X-FAB and Sivers: two broken parabolas, and why we didn't buy the dip

The standard read on a stock that's 50% off its high is that it's on sale. Both $XFAB (X-FAB Silicon Foundries, Euronext Paris) and $SIVE (Sivers Semiconductors, OMX Stockholm) closed July 3 right around half their June peaks, and both bounced hard the same day — X-FAB +8.9%, Sivers +14.9%. That's the exact moment the "buy the dip" reflex fires.

We ran both through our entry check instead. Neither passed. Not close — each name failed four of the five gates we require before capital moves. This piece walks through both charts, the check itself, and why a halved parabola is not a discount. It's an unwind that hasn't finished, in names where the fundamentals never supported the peak — or the current price.

Why it matters now

Both names put in their first big green candle after weeks of decline on July 3, 2026. A +9% and a +15% day on the same screen is what a bottom looks like — and also what every dead-cat bounce looks like. The difference is decided by volume, structure, and the business underneath, not by the size of the candle. All three said the same thing here.

Two parabolas, one shape

X-FAB is a European specialty foundry — analog/mixed-signal ICs, MEMS, SiC/GaN power, silicon photonics for automotive and industrial customers. On May 27, a single post from the influential retail account "Serenity" (@aleabitoreddit) citing X-FAB's photonics exposure and CHIPS Act grants sent the stock up as much as 77% intraday, through multiple trading halts (Bloomberg, Investing.com). The run peaked at €15.88 in June. By July 3 the stock closed at €7.86 — down 50.5% from that high, back below its own 50-day average.

Sivers is a Swedish fabless designer of 5G mmWave RFICs and III-V photonics. Its catalyst was more substantive on paper: a collaboration with GlobalFoundries to integrate Sivers laser arrays into GF's silicon photonics platform for AI datacenter optics (announcement coverage). The market took a company doing 61.9M SEK of quarterly revenue (down 22% year over year) and repriced it up more than 3,100% from its March low to a 110 SEK peak — roughly 16.6B SEK of market cap, 57x trailing sales. It closed July 3 at 56.30 SEK, down 48.8% from the high. Behind the tape: reported insider-trading investigations, board departures, and a pulled vote on a planned Nasdaq listing (ad-hoc-news).

Different catalysts, same chart: a flat base, a vertical repricing driven by narrative flow rather than earnings, and then the unwind. The historical pattern library is full of this shape, and it rarely resolves with a V.

The five-gate check we ran

Every discretionary entry at QuantAbundancia goes through the same screen before capital moves. Five gates, all required:

1. Trend alignment — at least two timeframes (weekly / daily / hourly) pointing the same way.
2. Setup quality — a recognizable, statistically defensible pattern: pullback-to-support in an uptrend, confirmed breakout, failed breakdown. Vertical/euphoric charts are an automatic fail, in either direction of the parabola.
3. Risk/reward — a structural stop and a structural target with at least 2.0x reward per unit of risk.
4. Fundamentals — would a long-only quant book own this on the numbers alone? Shrinking revenue plus weak margins plus a stretched multiple is a fail regardless of the chart.
5. Calendar — no binary event (earnings, macro print) inside the trade's opening window.

One failed gate downgrades an entry to WAIT. Two or more is a PASS — no position, no averaging in, no "starter size."

This check runs as an automated workflow — drop a chart, get the full multi-timeframe dossier and the gate-by-gate verdict. It's free for everyone at 5 reads a day; PRO raises the limit to 25. It's what produced both of these calls before a word of this article was written.

Gate by gate: where each name failed

Trend. X-FAB's daily is below both its 20-day EMA (€8.64) and 50-day SMA (€8.55) with MACD negative, while the weekly's momentum has rolled from RSI 76 to 56 in five weeks. Sivers' daily MACD histogram sits at -5.3 with the minus-DI dominant, and its weekly RSI has collapsed from 94.8 to 58.9. In both cases the timeframes disagree — and the disagreement is the downtrend winning.

Setup. Both charts are the textbook vertical/euphoric shape in its unwind phase. There is no pullback-to-support here because the move up never built support — it gapped and ran. When a chart goes vertical on flow, the round trip down has no structure to catch it.

Risk/reward. This is where the arithmetic gets stark. X-FAB's daily ATR is €0.82 — 10.4% of price. Sivers' is 13.36 SEK — 23.7% of price. A stop tight enough to be survivable sits inside one day's noise; a stop wide enough to be meaningful is 35% away on Sivers. Against the nearest real resistance, X-FAB's reward-to-risk pencils to roughly 1.1x and Sivers' to about the same on a stop that isn't real. Our floor is 2.0x. For the framework behind that floor, see R-multiples and expectancy.

Fundamentals. X-FAB: revenue down 4.2%, operating margin 1.2%, trailing PE above 60, and — worth stating plainly — a street average target near €5.50 against a €7.86 price, with one Buy rating out of eight before the frenzy. Sivers: revenue down 21.5%, negative gross margin, operating margin around -67%, at 57x sales. The GlobalFoundries deal may prove real over years; the current income statement does not support the current market cap, let alone June's.

The dead-cat arithmetic

The July 3 bounce is the trap in miniature. X-FAB's +8.9% day printed on 0.71x its 20-day average volume — buyers did not show up in size; sellers paused. Sivers' weekly volume is running at 0.58x its own 20-week average as the price halves. Falling price on falling volume after a parabola is distribution completing, not accumulation starting.

And the daily ranges are the position-sizing killer. When a stock moves 15–20% per day, any position sized to survive the noise is too small to matter, and any position sized to matter can hit a week's risk budget overnight. That asymmetry alone — before trend, before fundamentals — takes both names off the table for a systematic book. This is the mean-reversion trader's paradox: the more stretched the rubber band, the less tradeable the snap.

What would change the read

Observations, not predictions — the triggers that would re-open the file:

• X-FAB: a daily close back above the €8.55–8.64 moving-average cluster on at least 1.5x average volume, or a tag of the 200-day near €6.15 that holds and bases for multiple sessions. Earnings land 2026-07-30 — any trigger firing into that window inherits binary risk.
• Sivers: a multi-week base anywhere, on rising volume, plus a quarter that shows the GlobalFoundries collaboration converting into revenue that moves the -22% growth line. Earnings 2026-08-06.
• Either name: daily ATR compressing back under ~5% of price — the mechanical precondition for a stop that can be both survivable and meaningful.
• The theme: the AI-photonics driver is real; it's the vehicles that failed the check. The tracked names in the photonics theme and the semiconductors bubble carry the same exposure with financials that pass gate 4.

For readers trading European listings directly — both names sit on Euronext Paris and OMX Stockholm, which most US retail brokers don't reach; /stack/ibkr covers the access route we use.

Rule-based entry checks like this one — five gates, multi-timeframe dossier, on any ticker on demand — are free to run at 5 reads a day; /pro#chart-entry raises the limit to 25.

Bubble context: /bubbles/semiconductors and /bubbles/photonics — the tracked clusters that carry this theme. Neither $XFAB nor $SIVE is in the QA tracked universe; that's part of the point.

QuantAbundancia is educational research. Nothing here is investment advice. See /disclosures.