.
In that sense, the bullish case is less about a sudden breakout than about a steady accumulation of evidence: improving estimates, persistent data-center demand and a market increasingly willing to reward the companies building the backbone of the AI economy. That combination can be powerful, especially when investors begin to look for the next phase of the trade rather than the last one.