Fabrinet, Dycom and Qualcomm: 3 AI Fallen Angels Wall Street Just Marked Down

Key Points

• Fabrinet and Dycom Industries both beat earnings estimates by more than 10% and raised guidance, yet their stocks fell as investors focused on balance-sheet details instead.
• Marc Chaikin of Chaikin Analytics attributes the broader market pullback to elevated oil prices and a historical midterm-election pattern of weakness before a rally.
• Qualcomm's new chip joint venture with Amazon for AWS data centers could reset its stock trend, since all three companies remain profitable amid AI buildout risks.

Two weeks in, September is already doing what September usually does. The S&P 500 drifted lower toward its 21-day moving average, the AI trade lost its footing, and several companies that had just posted double-digit earnings beats got sold anyway.

That last part is the odd one. Fabrinet NYSE: FN beat estimates by more than 10% and guided higher. Dycom Industries NYSE: DY did the same and raised its outlook. Both stocks went down.

What changed was not the businesses. It was the bar. Expectations ran hot into July earnings, and once the numbers cleared, the market went hunting for something else to worry about, such as footnotes, short-term credit lines, and cash going out the door to build capacity that does not yet exist.

That combination—strong fundamentals, broken charts—is a specific kind of setup, and it only shows up when sellers have already done most of their damage.

Oil Prices and the Midterm Calendar Keep a Lid on Rallies

The pullback has a cause, and it is not complicated, per Marc Chaikin, founder of Chaikin Analytics. Energy prices have stayed stubbornly elevated, with crude pushing back toward triple digits as strikes on Saudi refining capacity and U.S.-Iran exchanges keep a risk premium in place. Refined products never came down at all.

Expensive energy makes rallies hard to sustain. Any unwelcome headline out of the Federal Reserve, Washington or the Middle East lands on a calendar that is already fragile.

Midterm election years have run to a pattern for more than a century: a late September or early October low, then a strong stretch of nine to 18 months once the election clears, per Chaikin. His team has been pointing subscribers at that window all year.

The Optical Layer That Moves the Data

Fabrinet builds the optical packages that move data around a data center floor. It recently beat estimates by more than 10% and guided higher.

The market chose instead to focus on the cash it is spending and the short-term credit lines it is drawing to get ahead of orders.

Sit with that for a second. The company got marked down for preparing to meet demand it can already see.

The stock traded above $700 in May and bottomed near $380 before bouncing, per Chaikin, who carries it as neutral-plus on the Chaikin Power Gauge: strong fundamentals, weak trend. Long-term debt is close to nil, the borrowing is short-term, and insiders have been buying.

Watch whether the electronic equipment and instrumentation group keeps its leadership, and whether the bounce builds a higher low instead of retesting.

The Unglamorous Layer of Digital Infrastructure

Dycom Industries handles specialty contracting for telecommunications and digital infrastructure, the physical work that turns capital expenditure (CapEx) budgets into working capacity.

It also beat estimates by more than 10% and raised guidance for the coming quarter and the full year.

It was still a sell-the-news quarter. The stock has given back more than 30% in the last three months at a market value near $9 billion and a fairly low price-to-earnings ratio.

The market is pricing a slowdown that has not appeared in the guidance. Chips from NVIDIA NASDAQ: NVDA have to sit somewhere, and somebody has to build the somewhere.

The risk is genuine. Data center construction has turned into a live political issue ahead of the midterms, with communities resisting projects sited near homes. The upside is that railroads and the interstate system met the same resistance and got built.

A Hyperscaler Deal Can Reset a Chipmaker's Trend

Qualcomm NASDAQ: QCOM is the outlier of the three. Its last quarter merely met expectations, which in this tape counts against you, and the stock sits more than 25% below its June peak. It traded under $140 in May, ran toward $250 on enthusiasm for a new class of chips, then gave the entire move back.

Then came a joint venture with Amazon.com NASDAQ: AMZN on customized silicon for AWS data centers. Chaikin calls it multi-generational, a platform rather than a single-cycle win, and notes the stock pushing back above its long-term trend line on the news, which would turn his Power Gauge rating bullish.

Analyst targets on Qualcomm imply less upside than on the other two names because estimates came down after that flat quarter. The gap is the opportunity, per Chaikin, since a design win with one of the three biggest data center operators forces the numbers to be revisited. Roughly $178 billion in market value against about $44 billion in revenue makes this an established business, not a story stock.

Earnings Are What Separate This Cycle From 2000

The dot-com comparison gets made constantly, but it misses something. Pets.com had no earnings, a broken model and too much debt. These three names print profits.

The vulnerability sits elsewhere in the AI space, with many companies carrying heavy debt and no earnings to service it. If OpenAI were public, it would not make the list, per Chaikin, and SpaceX (NASDAQ: SPCX) occupies similar ground: a compelling technology story still waiting on the money.

Meanwhile, Alphabet NASDAQ: GOOGL, Amazon, and CoreWeave NASDAQ: CRWV keep building, and everyday adoption keeps widening, which argues this buildout is early rather than finished.

The Feeders, Not the Headliners, Carry This Buildout

What ties Fabrinet, Dycom and Qualcomm together is more than a sector label. All three sell into the same construction cycle, all three cleared or matched their numbers, and all three got sold by a market that priced perfection in July and doubt in September.

The setup does not depend on AI demand accelerating. It depends on demand not collapsing and on hyperscaler CapEx landing near what has already been committed. It does not survive a genuine spending freeze, a cheap open-source model that guts compute requirements, or oil staying high long enough to squeeze the whole tape.

Stay focused on hyperscaler CapEx guidance and on estimate revisions following the Qualcomm-Amazon announcement, because that is what moves the companies feeding the buildout.

For more of his current positioning, see Marc Chaikin's free report on the AI stocks he would buy and the ones he would avoid.