Key Takeaways
S&P 500 Index earnings grew roughly 31% year-over-year in Q2 2026, and AI infrastructure companies contributed about half of that growth, with their own earnings rising around 54%.
The median S&P 500 Index company still grew earnings-per-share by 14% year-over-year, showing broad corporate health alongside the outsized AI infrastructure gains.
The WisdomTree Artificial Intelligence and Innovation Fund (WTAI) holds positions across semiconductors, memory, power infrastructure and connectivity, the layers where we see bottlenecks emerging.
The Q2 2026 earnings season is nearly complete, and one story is appearing above the rest.
Artificial intelligence infrastructure is not a speculative narrative. It is the single most powerful driver of corporate profit growth in the S&P 500 Index right now.1
For the WisdomTree Artificial Intelligence and Innovation Fund (WTAI), the season delivered important validation, and some equally important clues about where the investment opportunity might evolve from here.
The Numbers That Matter
S&P 500 Index earnings per share grew roughly 31% year-over-year in Q2, and AI infrastructure companies, specifically the hyperscalers and the ecosystem of companies benefiting from their capital expenditure, contributed approximately half of that growth, with their own earnings rising around 54% year-over-year. Put bluntly, that is roughly one segment of the market carrying an outsized portion of aggregate index profit growth.
Importantly, the median S&P 500 Index company also grew earnings-per-share (EPS) by 14% year-over-year, so this is not a story of AI running strongly while the rest of the market struggled. Broad corporate health appeared solid, but the AI infrastructure layer is operating in a different earnings atmosphere altogether.
Some of WTAI’s largest positions are concentrated precisely in this layer, as we have been discussing the ‘bottlenecks thesis’ through 2026.2
NVIDIA (5.1%) sits at the foundation of every major AI training and inference deployment.
Micron Technology (4.6%) and SK Hynix (2.2%) supply the high-bandwidth memory that makes modern GPU clusters function at the required speeds.
Taiwan Semiconductor (3.0%) manufactures the silicon that underpins nearly every AI chip worth discussing, while ASML (2.0%) and Applied Materials (1.7%) produce the equipment without which advanced nodes cannot be fabricated at scale.
Lam Research (2.4%) and KLA Corp (1.1%) complete that semiconductor capital equipment picture.
Broadcom (3.3%) designs custom AI accelerators for some of the largest hyperscalers.
Advantest (1.8%) and Teradyne (1.9%) handle the test and measurement side of an increasingly complex chip supply chain.
These are the companies whose revenue lines move when AI infrastructure spending moves, and in Q2, that spending moved sharply.
The Bottleneck Story Is Alive
What makes the current environment particularly interesting is that AI infrastructure spending is accelerating even as companies report that AI inference expenses currently represent less than 0.5% of S&P 500 Index revenues. That may seem paradoxical until you understand what it might imply.
The AI buildout is still in its early stages relative to the eventual installed base.
Demand is being pulled forward by hyperscaler capital expenditure commitments, and the bottlenecks being addressed today are precisely the ones that will determine which companies capture the next phase of that buildout.
One set of bottlenecks sits in connectivity and data movement. WTAI holds Credo Technology (1.4%), Astera Labs (1.2%), Lumentum Holdings (1.7%) and Coherent Corp (1.2%), which are companies whose optical interconnect and high-speed connectivity products address exactly this problem. Marvell Technology (1.9%) designs custom silicon for data center connectivity and AI networking. When cluster sizes grow from thousands of GPUs to hundreds of thousands, the interconnect layer becomes a critical path item, not an afterthought.
Power infrastructure is another bottleneck that we hear about nearly every day in different contexts. GE Vernova (1.9%), Vertiv Holdings (1.6%) and Bloom Energy (1.6%) are in WTAI because data center power density has reached levels that require rethinking grid connection, cooling architecture and backup power design simultaneously. These are not AI companies in the traditional software sense; they are industrial and energy companies whose growth trajectories have been fundamentally altered by the scale of the AI infrastructure buildout. Cummins (1.1%) belongs in this same conversation, providing generator and power systems solutions for facilities that cannot tolerate interruption.
At the memory and storage layer, possibly the most widely discussed of all the AI bottlenecks in 2026, the fund holds not only Micron and SK Hynix but Samsung Electronics (4.5%), Kioxia Holdings (2.9%) and SanDisk (2.6%). SanDisk was among the strongest performers in the market, up 34% on the week and 2,850% over the trailing 12 months.3 That extraordinary move reflects a semiconductor memory cycle that has shifted from oversupply to a structurally different demand profile driven by AI workloads. SK Square (2.5%) provides additional exposure to the Korean semiconductor ecosystem through its stake in SK Hynix.4
Cloud, Software, and the Productivity Layer
WTAI also holds positions that reflect what could be the next phase of the story, the one that may be still emerging but becoming increasingly legible.
