The seeds of a dramatic shift in markets over the past two weeks that has seen investors turn on sectors deemed vulnerable to AI disruption, were planted several years ago.

That's according to Adrian Cox and Galina Pozdnyakova, analysts at Deutsche Bank, who explained three key reasons for a sentiment shift that has driven the Nasdaq Composite COMP 5.3% lower since Jan. 28.

Magnificent 7 tech stocks - Apple, Amazon, Alphabet $(GOOGL)$, Microsoft $(MSFT)$, Meta $(META)$, Nvidia and Tesla $(TSLA)$ have dropped 8.3% in that time frame, with software, wealth managers, commercial real-estate and logistics hardest-hit sectors.

The first reason for the shift, said the analysts, is that investors are no longer betting on AI lifting all tides, but instead trying to differentiate between sectors and potential losers.

And that's actually been under way since ChatGPT launched in Nov. 2022, when suppliers of physical picks and shovels began outperforming services and software firms. Investors are now selling in cases of even slight doubts.

A major cause for that differentiation has been recent practical launches of AI model makers, such as a legal plugin from Anthropic hitting related stocks, and Altruist's AI tax-planning feature that put wealth managers in the firing line. Logistics companies were pummeled last week by news of a freight-scaling tool from former karaoke machine maker Algorhythm $(RIME)$.

Read: The stock market is reflecting fears of an AI apocalypse for white-collar jobs

Cox and Pozdnyakova said much of the risk is overstated for now, noting "deep moats of entrenched workflows" for hard-hit software-as-a-service companies, as well as proprietary data. To survive though, they warn companies will need to find ways to meet demand for seamless AI, they said.

They see a second reason for the sell-off tied to an increasingly tough job of trying to value AI in public markets. That's partly over concerns just seven companies account for 30% of the S&P 500 index SPX, but also contrasting views that AI is both overhyped and that its agents are improving so fast that entry-level jobs are disappearing.

The final reason for the spotlight turning on those vulnerable sectors is its impact seems more imminent than before. They noted a viral essay by Matt Schumer, the co-founder and CEO of autocomplete tool company OthersideAI, warning most people would be "blindsided" by AI.

That's as OpenAI's two-year dominance is starting to come into question ahead of potentially a trillion-dollar IPO, with Anthropic on its heels and proving popular with coders, while China's AI models are said to be nearly as good and cheaper.

"All of this signals that AI models may indeed be becoming commoditized.If that is the case, the true value of AI may rest ultimately in applications that have yet to be invented. That unpredictability gives investors yet another reason to take a break from the frenzy," said Cox and Pozdnyakova.