Matthew Tuttle argues the AI trade is not a March 2000-style bubble top, but it is shifting from a broad “buy anything AI” phase to a more selective bottleneck trade. He prefers memory, photonics, space, energy infrastructure, and selected software/cyber names over broad software ETFs, and he says investors should size positions small because a sharp correction is possible even if the secular trend remains intact.
Watch on YouTube ›Get the market thesis, key claims, assets, contradictions, and follow-up questions from any financial video — then unlock a version personalized to your portfolio, watchlist, and favorite speakers.
This is an interview with Matthew Tuttle, founder and CIO of Tuttle Capital Management, about where the AI trade goes next and how to position around it. His core thesis is that the AI boom is still alive into 2027, but the easy money has moved on from broad exposure to a more selective hunt for bottlenecks. In his framing, the market first bought “anything related to AI,” then shifted to bottlenecks such as memory, photonics, and space, and now needs to keep “peeling the onion” into substrates, inference, token costs, and related infrastructure. He repeatedly says this is not a March 2000-style episode, though he concedes “we might be” in a bubble and that nobody can reliably call the top. A major part of his argument is that memory is currently the strongest bottleneck. …
Near term, he would lean into pullbacks in memory and other AI bottlenecks while avoiding broad software baskets and day-one IPO chasing. The immediate risk is a sharper correction if rates rise or if the latest AI winners get crowded and unwind.
Over the next few months, he expects the market to keep rotating within AI from obvious winners into suppliers, bottlenecks, and infrastructure names. The setup stays constructive unless hyperscaler capex rolls over or the rate backdrop turns decisively hostile.
Structurally, he sees AI as a multi-year industrial buildout rather than a finished bubble, with value accruing to bottlenecks, energy, and adjacent infrastructure. The lasting implication is that stock selection matters more than thematic exposure, because broad baskets may underperform the true enablers.
The AI trade has moved from buying anything AI-related to focusing on bottlenecks — memory, photonics, space, substrates, inference, and token costs.
Speaker describes an evolution in the AI trade and lists specific bottleneck areas investors should focus on now.
Memory prices are skyrocketing because AI demand is structurally strong, and Micron's beat is demand-related rather than one-time.
Speaker points to rising memory prices, Micron's 9% gain, and expectations being high but still beaten as evidence of sustained AI-driven memory demand.
The IGV software index is a poor investment in the AI era because it includes legacy software companies that will be crushed by AI.
Matthew argues that in the age of AI, investors need to separate software companies that benefit from AI from those that will be disrupted, and a broad index like IGV includes too many losers.
Do you think Bitcoin and the IGV index trading alongside each other are indicative of more downturns or corrections to come?
Matthew doesn't read much into them moving together. He thinks Bitcoin is not a leading indicator for the market — air has just come out of that trade. On software (IGV), he says many software companies will get crushed by AI and some will benefit; he'd stay away from IGV and be careful on software generally, though he likes cybersecurity names and trades CRM and NOW occasionally.
What companies do you think will not be cannibalized by AI software companies?
Matthew says Microsoft is a whole different category and will be fine. Palantir will be fine. Anything cybersecurity — you need more of it, not less — will be fine. He also trades NOW and CRM when they get beaten down and thinks they'll be fine. Beyond that he'd be really careful with software.
Why should people look into the memory space if AI will continuously innovate and require less compute power — shouldn't we be shorting memory stocks instead?
Matthew says right now memory is a bottleneck. He is not sure it will always be one, but until it ceases to be a bottleneck it is an area he wants to be in. He notes they have three different memory ETFs plus a 2x long SanDisk fund. He advises watching position sizing (1-2% allocations) so a 50% correction would not be a big deal.
Unlock the full claims, asset map, scores, related transcripts, follow-up questions, and AI chat — shaped around your portfolio, watchlist, favorite speakers, and risks.
