AEHR — Aehr Test Systems (IT — Strict Pass, OM2s 95.0). The story: a tiny semiconductor test-equipment maker repositioned as an AI/silicon-carbide picks-and-shovels play. Why investors are excited: leverage to SiC and AI-chip burn-in testing, a small float that moves violently on any order news. Why the model hates it: worst-ranked short in the book — Score −179, SMP −382, 26.6x fwd EV/Sales, 451x EV/EBITDA, −8.2% ROIC, and a 2.5% Trifecta Ratio (near-lowest FCF-margin/ROIC/growth composite anywhere). Bull case: (1) SiC/AI test demand inflects; (2) 12.7% short interest + 4.87 beta = squeeze fuel; (3) niche monopoly-ish position in wafer-level test. Bear case: (1) trades at 26x sales on negative returns; (2) price is 1.68x the OM target, 1.02x SS — no margin of safety; (3) 96 volatility/squeeze score cuts both ways and revisions are weak. Verdict: the model's highest-conviction short — a lottery-ticket valuation on an unprofitable micro-cap; size for the squeeze risk.

HUT — Hut 8 Corp. (IT/crypto-miner — Strict Pass, OM2s 91.8). The story: a bitcoin miner pivoting into AI/HPC data-center hosting. Why investors are excited: the "miner-to-AI-hosting" re-rate narrative plus bitcoin beta. Why the model hates it: Score −146, SMP −345, 32.6x EV/Sales, −18.1% ROIC, −0.19 FCF/EV (burning cash), CAS just 10. Bull case: (1) AI-hosting contracts could re-rate the multiple; (2) bitcoin upside; (3) 11.8% short interest + 3.78 beta squeeze potential. Bear case: (1) 32x sales for a capital-intensive, negative-return business; (2) negative free cash flow; (3) price 1.49x OM target. Verdict: strict-pass short on valuation and cash burn; the AI-pivot hope is doing all the work holding the price up.

ALAB — Astera Labs (IT/AI-silicon — Strict Pass, OM2s 91.3). The story: the marquee AI connectivity-silicon name (retimers, smart cabling) for data-center scale-up. Why investors are excited: pure-play AI-infrastructure exposure with real growth and 90 EQS/20.8% ROIC — this one is actually profitable. Why the model hates it: valuation, not quality — Score −175, SMP −355, 39.0x fwd EV/Sales, 100x EV/EBITDA, a 100 Behavioral Dislocation score, and price 1.65x both OM and SS targets. Bull case: (1) best fundamentals of any short (99% Trifecta Ratio, 20.8% ROIC); (2) secular AI-connectivity tailwind; (3) high-quality growth that can grow into the multiple. Bear case: (1) 39x sales is priced for flawless execution; (2) 100 behavioral-dislocation — the most stretched-vs-targets name here; (3) any AI-capex wobble compresses it hardest. Verdict: the highest-quality short and therefore the most dangerous — this is a valuation short on a good company, best paired against a long, not held naked.

AIP — Arteris (IT/semi-IP — Near Pass, OM2s 90.6). The story: a network-on-chip IP licensor levered to custom-silicon and AI SoC design. Why investors are excited: royalty-model optionality as more chips get designed. Why the model hates it: −640% ROIC, 3,466x EV/EBITDA (effectively no earnings), 16.4x sales, CAS 13. Bull case: (1) IP royalties scale with design wins; (2) small-cap re-rate potential; (3) AI-SoC proliferation. Bear case: (1) deeply negative returns on capital; (2) 1.35x OM target; (3) 88 compression + 92 dislocation risk. Verdict: near-pass short on an unprofitable IP story priced well ahead of economics.

MXL — MaxLinear (IT/semis — Near Pass, OM2s 89.9). The story: a communications/broadband chipmaker pitched on infrastructure and connectivity recovery. Why investors are excited: cyclical rebound hopes and content-gain stories. Why the model hates it: Score −137, SMP −239, −21.2% ROIC, 12.1x sales, 99 Behavioral Dislocation, price 1.45x OM / 1.49x SS. Bull case: (1) chip-cycle recovery lifts estimates; (2) 3.06 beta squeeze potential; (3) 62 FRM hints at a growth inflection. Bear case: (1) negative ROIC at 45x EBITDA; (2) among the most stretched vs targets; (3) weak capital allocation (CAS 26). Verdict: near-pass short — a loss-making cyclical priced for a recovery that hasn't shown up in returns.

