Goldman Sachs models $765 billion in annual AI capital spending for 2026, climbing toward $1.6 trillion by 2031. That kind of multi-year buildout is the backdrop for three actively managed (or active-leaning) AI funds that have separated themselves from the passive crowd: the iShares A.I. Innovation and Tech Active ETF (NYSEARCA:BAI), the iShares Future AI & Tech ETF (NYSEARCA:ARTY), and the Dan IVES Wedbush AI Revolution ETF (NYSEARCA:IVES).
Each takes a different route to the same thesis. BAI runs concentrated active bets from BlackRock’s portfolio managers. ARTY tracks a curated thematic index but lets the methodology lean into emerging AI names rather than legacy tech. IVES is the outlier: a portfolio shaped by a single Wall Street analyst’s conviction list. With BAI up roughly 35% year to date and ARTY up about 32%, the active and active-leaning approaches have meaningfully outpaced broad tech benchmarks during the latest leg of the AI infrastructure cycle.
Why “barely started” is more than a slogan
The hyperscaler capex curve is still bending upward. Independent research outlets estimate roughly $690 billion to $700 billion in combined 2026 AI infrastructure spending from Microsoft, Amazon, and Alphabet alone, with about three-quarters of aggregate hyperscaler capex earmarked for AI workloads. Power generation, networking silicon, optical interconnects, and data center real estate are all being scaled simultaneously. An investor who believes the build is closer to the second inning than the seventh wants exposure that captures the next wave of beneficiaries beyond last year’s winners. That is where active selection earns its fee.
BAI: the active conviction play that is leading the pack
BAI is the cleanest expression of active management in this group. Run by iShares with a net expense ratio of 0.55%, it gives portfolio managers latitude to size positions based on conviction rather than index weights. The result is a top of the book that looks decisive: NVIDIA near 9% and Broadcom near 8% together account for roughly a sixth of the fund, with Alphabet around 5% and Microsoft around 4% rounding out the mega-cap core.
What makes the portfolio interesting is what sits underneath those names. Snowflake near 3%, Celestica near 3%, Fabrinet near 3%, and Tower Semiconductor near 3% are not the holdings you find at meaningful weights inside a Nasdaq-100 tracker. Celestica and Fabrinet are AI hardware supply chain plays. Tower is a specialty foundry. Snowflake is a software bet on enterprise AI data layers. Active managers can lean into those second-tier beneficiaries when conviction is high, which is the structural reason BAI has put up the strongest return in this group, with the fund gaining roughly 79% over the trailing year.
The tradeoff is concentration. When 16% of the fund sits in two semiconductor names, a single bad guidance update from NVIDIA or Broadcom hits harder than it would in a diversified tech index. Investors who want active conviction need to be comfortable owning that risk explicitly.
ARTY: thematic indexing with a forward-looking tilt
ARTY is technically passive, but it earns a spot on this list because the index it tracks behaves nothing like a generic tech ETF. The fund (formerly iShares Robotics and Artificial Intelligence Multisector, rebranded in August 2024) now tracks a Morningstar global index focused on companies driving future AI and tech, and it carries a 0.47% expense ratio with roughly $3.1 billion in total assets.
The portfolio reads as a deliberate bet on the AI supply chain rather than the platform giants. Recent top weights include Marvell Technology around 6%, AMD near 6%, CoreWeave around 5%, and Micron near 5%. CoreWeave is the standout signal here: a GPU cloud operator that did not exist as a public name a few years ago now ranks among the top holdings, alongside memory and custom silicon suppliers. The index methodology refreshes those weights as the AI ecosystem evolves, which is closer to active spirit than a market-cap tech fund.
The tradeoff is that ARTY’s global reach and equal-leaning weights mean less direct exposure to the very largest US mega-caps. If the next leg of AI returns is concentrated in NVIDIA and Microsoft, ARTY will lag funds like BAI. If it broadens to the picks-and-shovels names, ARTY is positioned for it.
IVES: the analyst conviction wildcard
IVES is the unconventional pick. The portfolio is built around Wedbush analyst Dan Ives’ AI Revolution thesis and crossed $1 billion in assets within months of its June 2024 launch. The fund carries a 0.75% expense ratio, the priciest in this group, which is the cost of buying an analyst’s specific stock list in fund form.
The holdings reflect Ives’ published views rather than a quant screen. The top of the book sits at AMD near 7%, Micron around 6%, Alphabet around 5%, and Broadcom around 5%, but the more telling positions are further down: Palantir, CoreWeave, Nebius, Innodata, and GE Vernova all appear in meaningful size. That mix combines the AI software layer (Palantir), neocloud infrastructure (CoreWeave, Nebius), data labeling (Innodata), and the power generation angle (GE Vernova) that capex-heavy data centers ultimately need.
IVES has lagged this year, with a 13% year-to-date gain, because several of its conviction names trade with higher beta and have been more exposed to the recent capex-digestion debate. The fund’s relatively concentrated 31-holding portfolio means a few names drive most of the outcome, which is the point: investors buy IVES specifically to own Ives’ calls in fund form.
Picking between them
The three funds map cleanly onto different investor profiles. BAI fits the investor who wants the largest AI infrastructure leaders sized by professional conviction and is willing to live with two-name concentration at the top. ARTY suits the investor who believes the next phase of returns will broaden into the supply chain (memory, custom silicon, GPU cloud) and prefers a rules-based portfolio that refreshes as the theme evolves. IVES is for the investor who specifically wants Dan Ives’ published thesis expressed in fund form and is willing to pay the highest expense ratio in the group for that single-analyst conviction.
The shared risk is straightforward. If hyperscaler capex growth slows, if AI revenue fails to keep pace with infrastructure spend, or if the market decides to compress the valuation multiples on AI-exposed names, all three funds will draw down together. The case for owning any of them rests on the longer-term spending curve continuing to bend up over multiple years. For investors who believe the buildout is still early, the choice comes down to which selection process they want managing their AI exposure.
Top Gaining Stocks
Top Losing Stocks