The AI sell-off today looks more like a reset in expectations than a change in the long-term AI thesis.
Bank of America called the recent AI controversy around Anthropic, OpenAI and frontier-model development “noise” within what it views as a secular AI bull market.
The biggest point: AI companies may want the industry to slow down, but in a highly competitive market, nobody wants to be the first to slow down.
The competition is happening on multiple levels—U.S. vs. China, hyperscalers vs. specialized cloud providers, and AI labs competing for customers, talent and technological leadership. That makes a synchronized slowdown extremely difficult.
More importantly, the physical demand for AI infrastructure remains strong. Network capacity is reportedly operating near full utilization, while even older-generation accelerators are seeing strong rental demand. If demand were truly collapsing, we would expect pricing and utilization to deteriorate—not remain this tight.
That’s why I’m not reading today’s weakness as “AI is over.”
I’m looking at it as an opportunity to become more selective.
The semiconductor group has already experienced a meaningful correction, while earnings growth remains strong. BofA expects AI capex could potentially reach $3T+ annually by 2030, driven by continued demand for compute, networking and infrastructure.
Names I’m watching closely on weakness:
$NVDA — AI compute leader
$MRVL — networking/connectivity exposure
$MU — HBM and memory demand
$INTC — turnaround + advanced manufacturing opportunity
$AMD — AI accelerators and data center growth
$LRCX — semiconductor equipment exposure
$ADI — analog/industrial semiconductor exposure
I especially like the idea of buying quality companies after fear-driven pullbacks, rather than chasing them after big runs.
The AI trade will continue to face headlines about capex, valuations, power constraints, regulation, competition and potential overspending.
That’s normal in a secular growth cycle.
For me, the important question isn’t whether there will be more AI fear.
There will be.
The question is whether AI demand, infrastructure utilization, earnings and hyperscaler spending continue to grow.
So far, the underlying demand remains strong.
I’m staying selective, keeping cash available, and looking for opportunities on weakness rather than trying to predict every headline.