Earnings Gap-Up Review: April 9 – July 8, 2026 (58 Names, 20%+ Gaps)

58 small-cap-and-up names (mkt cap ≥$100M) gapped up 20%+ on earnings between April 9 and July 8, 2026. The question that matters isn't "who popped" — it's "who kept it." This is a follow-through study, anchored on the BAND trade you missed.

The method

For every name, I compared the initial gap% (open vs. prior close on the reaction day) against the total return from the pre-earnings base price to the live price as of today (9/28/26) — call it the follow-through ratio. A ratio above 1.0x means the stock kept compounding after the pop; below 1.0x means it gave part of the gap back; negative means the whole move round-tripped and went red.

This is exactly the question behind your BAND regret: the stock gapped +27.9% on the 4/30 beat, and the instinct after a violent pop is to assume the easy money is gone. Instead BAND is a 8.02x follow-through ratio — the single best name on the list. It kept climbing for ten weeks, shrugged off a manufactured air-pocket from the $275M convertible notes pricing on 6/16 (converts priced off a $52.83 reference vs. today's ~$78, so that "76 → 48" pullback you remember was the market digesting dilution/hedging flow around the deal, not a change in the fundamental story), and is at new highs now. The lesson: on this list, a post-gap retracement was overwhelmingly a re-entry, not an exit signal — see the tiering below.

Full ranked list — follow-through ratio (best to worst)

Group 1 — Beaten-down SaaS/software snapback (your BAND cluster)

BAND (8.02x), TWLO (2.37x), COMP (2.32x), EVER (2.26x), AGYS (1.75x), TEAM (1.54x), PD (1.60x), RXT (1.23x), GBTG (1.02x)

This is the standout group and the one most relevant to you personally. Nine names, all beaten-down or out-of-favor software/internet stocks going into earnings, all beat and raised, and every single one held or grew the gap over the following weeks — zero round-trips. Compare that to the chip/hardware group below, where roughly half gave the gap back. The pattern: when a low-multiple, low-sentiment software name beats and guides up, the re-rating tends to be a regime change (multiple expansion off a depressed base), not a one-day event — and dips inside that re-rating (BAND's notes-deal air pocket, COMP/EVER's typical day-2 giveback) are buyable, not sells.

Recommendation: Build a standing scan for "software/internet name gaps 20%+ on earnings, market cap <$10B, previously de-rated" — this was close to a 100% hit rate for continuation in this window. RXT and GBTG are the two in this group still worth a look now (RXT only 1.23x — hasn't fully re-rated yet; GBTG flat at 1.02x, worth checking if travel-spend guidance is still improving).

Group 2 — AI datacenter chip/hardware chain: same theme, coin-flip outcome

Winners: MXL (3.12x), SIMO (3.82x), INTC (2.96x), NXPI (1.18x), HIMX (1.09x), DDOG (2.83x)
Faders/failures: VECO (0.62x), TTMI (0.30x), HPE (0.13x), SITM (0.12x), NTAP (0.86x)

This is the single biggest cluster on the list by ticker count (~18 of 58), confirming AI-capex is still the dominant catalyst driving violent earnings reactions across the whole stack — chips, optical, systems, storage. But unlike Group 1, the outcome inside the theme was a coin flip. Confirmed detail: MXL's move was real — infrastructure/optical-datacenter revenue +136% YoY, guide raised to $160–170M for Q2, Stifel/Needham upgrades followed (the stock is +187% over 30 days per Tickeron). SIMO and INTC show the same shape. But SITM and HPE gave back 85–90% of the pop despite comparable-sounding beats (SiTime Q1 revenue +88% YoY) — the difference wasn't "is AI capex real," it was whether the print re-rated forward estimates or just cleared an already-priced-in bar.

Recommendation: Don't treat "AI datacenter beat" as a single trade — MXL/SIMO/INTC/DDOG were genuine estimate-revision events; SITM/HPE/TTMI/VECO were sell-the-news despite decent headlines. The tell in real time is guide magnitude relative to the run-up into the print, not the headline beat.

