AI ETFs: Beyond Chips & Hyperscalers

While AI investing initially centered on hyperscalers and chips, the opportunity has been broadening as infrastructure grows. Opportunities are moving further down and across the supply chain — into memory, data center power, components, and robotics. The ~15 AI and tech-related ETFs launched in September alone emphasize the growing opportunities for targeted AI exposure.Key Takeaways: AI investing is moving beyond hyperscalers and semiconductors toward increasingly specialized segments across the broader AI ecosystem. Around 15 AI- and tech-related ETFs launched in September alone, targeting memory, power infrastructure, electronic components, robotics, and more. DRAM’s rapid success demonstrated demand for niche AI exposure, encouraging issuers to pursue increasingly granular themes. Semiconductors Still Lead AI DemandInvestor interest in AI has remained strong, but some of the clearest demand has shown up in semiconductor ETFs. The iShares Semiconductor ETF (SOXX B) and the VanEck Semiconductor ETF (SMH B) have brought in year-to-date net inflows of approximately $11.1 billion and $8.7 billion, respectively, although shorter-term demand has fluctuated. From a volume perspective, semiconductor ETFs have also been high volume. Looking at 30-day average volume across all U.S.-listed ETFs, the top ten currently include four AI names: the GraniteShares 2x Short NVDA Daily ETF (NVD A-), the Tradr 2X SNDK Long Daily ETF (SNXX), and the Direxion Daily Semiconductor Bull 3X ETF (SOXL B).AI Extends Beyond Chip CompaniesAI as a theme beyond the semiconductor industry isn’t necessarily new. ETFs like the ROBO Global Artificial Intelligence ETF (THNQ B-) were launched in May 2020. The rapid rise of the Roundhill Memory ETF (DRAM) demonstrated just how much investor demand existed beyond traditional semiconductor exposure. Launched in April, DRAM crossed $1 billion in assets in 10 trading days and reached $10 billion within six weeks — making it the fastest ETF on record to hit the latter milestone. Capital kept flowing even as market volatility picked up. By July, DRAM had accumulated roughly $26 billion in net inflows, placing a niche, specialized fund alongside the industry’s largest flagship products on monthly flow leaderboards. That success predictably sparked competition. DRAM initially offered one of the few direct access points to global memory leaders like Samsung, SK Hynix, and Micron; by mid-summer, several rival issuers were preparing competing products. This rapid response illustrates a new ETF playbook: isolate a critical, underrepresented link in the AI supply chain and package it into a targeted pure-play fund. The Tema Photonics & Optical ETF (LAZR) is another example of the shift. Launched in late June, the actively managed LAZR was one of the first in the space and surpassed $100 million in assets in under three months, proving investor willingness to move beyond processing chips toward the optical technologies which move data between servers and data centers. The segment quickly grew to include at least six ETFs including passively managed ETFs like the Aura AI Photonics ETF (PHOX).Large Wave of AI & AI Adjacent ETFs Launched in SeptemberThe third quarter brought a lot of new ideas like MANGOS, but taking a smaller more manageable slice of September it appears that there around 15 ETFs launched in the AI and robotics space (not including any single stock ETFs).Memory Continues to Gain AttentionDRAM’s success encouraged issuers to find more granular ways to access the memory landscape. In September, Yorkville America introduced the Next Generation Memory Index ETF (NRAM) which targets companies associated with the design, development, manufacturing, and supply of physical memory, including advanced semiconductor memory technologies. Defiance took a targeted regional approach with the Defiance China Memory ETF (CRAM), the first U.S.-listed product dedicated specifically to China’s memory supply chain. Rather than limiting exposure to pure-play fabricators, CRAM spans 11 distinct nodes of the value chain including controllers, memory modules, packaging, testing, and production equipment — offering comprehensive exposure to China’s semiconductor push.MLCCs & Electric ComponentsMultilayer ceramic capacitors (MLCCs) are critical passive components that stabilize power delivery across circuit boards, and represent another hyper-targeted AI infrastructure theme gaining traction. On September 9, Roundhill launched the Roundhill MLCC & Electronic Components ETF (CCML) to capture key Asian suppliers across Japan, South Korea, Taiwan, and China, where component density per server is growing rapidly to meet AI power demands. Soon after, Global X launched the MLCC & Electronic Components ETF (MLCC). Global X cites projections that AI-driven MLCC demand could surge 4.3x between 2025 and 2030 against a highly concentrated supply base. This ETF takes an actively managed approach that broadens exposure to adjacent passive components like chip resistors and inductors.Data CentersThe Harbor 800VDC AI Datacenter Ecosystem ETF (HUMM) pushes specialization even further by isolating the shift toward next generation 800-volt direct-current (800VDC) power architecture in data centers. As high-density AI clusters strain existing facility design, legacy lower-voltage distribution systems suffer from energy loss, thermal bottlenecks, and physical space constraints. Upgrading to an 800VDC architecture allows facilities to deliver significantly higher power density with greater efficiency. HUMM targets the entire hardware supply chain enabling the transition including power semiconductors, conversion equipment, high-voltage connectors, cooling systems, and electrical infrastructure.Robotics ActuatorsOn September 15, Defiance launched the Defiance Robotics Actuators ETF (AT) to capture pure-play exposure to precision-motion systems including motors, gearboxes, and encoders. According to McKinsey, actuators represent the primary cost driver in humanoid robotics, accounting for an estimated 40% to 60% of a robot’s total bill of materials. The majority of holdings in this ETF are Asian stocks (e.g., Japan, China, South Korea). KraneShares followed later in the month with the KraneShares Actuator ETF (TORK), which similarly targets the “muscles and joints” that convert digital AI commands into physical movement. TORK provides global exposure to component makers across key manufacturing hubs in Japan, South Korea, Taiwan, and China, spanning electric motors, controllers, speed reducers, and sensors.Other AI BeneficiariesIn contrast to hardware and infrastructure launches, the iShares Future AI Beneficiaries ETF (AIBF) shifts from AI enablers to AI adopters. Rather than holding tech companies building models or hardware, AIBF targets U.S. corporations leveraging AI to drive operational efficiency, margin expansion, and top-line growth like Walmart (WMT), Costco (COST), and Abbvie (ABBV). This strategy shifts exposure beyond traditional technology into non-technology sectors such as healthcare, industrials, and consumer goods.Bottom Line:The AI trade is broadening to included more targeted subsectors now appearing in ETFs. These ETFs, while small, allow access to more niche equity securities including international equities and more specialized themes in the AI space. These are best used as targeted thematic complements instead of a replacement for a broader technology allocation. For more news, information, and analysis, visit the Artificial Intelligence Content Hub. VettaFi LLC (“VettaFi”) is the index provider for THNQ and PHOX, for which it receives an index licensing fee. However, THNQ and PHOX are not issued, sponsored, endorsed, or sold by VettaFi, and VettaFi has no obligation or liability in connection with the issuance, administration, marketing, or trading of THNQ or PHOX.