NBIS — Nebius Group (IT/AI-cloud — Near Pass, OM2s 89.7). The story: the ex-Yandex "neocloud" renting GPU capacity for AI. Why investors are excited: one of the purest AI-compute-capacity plays with explosive revenue growth. Why the model hates it: −0.44 FCF/EV (heavy cash burn on GPU capex), 10.5x sales, 94 Volatility/Squeeze, 19.9% short interest, 2.98 beta. Bull case: (1) AI-compute demand is real and NBIS has capacity; (2) huge short interest = squeeze risk; (3) sovereign/enterprise AI contracts. Bear case: (1) massively negative free cash flow funding capex; (2) 1.31x OM target; (3) commodity-GPU-rental economics under pressure as supply grows. Verdict: near-pass short on cash burn and valuation — but the 19.9% short interest makes it the single most squeeze-prone name; handle with care.

NVTS — Navitas Semiconductor (IT/GaN-SiC — Strict Pass, OM2s 89.2). The story: gallium-nitride/silicon-carbide power chips pitched on AI-datacenter power and EVs. Why investors are excited: an NVIDIA power-partner narrative and electrification tailwind. Why the model hates it: worst Score in the book (−254), SMP −394, 70.3x fwd EV/Sales (highest here), −34.6% ROIC, negative EBITDA, price 2.35x OM target. Bull case: (1) AI-power-delivery design wins; (2) 15.3% short interest + 4.18 beta; (3) GaN adoption curve. Bear case: (1) 70x sales on deeply negative returns is extreme; (2) 2.35x OM target — furthest above target in the book; (3) tiny profits, big losses. Verdict: strict-pass short — the most valuation-detached name; a violent squeeze risk wrapped around a broken income statement.

DDOG — Datadog (IT/software — Near Pass, OM2s 89.0). The story: the premium cloud-observability platform, a quality SaaS compounder. Why investors are excited: durable growth, high retention, AI-observability upside. Why the model hates it: 96 Multiple Compression + 95 Behavioral Dislocation — 77x EV/EBITDA, 19.6x sales, only 2.7% ROIC despite an 88.9% Trifecta Ratio; FRM is a weak 33. Bull case: (1) best-in-class SaaS with 72 EQS; (2) AI-monitoring TAM expansion; (3) low beta (1.18) makes it a "safer" short-hedge. Bear case: (1) 19.6x sales for decelerating growth (FRM 33); (2) 1.21x OM target; (3) among the highest compression-risk scores here. Verdict: near-pass short — a great company at a multiple the model reads as unsustainable; a valuation-normalization short, not a broken-business short.

RIOT — Riot Platforms (IT/crypto-miner — Composite, OM2s 88.8). The story: a large bitcoin miner with an AI/HPC data-center pivot. Why investors are excited: bitcoin beta plus AI-hosting optionality. Why the model hates it: Score −49, −25.4% ROIC, −0.08 FCF/EV, EQS 22, CAS 10, 6.1% Trifecta Ratio (near-bottom). Bull case: (1) bitcoin rally + halving dynamics; (2) 14.7% short interest, 3.34 beta squeeze; (3) power/site assets have AI-hosting value. Bear case: (1) cash-burning, negative-return miner; (2) 13.5x sales; (3) dilution history. Verdict: composite short — lower Score than the strict names but the same burn-plus-hope profile; a bitcoin/AI-narrative fade.

