Wall Street has rarely been louder with bubble warnings on the AI trade after the parabolic moves we've recently seen in semiconductors like Micron Technologies Inc. (MU) or Sandisk Corp. (SNDK).

But Jordi Visser, head of AI Macro Nexus Research at 22V Research, believes investors obsessing over speculative excess may be misreading the cycle entirely.

Now he is telling them which part of the AI trade is about to take the baton.

The AI buildout is moving into a new phase, one where the early-cycle semiconductor winners hand off to the late-cycle physical bottlenecks: power, chemicals, and silver. Visser is putting his own portfolio behind that view.

"I've now sold out of two-thirds of my Micron," Visser said in a video posted on Sunday. "I still think it's going higher, but I think there's other bubbles and parabolas I'd rather be involved in."

The AI Cycle Has A New Map

Visser's framework breaks the AI economy into five layers, with applications and models at the top. Below that sit data infrastructure and chips, and at the base, energy, hardware, and commodities.

He calls it the "five-layer AI cake," a structure he has used to build thematic portfolios since the agentic phase of AI took off in late November.

The opening leg of that cycle, in his telling, belonged to memory, advanced packaging, optical fiber, and racks — the layers companies hoarded first as compute demand exploded.

That phase has now matured.

"Right now, my focus has been on the companies in the first three themes because they are early and mid-cycle," Visser said in an email. "However, now I believe the bottlenecks for power are going to dominate while inflation picks up, so I am more interested in commodities and chemicals due to being more late cycle in the AI cycle," he added.

Why Silver Could Be The Next Micron

Visser has been bullish on silver as a structural input to the AI buildout for months.

The metal — tracked by the iShares Silver Trust (SLV) — sits near $80 per ounce, down roughly 14% from its peak ahead of Gulf War III in late February but still up more than 140% year-over-year.

The recent pullback, in his view, looks like the same setup that preceded the memory rally.

"Silver is very attractive to me for the reason I just mentioned from the last cycle," Visser said in the interview. "Silver has lagged behind this recent run up in DRAM due to its overshoot last year. I think now that we are moving into the late cycle in my work of the AI cycles, I look for it to play catch up," he added.

The inflation backdrop is the second leg of the thesis. April Consumer Price Index (CPI) lands Tuesday with consensus near 3.7% year-over-year. The 3-month Treasury bill yielded 3.69% on May 8.

"This week we are likely to see CPI YoY be higher than 3m bill yields for the first time since 2023," Visser said. "I think we see a regime shift towards more inflation investments for the second half of the year."

Negative real rates on cash, in his model, force the rotation. That regime favors silver, gold, and Bitcoin (BTC) as core holdings rather than satellites.

According to Visser, a Warsh-led Fed will refrain from hiking interest rates. "I do not think it is likely the Fed will raise rates although I do expect the pressure will grow at a time that the Fed is very polarized and dissention in views seems to be very high," he said.

Why Power, Why Chemicals Now?

Caterpillar Inc. (CAT) closed its most recent quarter with a record $62 billion backlog and guided to power-generation equipment sales tripling by 2030. Sterling Infrastructure Inc. (STRL), Fluence Energy Inc. (FLNC), Vistra Corp. (VST) and Eaton Corp. (ETN) sit in the same trade, along with the broader utilities and transmission grid complex.

Chemicals is the layer Visser believes is most underappreciated — specialty inputs for advanced packaging, optical fiber tubing, batteries and the AI upgrade cycle across autos, phones and appliances.

Visser has called The Chemours Company (CC) one of his highest-conviction names in the chemicals layer. The stock is already up 88% year to date.

When Does The Trade Rotate Back To Software?

When asked whether software stocks represent an opportunity after the selloff, Visser argues software stocks broadly remain a poor use of capital relative to the physical buildout, but he is selective.

He pointed to Cadence Design Systems Inc. (CDNS), Synopsys Inc. (SNPS), Palantir Technologies Inc. (PLTR), and Oracle Corp. (ORCL) as exceptions because they sit closer to the physical compute layer than traditional SaaS models do.

"I think SaaS seat-based models are a bad use of mental investment energy at this point," Visser said in the interview. "There is 90 trillion-dollar physical AI buildout, according to Nvidia Corp. (NVDA)'s CEO Jensen Huang. Why do people need to find a cheap SaaS company when we just started this buildout?" he added.

According to Visser, software companies leveraged to enterprise compute and the AI agent rollout are working. Traditional SaaS, where revenue is anchored to a seat count that AI agents are now eliminating, is not.

The iShares Expanded Tech-Software Sector ETF (IGV) has spent most of the year below its 200-day moving average. That divergence is the answer.

What's The 'Benchmark Arbitrage'?

The thread connecting all of Visser's calls is what he labels "benchmark arbitrage" — the structural mismatch between an index built for the software age and an economy being rebuilt for the AI age.

For the last fifteen years, he noted, the dominant investment phrase was Jeff Bezos's line, "your margin is my opportunity."

The new phrase, in his framing, is "your CapEx is my opportunity."

The hyperscalers — Microsoft Corp. (MSFT), Amazon.com Inc. (AMZN), Alphabet Inc. (GOOGL) and Meta Platforms Inc. (META) — are guiding to roughly $725 billion of combined capex in 2026, with contracted cloud backlogs north of $2 trillion.

What changes, in his view, is who receives the marginal dollar.

The receivers are not the Magnificent Seven that built moats on code in the 2010s. They are the companies selling power, copper, silver, chemicals, advanced packaging, optical fiber, and grid equipment to everyone else trying to scale intelligence.

According to Visser, passive funds are mechanically anchored to old weightings — software, consumer staples, financials, large-cap services — that no longer reflect where economic value is being created.

Every active manager benchmarked to the S&P 500 is, by definition, underweight the names actually driving the market higher.

As the agentic stage of AI accelerates the gap between benchmark weights and where capital should sit only widens.

The first leg paid out through Micron and Nvidia. The next leg, in his view, runs through silver, power, and chemicals — and the bottlenecks that come with them.