Strategist Defends Semiconductor Rally Despite Bubble Fears: ‘Price Might Be Understanding Something You Don’t Quite Get’

For much of the past year, Wall Street has debated whether the semiconductor rally is a once-in-a-generation earnings cycle or bubble. The Compound and Friends podcast turned that into a clean event-versus-precedent debate, with Josh Brown invoking Cisco Systems (NASDAQ:CSCO | CSCO Price Prediction) as the 2000 ghost in the room and Fidelity strategist Denise Chisholm pushing back with a line that has bounced around chip desks all week: "price might be understanding something you don’t quite get." But the data she brought does more work than the quote. Her pattern study lands on the opposite conclusion from what a casual look at the charts would suggest.

The SOX index, per the panel, is sitting 69% above its 200-day moving average, which the show called unprecedented over the last 25 years. Semiconductors as a group are up roughly 100% over the past year. NVIDIA (NASDAQ:NVDA) is up 62% over the past year and carries a $5.4 trillion market cap. Broadcom (NASDAQ:AVGO) is up 95% in twelve months. Micron Technology (NASDAQ:MU) has vaulted 986% over the same window, which Brown framed by noting the stock moved from the 127th largest S&P 500 name a year ago to number 11 today.

The 2000 Mirror

The precedent Brown reached for is the one every veteran knows. Cisco was briefly the most valuable company on the planet in March 2000, traded at a triple-digit multiple, then surrendered most of its value over two years. Cisco today sits at a $504.5 billion market cap with a trailing P/E of 43 and a forward P/E of 26, and even after a 68% year-to-date run, it remains a reminder that infrastructure-buildout darlings can take decades to revisit a parabolic top. The fiber that fueled Cisco’s 1999 boom sat unlit in the ground for years. For a deeper look at how today’s AI leaders compare to that era, see our prior coverage of the AI infrastructure trade.

I’ve been watching this sector since the original GPU compute thesis took shape, and the Cisco parallel is the first thing any honest bull must address.

Chisholm’s Counter

Chisholm’s pushback rests on base rates. She argued that semiconductors are up roughly 100%, exactly in line with their earnings growth, and that historically, when an industry runs 70% to 100%, about 65% of the time it outperforms the following year. Layering valuation onto the pattern, she put the odds of semis outperforming over the next twelve months at roughly 70%.

The structural distinction she drew from 2000 matters. Micron’s memory "is put into use immediately," she said, contrasting that with the dark fiber problem two decades ago. Data centers, in her words, are "literally saying we’re short compute based on today’s demand." Micron just posted Q1 FY26 revenue of $13.64 billion, up 56.6% year over year in its most recent 10-Q filing, with non-GAAP EPS of $4.78 and a Cloud Memory unit running at a 66% gross margin. NVIDIA’s most recent quarter delivered $81.61 billion in revenue, up 85.2% year over year, and Data Center Networking grew 199%. Broadcom’s AI semiconductor revenue jumped 106% year over year to $8.4 billion, and CEO Hock Tan has staked out a goal of $100 billion-plus in AI sales by 2027.

The Valuation Gut-Check

NVIDIA trades at a trailing P/E of 34 and a forward P/E of 26, with a PEG of 0.69. Micron, even after its run, sits at a forward P/E of 11. Cisco in 2000 traded north of 150x earnings. This setup looks structurally different. UBS captured the bullish edge with its Micron price target of $1,625, raised from $535, more than double the $700.51 close cited on the show.

The speed remains a problem even inside the bull case. One portfolio manager on the podcast described watching Micron rise 5% daily, which historically precedes air pockets. Chisholm conceded semis are still cyclical, only suggesting the cycle may be stretching from roughly three years to seven. Cycles mean-revert harder when margins are at all-time highs, which is exactly where NVIDIA’s 75% non-GAAP gross margin and Broadcom’s 68% adjusted EBITDA margin sit now.

What the Pattern Says

The Long Memory verdict is clean. The 2000 precedent screams caution at the price. The base-rate study Chisholm cited says price has historically been a leading indicator of fundamentals catching up. Both can be partly right. Cisco was a real company that earned its way back. The internet still ate the world. Long term, the buildout of AI infrastructure looks more like rails and grids than unused fiber, and Wall Street has a way of paying off patient owners of pick-and-shovel cycles even when entry was uncomfortable.

What I’m watching now: whether order books stretching into 2027 hold up when hyperscaler capex guidance refreshes, and whether NVIDIA’s $119 billion in total supply commitments turns into delivered revenue or sticky inventory. The pattern favors the bulls. The speed of Micron’s chart says not to confuse a favorable base rate with a free lunch.