BE — Bloom Energy (Industrials — Near Pass, OM2s 88.5). The story: solid-oxide fuel cells pitched as AI-datacenter on-site power. Why investors are excited: 91 FRM (strong forward growth) and the data-center-power thesis. Why the model hates it: ~0% ROIC, 89x EV/EBITDA, 18.9x sales, 96 Behavioral Dislocation, 4.60 beta, price 1.39x OM target. Bull case: (1) genuine data-center power demand (best FRM in the short book at 91); (2) 10.5% short interest + high beta squeeze; (3) energy-transition policy support. Bear case: (1) barely breakeven at 89x EBITDA; (2) capital-intensive with thin returns (CAS 38); (3) among the most stretched-vs-target names. Verdict: near-pass short — the only non-tech name with a real growth story, which makes it the riskiest short here; the 91 FRM is a genuine warning against pressing it.

WYFI — WhiteFiber (IT/AI-infra — Near Pass, OM2s 88.4). The story: an AI/HPC data-center and connectivity infrastructure name. Why investors are excited: AI-buildout exposure, small float. Why the model hates it: Score −96, −9.1% ROIC, −0.15 FCF/EV, EQS 21, CAS 9, 4.71 beta. Bull case: (1) AI-infra demand; (2) small-cap re-rate potential; (3) high beta squeeze fuel. Bear case: (1) cash-burning with weak returns; (2) 8.6x sales, 1.28x OM target; (3) 93 dislocation risk. Verdict: near-pass short — a small, unprofitable AI-infra name priced on narrative; low quality (33% Trifecta).

WULF — TeraWulf (IT/crypto-miner — Strict Pass, OM2s 87.6). The story: a bitcoin miner leaning into AI/HPC hosting with nuclear-adjacent power. Why investors are excited: the AI-hosting pivot and power-asset value. Why the model hates it: EQS 11 and CAS 8 (near-lowest in the book), −34.8% ROIC, −0.28 FCF/EV (heaviest burn), 25.6x sales, 21.9% short interest. Bull case: (1) AI-hosting contract catalysts; (2) 21.9% short interest — big squeeze risk; (3) low-cost power assets. Bear case: (1) worst earnings-quality/capital-allocation scores here; (2) deeply negative free cash flow; (3) 25x sales. Verdict: strict-pass short on fundamentals, but the 21.9% short interest makes it a squeeze magnet — a fundamentals-vs-positioning tug-of-war.

LSCC — Lattice Semiconductor (IT/FPGA — Near Pass, OM2s 87.4). The story: a low-power FPGA maker with a quality reputation and edge/AI exposure. Why investors are excited: high-margin franchise, design-win momentum, 88.5% Trifecta Ratio. Why the model hates it: valuation — 94 Multiple Compression, 63x EV/EBITDA, 23.4x sales on just 2.6% ROIC; Score −56. Bull case: (1) genuinely high-quality FPGA franchise; (2) edge-AI content growth; (3) low short interest (4%) — less squeeze risk to a short. Bear case: (1) 23x sales for modest current returns; (2) 1.15x OM target; (3) FRM only 40 — growth has cooled. Verdict: near-pass short — a good chip company the model fades purely on multiple; a mean-reversion short.

VPG — Vishay Precision Group (IT/sensors — Near Pass, OM2s 87.3). The story: precision measurement/sensor components. Why investors are excited: niche industrial-tech exposure. Why the model hates it: 97 Behavioral Dislocation (price far above targets — 1.32x SS), 38x EV/EBITDA on 1.6% ROIC, Score −51. Bull case: (1) diversified sensor end-markets; (2) reasonable 4.45x sales (cheapest of the shorts on that metric); (3) recovery optionality. Bear case: (1) price stretched 1.32x vs SS target; (2) low returns at a high EBITDA multiple; (3) weak momentum/quality. Verdict: near-pass short — more a "price ran ahead of a mediocre business" fade than a broken-model short.