Group 3 — Data-center/fiber infrastructure builders: priced-in already

STRL (0.90x), DY (0.06x), AAON (0.27x), BBCP (0.78x), AVAV (0.25x, too early)

The literal builders of the AI-capex boom (engineering & construction, fiber buildout, HVAC for data centers) beat big — STRL specifically posted record Q1 revenue +92% YoY and raised full-year guidance to $18.40–19.05 adjusted EPS — and then still gave almost the entire gap back. Confirmed cause for STRL: a sector-wide E&C selloff on 7/2/26 (STRL -10.8% that day alone, with all six comparable peers red) — a rotation event, not a fundamental crack. The read here: this group had already re-rated hard over the prior two years on the AI-buildout thesis, so even blowout prints had no incremental multiple left to expand into, and they're first to get sold when sentiment on the broader capex theme wobbles.

Recommendation: Treat this group as the "tell" for the whole AI-capex trade's temperature — when STRL/DY/AAON fade a beat, that's a sentiment gauge on the capex cycle overall, worth watching before adding to Group 2 names.

Group 4 — Healthcare re-rating

AGL (5.60x), PGNY (2.82x), CERS (1.69x), GKOS (1.56x), VISN (0.93x)

AGL is the #2 name on the entire list. Confirmed: Q1 EPS $1.80 vs. $0.95 est (+89%), net income $49M vs. $12M a year ago, FY26 revenue guidance raised to $5.68–5.81B — even with membership down 13% YoY and topline down 7%, the medical-cost-ratio improvement was the story, and the market re-rated the stock over 5x from a badly depressed base. This is the same mechanic as Group 1 — a low-multiple name proving the bear case wrong.

Group 5 — Deep-value/distressed low-dollar names: mixed, split on real vs. meme

Real turnaround, held gains: XRX (3.19x), PACK (2.84x), OSG (1.64x)
Squeeze that round-tripped: BBBY (0.09x), JMIA (-0.33x), BOOM (0.17x)

Sub-$10 stocks with small-dollar EPS beats produce outsized percentage gaps almost mechanically — but follow-through separates a real balance-sheet/demand story (XRX, PACK) from a pure short-covering pop with no underlying change (BBBY, JMIA gave essentially the whole thing back within days).

Group 6 — Energy storage/climate hardware: avoid regardless of the print

FLNC (0.57x), BE (0.61x), PCT (-0.09x), EOSE (-1.90x)

Worst-performing group as a whole, and EOSE is the single worst name on the list. Confirmed: the +22.8% earnings pop was fully erased by a $75M registered-direct raise (Hudson Bay Capital, 13.7M shares + warrants at $5.48) stacked with a $150M rights offering to fund a JV stake — stock went from ~$8 mid-June to the mid-$4s by 7/7, a ~40% drawdown in weeks. The lesson: in this sector, balance-sheet/dilution risk dominates the fundamental print. A good quarter doesn't matter if a capital raise is coming behind it — check cash runway and ATM/shelf status before trusting any gap in this group.

Bottom line / how I'd use this going forward

1. Your BAND instinct was right, the exit wasn't. The retracement you saw on 6/16 was a financing-driven air pocket (converts priced off $52.83 vs. today's ~$78), not new information about the business. Across this whole 3-month sample, beaten-down software names that gapped up on earnings had a 100% continuation rate — that pattern deserves a permanent scanner entry, not a one-off memory.
2. "AI capex beat" is not one trade. Split it: components with a genuine forward estimate re-rate (MXL/SIMO/INTC/DDOG-style) vs. names clearing an already-priced bar (SITM/HPE/TTMI/VECO) — the guide-vs-run-up math, not the headline, is the discriminator.
3. Use STRL/DY/AAON as a capex-sentiment thermometer — when the builders fade a beat, it's telling you something about the group's crowdedness, independent of any one company's numbers.
4. In energy-storage/climate names, check the balance sheet before trusting the gap — EOSE is the cautionary tale for the whole book.