EOD Sync: AEHR Was a Miss-Then-Catch, IBM Was the Only Real Loser Today
EOD analysis — Tuesday, July 14, 2026 · 7:20 PM ET (updated) · brought to you by SignalDeck.live.
Situational Awareness
📈 Indices actually closed HIGHER, not lower — correcting the earlier version of this report. Verified against two independent sources: S&P 500 +0.38% (7,543.59), Nasdaq Composite +0.9% (26,107.01), Dow +0.02% (52,508.27), “boosted by semiconductor stocks.” An earlier version of this report stated the opposite (Dow -0.3%, S&P -0.8%, Nasdaq -1.6%) — those figures were wrong, most likely cross-contaminated with Monday’s actual closing numbers (Nasdaq -1.55% on 7/13, suspiciously close to the erroneous figure used here). The tell we should have caught in the moment: that version’s own text noted an internal contradiction (“SOXX +2.5% yet Nasdaq -1.6%, weakness concentrated elsewhere”) without resolving it — an unexplained internal inconsistency like that should trigger a second source check before publishing, not a hand-wave.
📉 IBM closed -24.6% ($218.86) — worst single session since at least 1972 per FactSet. The single heaviest weight on the S&P 500 today. This part of the earlier report was correct and independently corroborated across sources.
💾 The semis/memory bounce faded relative to its premarket peak but the broader tape still closed green, “boosted by semiconductor stocks” per the close-of-day wire. This morning’s technical-bounce caution (bargain-hunting after KOSPI’s oversold plunge, not a fundamental resolution) is still the right frame for why the premarket-to-close fade happened — it just happened against a green tape, not a red one.
🇮🇷 Iran/Hormuz tension remains the active macro overhang into tomorrow — no de-escalation today.
Today’s Full Arc — Premarket → Midday → Close
TSEM: premarket +19.3% (peaked ~+25% at the open) → midday +12.7% → closed +11.7%. Real, unchanged catalyst ($3B Japan expansion, raised 2028 outlook) — ordinary profit-taking into a large gap, not a thesis break.
⚠ CLSK: premarket +18.8% → midday +10.2% → closed +9.2%. No catalyst ever confirmed — consistent with this morning’s “unconfirmed” flag.
Semis broadly (SOXX proxy): opened +5.7% → midday +3.4% → closed +2.5%.
MU: closed +4.5%. SNDK: closed +4.4%.
AEHR — corrected, and the day’s real headline. Initially scored SKIP in the first version of this report on a relative-volume reading of 1.46x (below the 2.0x gate) taken before the full session’s volume was in. Rescored after close with the accurate day-end relative volume of 2.01x — which clears the gate — and the verdict flips completely: STRONG⭐⭐⭐, EP-γ 94 (confirmed on independent second-pass scoring at 91). Now banked as a recommendation. Full detail below.
COHU (also Semiconductor Equipment & Materials) +8.0% after hours — not a fresh print (next earnings 7/30); riding Stifel’s 7/10 price-target raise ($50→$70) and Baird’s 7/9 Outperform initiation, plus likely same-industry sympathy with AEHR’s beat tonight. Classified
analyst_action/sector_sympathy, not scored via EP-γ.
AEHR — Full Detail
Fiscal Q4 2026 (reported after close 7/14): EPS $0.11 vs. -$0.02 estimated; revenue $18.8M vs. $14.1M prior year (+33% YoY); GAAP net income $1.4M vs. a prior GAAP net loss; record quarterly bookings $60.7M; backlog $80.6M ($100.6M effective including post-quarter orders); FY2027 revenue guidance $130-150M. First fresh-EP candidate in 6 straight sessions.
Second-Degree: AEHR’s Print as an AI-Infrastructure Read-Through
Per operator request — tracing what a small-cap semiconductor test-equipment beat actually implies for the broader AI buildout, not just the stock itself.
What AEHR actually does, and why that matters for the read-through. Burn-in is stress-testing every chip at extreme temperature and full power for hours or days, forcing weak units to fail in the factory instead of in a customer’s data center. AEHR is the only company offering both wafer-level (FOX-XP/FOX-NP, up to 18 wafers simultaneously) and package-level (Sonoma, handling 2,000+ watts per device) burn-in for AI processors, silicon photonics, SiC power devices, GaN, and flash memory. This matters structurally: because burn-in happens before a chip ships, AEHR’s bookings and backlog are a forward-looking, units-based signal about actual AI silicon production ramps — complementary to, and running slightly ahead of, the dollar-based capex signals (hyperscaler capex guides, memory pricing) this week’s reports have been tracking.
The business has completely flipped in two years — this is the real story, not a one-quarter beat. Two years ago, over 95% of AEHR’s business was tied to silicon carbide for electric vehicles. Today, almost 95% of FY2026 revenue comes from markets other than EV silicon carbide — overwhelmingly AI processors, silicon photonics, and power semiconductors. The ~$8M in SiC/EV orders mentioned in recent press releases is now the small residual, not the main business. Don’t read tonight’s beat as “an EV-testing company having a good quarter” — that company doesn’t really exist anymore.
The concrete AI-infra data point: in April 2026, AEHR announced a $41M production order from an undisclosed “lead hyperscale AI customer,” pushing second-half bookings above $92M even before tonight’s Q4 print. A production order (not an evaluation or pilot) from a hyperscaler is a real signal that whichever custom-silicon program that customer runs (the obvious candidates given the “hyperscale” framing: Google TPU, Amazon Trainium/Inferentia, Microsoft Maia, or Meta’s MTIA) is moving into volume manufacturing, since burn-in capacity gets bought ahead of the chips it will test. AEHR’s press releases don’t name the customer, so treat the specific-chip-program attribution as informed inference, not confirmed fact.
Where this sits relative to the house AI-supplier framework (see 2026-06-25 memory-cycle piece): that framework’s Camp A (commodity memory) / Camp B (high-multiple connectivity) taxonomy doesn’t cleanly fit test/burn-in equipment. AEHR’s moat is closer to Camp B’s differentiated-IP shape (it explicitly claims to be the only company with this combined capability) than Camp A’s commodity cyclicality — but it sits one step further back in the supply chain than either: it doesn’t compete for the same GPU/HBM capacity everyone else is fighting over, it profits from whoever wins that fight needing to test their output. That’s a structurally different risk profile — less exposed to a single company’s product cycle, more exposed to aggregate AI-silicon unit volume broadly.
Caveats, stated plainly: this is one earnings print from a $2.26B company; the exact AI-vs-other split of tonight’s $60.7M bookings wasn’t disclosed in what we found (only the trailing two-year mix shift); and a single-quarter bookings record can be lumpy/order-timing-driven. The COHU sympathy pop tonight is a real, live signal the market is already pricing a segment-wide re-rating of test/reliability-equipment names off this print, not just an AEHR-specific move — worth watching whether that broadens further tomorrow.
Recommendation Review Status
AEHR banked (STRONG⭐⭐⭐, D1 due 7/15). No other pending D1/D3/D5/D10/D15 reviews due today. No R10.1 second-entry candidates.
Bottom line: The day’s real story was almost missed by our own process — a real, high-conviction AI-infra beat (AEHR) got scored SKIP on stale intraday volume data before flipping to STRONG⭐⭐⭐ once the close was in, and this report’s own index numbers were wrong until corrected. Two verification lessons in one report: always re-check a volume-gated SKIP after the close, and never publish index-level numbers without a second source, especially when your own data contains an unexplained internal contradiction.
📊 Read this on SignalDeck.live.
Not financial advice. For informational purposes only. Do your own research.