Wall Street’s AI trade is entering a new phase — and this time the winners are no longer the mega-cap software platforms that defined the market’s first generative AI rally.
Through the first half of 2026, investor leadership rotated decisively toward the infrastructure layer powering the artificial intelligence economy: semiconductors, high-bandwidth memory, data-center architecture, networking systems and power infrastructure.
The shift marks a meaningful evolution in how Wall Street is pricing the AI cycle.
Wall Street’s AI Rally Moves Beyond the Magnificent Seven
During the initial phase of the boom, capital concentrated heavily in the so-called Magnificent Seven, as investors rewarded platform dominance, cloud distribution and AI application optionality.
But the current market regime increasingly reflects something different: a recognition that the next stage of AI monetisation will be constrained less by software demand and more by physical compute capacity.
That distinction has become increasingly visible across equity performance.
The Technology (XLK) sector rose roughly 33 per cent in the first half of the year, its strongest start since 2023.
Yet beneath the headline performance, Wall Street has aggressively rotated away from crowded mega-cap positioning and into companies supplying the computational backbone of large-scale AI deployment.
Microsoft shares, despite the company’s central role in enterprise AI, fell roughly 23 per cent year-to-date. Nvidia, while still strategically dominant, no longer monopolised investor enthusiasm.
Instead, Wall Street’s attention shifted toward a broader set of infrastructure beneficiaries — particularly memory and storage suppliers exposed to the explosive growth in AI training and inference workloads.
Why Memory Chips Have Become Wall Street’s New AI Obsession
Micron Technology became one of the clearest expressions of that trend.
The US memory-chip producer, now viewed as a critical supplier of high-bandwidth memory used in AI accelerators, emerged as one of Wall Street’s strongest-performing large-cap technology stocks, rallying roughly 700 per cent over the past year.
The move reflects mounting concern across the industry that HBM supply could become one of the key bottlenecks limiting AI scaling economics.
“The moves we’re seeing are fundamental in nature, not merely hype”, said Alex Liberfield, Managing Partner at Miami-based Liberfield Capital, a private fund specializing in alternative investments.
Across Wall Street, investors increasingly appear to believe that the scarcity value inside AI is migrating downward through the stack — away from applications and toward the underlying compute infrastructure itself.
That dynamic has driven sharp repricing across the semiconductor ecosystem.
Shares in Micron, Western Digital, Seagate Technology and Intel have each risen more than 250 per cent this year, helping propel the iShares Semiconductor ETF roughly 110 per cent higher over the past six months.
The underlying thesis is becoming more widely accepted on Wall Street: as AI models become larger, more inference-intensive and increasingly multimodal, the real economic bottlenecks are shifting toward memory bandwidth, networking throughput, energy availability and data-centre scalability.
In other words, AI is starting to resemble an industrial infrastructure cycle as much as a software revolution.
Even so, investors remain wary of overheating valuations.
“There could still be room to run if the AI Capex cycle holds”, Liberfield said. “But these will likely be volatile names and if the Korean produces add supply, the pricing power of these companies erodes fast”.
Liberfield Capital: It could definitely be the beginning of a new cycle
Still, Liberfield believes Wall Street may only be in the early stages of a broader structural repositioning:
“It’s likely a real and broad and could definitely be the beginning of a new cycle. The Mag7 mega-cap tech space is very crowded and it is now unlikely to offer the price appreciation of the last few quarters. In my view the next leaders will be names in memory and storage, custom silicon, and the broadening of the rest of the market, especially industrial and manufacturing names that sat out the AI craze”.
Liberfield added that “in the US the sectors I’d favor in this space are the AI infrastructure pick-and-shovel names as well as potentially re-shoring industrials”.
At the same time, “you could argue you can find better opportunities in the non-US space, Europe looks cheap, especially its defense and industrial names, which look better positioned than its consumer tech space”.
Joseph Sroka (NovaPoint Capital): “AI infrastructure spending could remain structurally elevated for years”
Joseph Sroka, co-founder and chief investment officer at NovaPoint Capital, argues that Wall Street is still underestimating both the scale and duration of the AI infrastructure buildout.
“The compute demand for AI has been growing. This creates demand for all the components that are going into a data center ranging from semiconductors to cables to cooling to electrical equipment to power generation”.
Importantly, much of the spending announced by hyperscalers reflects multi-year capital expenditure programmes rather than short-term cyclical investment.
That distinction matters because it suggests AI infrastructure spending could remain structurally elevated for years rather than quarters: “Much of the AI spending announcements we have seen are on multi-year timelines so the investment should continue to flow for several quarters and years”.
Sroka believes Wall Street is beginning to appreciate that sustainable AI adoption requires a much broader industrial ecosystem than initially assumed.
Investors are increasingly looking beyond software platforms and chip designers toward utilities, industrials, electrical infrastructure providers and data-centre real estate owners:
“I think this is a broadening of the investment opportunities. Investors are moving beyond just the technology components of chips and hardware and seeing that for long-term viability of AI adoption it involves companies in the industrial, utility, and even real estate sectors. A broadening of opportunity in a growing part of the economy is healthier and likely more stable than a narrowly focused opportunity that only benefits a small number of participants. It is a cycle, but a cycle with some breadth to it”.
“AI as a digital supply chain”. The New Ideal AI Portfolio Beyond the Magnificent Seven
In that sense, AI increasingly resembles a vertically integrated supply chain rather than a narrow technology subsector.
Sroka described the AI as a “ digital supply chain ”, since “we have real estate that houses data centers, power generation that supplies the electricity, the technology hardware and components that support the work, and the hyperscalers and their models that produce the work”.
Based on that framework, NovaPoint Capital favours companies across the full infrastructure stack.
“Companies that are leaders across this supply chain should be stock market leaders. We like companies in power and electrical infrastructure to include GE Vernova (GEV), Eaton (ETN), and Vertiv (VRT). We like the data center REITs to include Digital Realty (DLR) and Equnix (EQIX). We like the utility companies supply power in the large data center markets to include Southern Company (SO), Dominion Energy (D), and Duke Energy (DUK). We like tech hardware and components to include Nvidia (NVDA), Corning (GLW), Broadcom (AVGO), and Amphenol (APH). Finally we like hyperscalers to include Alphabet (GOOGL), Palantir (PLTR), Microsoft (MSFT) and Oracle (ORCL). We own all these stocks in our investment strategies”.
For Wall Street, the broader message is becoming increasingly difficult to ignore.
The AI trade is no longer simply about software narratives, chatbot adoption or multiple expansion among mega-cap technology groups.
It is evolving into a full-scale infrastructure and industrial investment cycle — one that could ultimately prove larger, broader and considerably more durable than the market’s original generative AI thesis.
