Tech Crowded Trade Unwinds: Software Stocks Bottom Out vs. Semiconductors
Miles Bennett
The year-long long-semis / short-software crowded trade is unwinding — the software ETF-to-semiconductor ETF ratio bounced from a 0.13 low and broke above its downtrend line. Money is rotating out of tech's most lopsided bet, signaling a potential trend reversal in software's relative performance.
How crowded was this trade?
Capital piled into Nvidia, TSMC, and Micron while draining from Palantir, ServiceNow, and Salesforce — one of the most concentrated one-way positions in tech in recent years.
This means → the market treated "AI winner" as synonymous with "buy semiconductor hardware," and dumped software as AI's collective loser.
In plain terms = all the money crowded onto one side of the boat — once the signal flips, the unwind alone can drive violent relative-price swings.
What does the reversal signal look like?
The software ETF (IGV) to semiconductor ETF (SMH) ratio bottomed near 0.13 last month, then broke above downtrend-line resistance.
Software's ratio to the S&P 500 has formed a triple bottom — a classic reversal pattern where price tests the same low three times — building over roughly five months since early 2026.
This reflects selling pressure drying up: sellers are exhausted, the price won't go lower, and technicals are flashing first.
What's Jefferies' take — "software isn't dead, weak software is"?
Jefferies published a 65-page report titled *What's Wrong with Software*, cataloging 10 "walls of worry" around AI's threat to the sector — and argued most are already priced in.
Core logic: unless a single AI model ends up ruling everything, enterprise data is inherently siloed and messy. AI penetrates "clean" domains like coding fast, but spreads far more slowly across other business functions.
This means → incumbents that own chaotic enterprise data have a natural moat. AI agents — autonomous programs that execute tasks — actually need to plug *into* these systems, reinforcing rather than replacing existing vendors.
Where is Jefferies placing its bets?
The strategy overweights the "compute channel" — infrastructure and hyperscale cloud providers that scale with AI usage.
Within application software, the firm is selective: top picks are ServiceNow (application software) and Snowflake (infrastructure software).
In plain terms = not all software deserves a bid — Jefferies is picking the names AI can't bypass and must plug into.
ServiceNow — what do the technicals and fundamentals say?
ServiceNow (NOW) has fallen roughly 50% from its all-time high and is building an inverse head-and-shoulders pattern — a classic bottom-reversal formation — near $100, targeting the 200-day moving average at $132.
Of the 48 analysts covering the stock, 44 rate it buy or strong buy. The consensus target is $141.12; Tuesday's close was $110.73 — about 27% upside to consensus.
Jefferies says the margin dip worrying investors stems from the Armis acquisition — a short-term cost management expects to recover by 2027. EBITDA has grown at roughly 30% annually over the past three years.
Can the reversal stick — what's the key checkpoint?
Whether software's relative strength holds ultimately depends on Q2 earnings delivering the fundamental support to justify the valuation re-rating.
This means → technicals and fund flows have already flashed green, but fundamentals haven't turned in their paper yet — earnings season is the exam.
Guggenheim upgraded ServiceNow to buy on July 1 with a report titled "Apocalypse Cancelled" — institutions are voting for the reversal with real ratings and real money.
Content is for reference only, not financial advice.