Goldman's AI pitch is really a case for the active wrapper

Goldman Sachs Asset Management brought its AI investing case to an ETF audience in a VettaFi webinar segment that ETF Trends wrote up on September 10, with VettaFi head of research Todd Rosenbluth and research analyst Ben Hernandez hosting and GSAM's Katherine Bordlemay and Brook Dane doing the talking. The segment ran under the title "The AI Infrastructure Buildout: Where Investors May Find the Next Opportunities with Active ETFs," which carries the structure of the argument: artificial intelligence is the theme, the active ETF is the product, and the hour made as strong a quantitative case for active management as the industry has seen since 2009 while also showing the condition under which that case fails.

Bordlemay asked listeners to take away three points: active management matters more in a market defined by AI and disruption, innovation is the largest wealth-creation opportunity available even as it threatens countless incumbents, and GSAM sells active ETFs built for a market where leadership is broadening and index results are distorted. The quantification that followed was the most quotable stretch of the hour, with stock-level dispersion inside the S&P 500 running above 70%, which she described as the widest reading since 2009, and rates no longer low while AI and geopolitics compound the disruption.

The number an allocator should test came next: 20% of active managers beat their large-cap growth benchmarks over the last five years, and almost 60% have in what Bordlemay called the current era. Together they make one sales argument: the market has stopped paying for the average, so an investor should pay a manager to own something other than the average.

The second takeaway disciplines the first: if innovation creates wealth and destroys incumbents, an AI strategy is a wager on which side of the line a company sits, and wide dispersion is as much a warning about owning the disrupted as an advertisement for owning the disruptors.

Dane supplied the map, sorting the opportunity into three semiconductor buckets—large GPU names, ASIC chips, and CPUs—and tracking where the constraint sits, from memory earlier this year to optical networking now. Cybersecurity firms stand to benefit, a trend he tied to what the segment called the Hugging Face incident, and on spending he argued the headline capital expenditure has further to run, with the better question being where those dollars land, not how big the total gets.

The 20-to-60 jump deserves more scrutiny than a webinar can give it, because the comparison sets a five-year window against an undefined current era and measures both against large-cap growth benchmarks—the category where the largest names have been doing the most work if Bordlemay's dispersion point holds. A hit rate against one benchmark family is evidence about that family, not about active management in general, and it sits awkwardly beside the dispersion figure, since dispersion above 70% usually traces to a narrow set of winners carrying the index average and suggests the mechanics that made indexing hard to beat are still in place. The two claims point the same direction only if a manager holds weights that look very different from the benchmark, a portfolio construction commitment more than a market observation.

The framework is also directional in a way that matters for construction: naming three semiconductor buckets implies they will not be held in fixed proportion, and calling optical networking the new constraint asserts the bottleneck has moved off memory and will move again. An index rebalances into those shifts after the fact, which is the practical case for the wrapper in a supply chain that reprices by segment. The counterweight is capacity: the narrower the bottleneck, the fewer liquid names express it, and at a firm reporting $2.65 trillion in regulatory assets per ETF's records, that becomes a constraint on the strategy more than an advantage for it.

The wrapper is the more interesting half of the story, because an active ETF removes a manager's ability to narrate a bad stretch—NAV prints daily, flows are visible, and the index comparison is made whether the manager wants it or not—which is a demanding format for a franchise that asks to be judged on process. GSAM has chosen that format across its ETF shelf, where the related reading in the same ETF Trends piece points to the GPIQ and GPIX income pair; AI dispersion and monthly income do not compete for the same buyer, though they do compete for the same wholesaler's calendar.

The shelf has been expanding in public: GSAM's 16-basis-point ultra-short active ETF is the instrument now under test by a Japanese 10-year yield above 3%, and the August 20 piece on the IAUI gold-income fund noted that Neos shareholders have been weighing a Goldman acquisition. A product priced at 16 basis points and a deal still being weighed leave the AI message carrying a heavy load for a franchise whose economics are set by fees and shelf space.

Dispersion at 2009 levels is a condition that mean-reverts, and the managers now beating large-cap growth benchmarks are winning in part because those benchmarks are concentrated. The wrapper migration is settled; managers still defending mutual fund shelves are conceding the next several years of flows, and GSAM sits on the right side of that, with AI as the vehicle, not the destination. A GSAM AI-infrastructure product that holds the three buckets at different weights, in a prospectus, would turn the dispersion slides into something an allocator can buy. Until one exists, the regime-change language is doing the work a fact sheet usually does.