HIMS — Hims & Hers Health (Health Care — Near Pass, OM2s 86.8). The story: the DTC telehealth/GLP-1 growth story. Why investors are excited: viral consumer growth and weight-loss/compounded-drug TAM. Why the model hates it: FRM just 19 (growth decelerating in the model), 28.3% short interest (highest in the book), −1.2% ROIC, 97 Behavioral Dislocation, price 1.35x SS. Bull case: (1) huge consumer brand + GLP-1 demand; (2) 28.3% short interest = extreme squeeze fuel; (3) subscription stickiness. Bear case: (1) regulatory risk to compounded-GLP-1 revenue; (2) decelerating forward growth (FRM 19); (3) stretched 1.35x vs target. Verdict: near-pass short with the single highest squeeze risk in the book — thesis is regulatory/valuation, but 28% short interest means never press it into strength.

AXTI — AXT, Inc. (IT/compound-semi materials — Strict Pass, OM2s 86.7). The story: a maker of compound-semiconductor substrates (GaAs/InP) tied to AI/photonics. Why investors are excited: AI-optical and datacenter-photonics substrate demand. Why the model hates it: Score −169, SMP −345, 22.4x sales, 122x EV/EBITDA, −3.9% ROIC, 11.9% short interest. Bull case: (1) photonics/optical-AI substrate tailwind; (2) squeeze potential (11.9% SI, 3.0 beta); (3) China-listing/subsidiary optionality. Bear case: (1) negative returns at 122x EBITDA; (2) 1.62x OM target; (3) weak capital allocation (CAS 14). Verdict: strict-pass short — an unprofitable materials name riding the photonics narrative well past its economics.

CRWD — CrowdStrike (IT/software — Strict Pass, OM2s 86.7). The story: the premier endpoint/cloud-security platform and a market darling. Why investors are excited: security-platform consolidation, strong retention, AI-security upside. Why the model hates it: 31x fwd EV/Sales, 102x EV/EBITDA, −0.6% ROIC, 97 Behavioral Dislocation, price 1.41x OM / 1.11x SS — priced for perfection. Bull case: (1) best-of-breed platform with an 84.2% Trifecta Ratio; (2) low 3% short interest and 1.49 beta (a "cleaner" short); (3) durable growth into the multiple. Bear case: (1) 31x sales is one of the richest valuations in the market; (2) negative current ROIC; (3) 89 compression risk. Verdict: strict-pass short — a top-tier company at a multiple the model won't defend; a valuation short best expressed vs a cheaper software long.

WOLF — Wolfspeed (IT/SiC — Near Pass, OM2s 86.6). The story: a silicon-carbide pure-play for EVs and power, post-severe-drawdown. Why investors are excited: SiC secular demand and a potential turnaround/restructuring bounce. Why the model hates it: 53.8% short interest (by far the highest), −10.7% ROIC, negative EBITDA, CAS 6, 96 Volatility/Squeeze. Bull case: (1) SiC demand recovery; (2) 53.8% short interest = enormous squeeze risk; (3) restructuring/refinancing catalysts. Bear case: (1) balance-sheet and dilution overhang; (2) negative returns and free cash flow; (3) 1.35x OM target. Verdict: near-pass short on fundamentals, but 53.8% short interest makes it the most dangerous squeeze in the entire book — the model's bearish, but this is not one to short naked.

MRAM — Everspin Technologies (IT/memory — Near Pass, OM2s 86.6). The story: a niche MRAM (persistent memory) maker. Why investors are excited: specialty-memory demand and defense/industrial niches. Why the model hates it: micro-cap (~$449M), ~0.4% ROIC, 22x EV/EBITDA, 93 Behavioral Dislocation, Score −66. Bull case: (1) defensible MRAM niche; (2) 9.6% short interest + 3.20 beta; (3) design-win optionality. Bear case: (1) negligible returns at a high multiple; (2) 1.15-1.17x targets; (3) thin liquidity cuts both ways. Verdict: near-pass short — a stretched micro-cap with barely-there profitability.

VRNS — Varonis Systems (IT/security software — Near Pass, OM2s 86.5). The story: a data-security software vendor mid-transition to SaaS. Why investors are excited: data-security TAM and SaaS-transition margin story. Why the model hates it: −12.4% ROIC, 156x EV/EBITDA, 96 Behavioral Dislocation, Score −78, price 1.19-1.21x targets. Bull case: (1) SaaS-transition inflection; (2) very low 0.86 beta (a low-volatility short); (3) data-security demand. Bear case: (1) negative ROIC through the transition; (2) 156x EBITDA; (3) stretched vs targets. Verdict: near-pass short — a transition story priced ahead of the payoff; low beta makes it a calmer short.

