Most capital chasing the generative AI theme has flowed into chips. Semiconductor funds have absorbed the bulk of inflows, and the price action shows it: the iShares Semiconductor ETF (NASDAQ:SOXX) is up 87% year to date. The software layer, where applications, model providers, and enterprise platforms turn raw compute into recurring revenue, has trailed that narrative. For investors looking specifically at that second wave, three ETFs stand out: Invesco AI and Next Gen Software ETF (NASDAQ:IGPT), Roundhill Generative AI & Technology ETF (NYSEARCA:CHAT), and iShares Future AI & Tech ETF (NYSEARCA:ARTY).
Each fund routes investor capital into the software side of generative AI through a different mechanism. IGPT carries an explicit software mandate baked into its index, while CHAT is an actively managed, concentrated bet on the names building and selling generative AI products. Finally, ARTY blends emerging AI software startups with semiconductor exposure, sitting closer to a full-stack approach. Sorting which one fits a given portfolio comes down to how pure a software tilt the investor actually wants.
Why the Software Layer Has Been the Quieter Trade
The chip-buying phase of the AI build-out is visible in capex disclosures and earnings calls. Outlooks from Goldman Sachs, JPMorgan, PineBridge, and Vanguard for 2026 all describe AI infrastructure as the dominant growth engine, with hyperscalers investing heavily in datacenter expansion. PineBridge notes that software remains a “show me” story, with AI-driven disruption and the potential reallocation of IT budgets posing persistent concerns. That hesitation is the setup. Once enterprise budgets reallocate toward AI-enabled software, the funds carrying those names benefit from a layer of the stack that has not yet been fully priced in.
Ultimately, it’s the hardware-heavy funds like SOXX that capture the first leg, but it’s the three ETFs below that capture the second.
IGPT: The Most Literal Software Bet
IGPT tracks the STOXX World AC NexGen Software Development index, and the index methodology reflects that label. Where most AI ETFs mix chip designers, foundries, and packaging companies into the portfolio, IGPT’s index methodology screens for next-generation software developers, meaning the cloud platforms, data analytics vendors, cybersecurity companies, and enterprise software names that embed generative AI into products customers already pay for. That is the cleanest mechanism for capturing budget reallocation toward AI-enabled software.
The performance picture lines up with the thesis. IGPT is up 64% year-to-date and 119% over the past year, with shares trading around $97. Over five years, the fund has roughly doubled, returning 97%, while the ten-year figure of 618% reflects its longer history as an Invesco software-themed product before its current branding.
The tradeoff is concentration in a single layer of the stack. If the AI software narrative stalls and capex spending continues to flow primarily to chips, IGPT lags funds that hold both, but the fund has no meaningful semiconductor exposure to offset that.
CHAT: Actively Managed, Pure-Play Generative AI
CHAT was the first ETF launched explicitly around the generative AI theme, debuting in May 2023 shortly after ChatGPT pushed the category into the mainstream. Roundhill actively manages the portfolio rather than tracking an index, which allows the manager to weight names like Meta, Microsoft, and Alphabet at the top of the book. CHAT is the purest software-tilted Gen AI ETF with NVIA, AMD, and Alphabet at the top. Those three are the public-market proxies for closed-source frontier models, internal AI assistants, and the infrastructure layer that monetizes them through cloud subscriptions.
The active mandate matters more here than for the other two funds. Generative AI as a category did not exist in any meaningful index a few years ago, and the universe of relevant companies continues to shift quickly as private model labs go public and incumbents pivot. An active manager can rotate without waiting for an index reconstitution.
Performance reflects the concentrated approach. CHAT is up 60% year-to-date and 133% over one year, with shares near $93. On the plus side for investors, since its 2023 inception, the fund has returned a whopping 272%. The tradeoff is that active management on a narrow theme carries higher fees than index alternatives, and the concentrated book amplifies single-name risk. A bad quarter from Microsoft or Alphabet hits CHAT harder than it hits a broadly diversified tech fund.
ARTY: The Hardware-Software Hybrid
ARTY is the most diversified of the three and the one that most resembles a full AI-stack fund. The iShares product holds roughly $2.08 billion in net assets as of March 31, 2026. The portfolio is hardware-heavy at the top: TSMC, Marvell, NVIDIA, AMD, Broadcom, and Micron together account for a large share of the book, with TSMC at 5% and Marvell at 5% of net assets.
The reason ARTY captures meaningful software-layer exposure is that it appears just beneath its primary hardware anchors. CoreWeave leads this secondary block at 4.36%, while enterprise pillar Oracle commands a 3.91% position. This allocation strategy purposefully pairs infrastructural cloud power with pure-play artificial intelligence applications, ensuring investors capture the actual downstream deployment of compute capacity. By keeping a firm hand on enterprise software alongside high-margin services, the fund avoids the structural hazard of betting entirely on high-flying infrastructure. It grounds the portfolio with companies generating visible, subscription-based free cash flow.
ARTY is up 49% year-to-date and 98% over one year, trading near $71. The trailing performance lags both IGPT and CHAT in 2026, which is exactly what hybrid exposure yields when one layer of the stack leads.
The tradeoff is the hardware ballast. Investors who want clean software-layer exposure end up with diluted ownership of the chip names that dominate AI capex narratives. For investors who want a single fund covering both sides, that blend is the feature.
Choosing Between the Three
The three funds map to three distinct investor profiles. Starting off, IGPT fits the investor who has already taken chip exposure elsewhere and wants a dedicated allocation to AI-enabled software with no semiconductor overlap. CHAT fits the investor who wants concentrated active exposure to the largest public-market beneficiaries of generative AI and is comfortable paying for active management on a narrow theme. ARTY fits the investor who wants a single fund covering both the chip and software layers, accepting that hardware weighting will dominate near-term performance.
The broader point is that the chip trade has run hard, with SOXX up 183% over the past year, while the software layer is still earlier in its repricing. The three funds above are the most direct ways currently available to express that view.
Top Gaining Stocks
Top Losing Stocks