Amazon (4.2%), Alphabet (3.4%) and Meta (3.6%) are hyperscalers whose capital expenditure commitments are, in a very real sense, the demand signal that the rest of the fund’s holdings are built around. Their Q2 results demonstrated that inference revenue is accelerating, even as only 2% of S&P 500 Index companies have yet quantified the impact of AI on their own earnings. The hyperscalers are the ones selling AI productivity, and their revenue growth suggests enterprise adoption is building, even if the productivity return on investment (ROI) has not yet shown up clearly in their customers’ income statements.
Oracle (3.0%), Snowflake (2.8%), Cloudflare (1.4%) and Datadog (0.9%) represent the data infrastructure and platform layer, the companies that govern how enterprises store, move, query and secure the data that makes AI models useful. Adobe (1.1%) has been systematically integrating generative AI into its creative workflows. ServiceNow represents a company whose workforce has meaningful AI automation exposure; it is a position WTAI holds through its broader software allocation. CrowdStrike (1.2%) anchors the cybersecurity dimension, a non-negotiable component of any enterprise AI deployment at scale.
For compute infrastructure beyond the public cloud hyperscalers, CoreWeave (1.3%) and Nebius Group (1.2%) provide exposure to the independent AI cloud layer, GPU-as-a-service providers serving AI labs, model developers and enterprises that cannot get sufficient capacity through the hyperscalers alone. DigitalOcean (1.0%) extends that theme toward mid-market and developer-focused cloud compute.
Wiwynn (1.2%) in Taiwan deserves mention as a key original design manufacturer (ODM) server manufacturer supplying AI-optimized rack systems directly to hyperscalers, a company that sits squarely at the intersection of the infrastructure buildout and the Asian supply chain that makes it possible. ASE Technology (1.1%), Elite Material (1.0%) and Unimicron Technology (1.0%) similarly reflect the critical role of Taiwan’s advanced packaging and printed circuit board (PCB) supply chains in enabling next-generation AI hardware. Tower Semiconductor (1.0%), added in the most recent rebalance, expands the fund’s exposure to specialty analog and mixed-signal foundry capacity relevant to AI system design.
SoftBank Group (1.4%) rounds out the picture differently, as a major investor in AI infrastructure globally, including through its Vision Fund holdings and its ownership of Arm Holdings, SoftBank is a portfolio-level expression of the AI capital formation thesis.
Why the Infrastructure Thesis Holds
Investors have focused almost exclusively on AI infrastructure stocks rather than potential AI productivity beneficiaries, and the returns thus far in 2026 signal they have been correct to do so. Companies involved in the infrastructure boom have delivered large, visible, near-term earnings. Companies that have talked about AI productivity initiatives, by contrast, have roughly matched the broader market.
This dynamic reflects the difference between revenue that is already in the income statement and revenue that depends on successfully implementing AI at enterprise scale and demonstrating measurable ROI.
The infrastructure spending is happening now, at scale, with auditable results.
The productivity payoff is coming, but it has not yet translated broadly to earnings.
The AI inference expense data tells the same story. Spending has accelerated sharply, growing from roughly $5 per employee per month at the start of 2026 to $12 in July for the median company, but it still represents less than 0.5% of S&P 500 Index revenues.
WTAI is positioned at the layer of the AI value chain where earnings are confirmed, spending is accelerating, and bottlenecks are identifiable and addressable.
The semiconductor companies, the equipment makers, the optical interconnect providers, the power infrastructure names, these are not bets on which enterprise will figure out AI productivity first. They are positions in the companies that must be paid before any of that productivity can be unlocked.
1 Source for data in this piece, unless otherwise stated: Snider, B., Hammond, R., Ma, J., Chavez, D., Jayachandran, K., & Sung, C. (2026, August 14). What Q2 earnings reports signaled about the state of corporate AI adoption (US Weekly Kickstart). Goldman Sachs Global Investment Research.
2 The percentages following each company name refer to the announced targeted weights for the WisdomTree Artificial Intelligence & Innovation Index, which WTAI is designed to track before fees and expenses. Data was posted as of August 13, 2026 and will be effective on August 21, 2026. Holdings subject to change.
3 Data referenced as of August 14, 2026.
4 Source: SK Square. (2026, August 13). Shareholding structure & NAV. Largest single shareholder of SK Hynix at approximately 20% as of June 30, 2026.
Comments
Log in or sign up to join the conversation.