BFLY — Butterfly Network (Health Care — Near Pass, OM2s 86.3). The story: handheld point-of-care ultrasound with an AI-imaging angle. Why investors are excited: democratized-imaging vision and AI diagnostics. Why the model hates it: Score −146, −32.0% ROIC, negative EBITDA, 15.3x sales, 99 Behavioral Dislocation, price 1.49x OM / 1.27x SS. Bull case: (1) disruptive imaging TAM; (2) AI-software attach; (3) small-cap re-rate potential. Bear case: (1) deeply negative returns, cash burn; (2) among the most stretched vs targets (99 dislocation); (3) commercialization has repeatedly disappointed. Verdict: near-pass short — a story stock priced far above a loss-making reality.

CORZ — Core Scientific (IT/crypto-AI — Near Pass, OM2s 86.3). The story: a bitcoin miner turned AI/HPC hosting landlord (the CoreWeave-hosting narrative). Why investors are excited: long-term AI-hosting contracts and data-center asset value. Why the model hates it: −130.2% ROIC (near-worst), −0.06 FCF/EV, CAS 8, EQS 36, 23.5% short interest, 95 Volatility/Squeeze. Bull case: (1) contracted AI-hosting revenue visibility; (2) 23.5% short interest squeeze risk; (3) power/site asset value. Bear case: (1) catastrophic current returns on capital; (2) 10.9x sales; (3) execution/financing risk on the buildout. Verdict: near-pass short on fundamentals, tempered by heavy short interest — the AI-hosting story is holding up a very weak income statement.

AMBQ — Ambiq Micro (IT/ultra-low-power semis — Near Pass, OM2s 86.3). The story: ultra-low-power chips for edge-AI and wearables (recent IPO). Why investors are excited: edge-AI/tinyML exposure and IPO momentum. Why the model hates it: worst-ish Score (−214), −24.2% ROIC, negative EBITDA, 13.8x sales, price 1.93x OM target (2nd-furthest above target). Bull case: (1) edge-AI power-efficiency niche; (2) IPO/growth narrative; (3) small float re-rate potential. Bear case: (1) 1.93x OM target — extreme; (2) negative returns and cash flow; (3) CAS 6. Verdict: near-pass short — a freshly-public edge-AI story priced nearly 2x above the model's target.

CEVA — CEVA, Inc. (IT/semi-IP — Composite, OM2s 86.2). The story: a wireless/edge-AI IP-licensing house (DSP, connectivity). Why investors are excited: royalty exposure to edge-AI and IoT proliferation. Why the model hates it: Score −43 (mildest here), −3.7% ROIC, 41x EV/EBITDA, 92 Behavioral Dislocation, CAS 27. Bull case: (1) IP-royalty leverage to unit growth; (2) cheapest-Score short (least extreme); (3) edge-AI licensing tailwind. Bear case: (1) negative returns at 41x EBITDA; (2) 7.3x sales; (3) stretched vs targets. Verdict: composite short — the softest-conviction name in the book; a valuation/quality fade rather than a broken model.

QUIK — QuickLogic (IT/FPGA — Near Pass, OM2s 85.9). The story: a micro-cap embedded-FPGA and IP company with defense/eFPGA exposure. Why investors are excited: eFPGA design wins and defense-program optionality. Why the model hates it: micro-cap (~$292M, smallest here), −41.0% ROIC, 11.3x sales, Score −112, price 1.34x OM / 0.71x SS. Bull case: (1) eFPGA/defense contract catalysts; (2) tiny float can rip on news; (3) IP-licensing optionality. Bear case: (1) deeply negative returns on capital; (2) micro-cap execution/liquidity risk; (3) 85 compression risk. Verdict: near-pass short — a loss-making micro-cap priced on program hopes; the SS target below price (0.71x) is the one crumb of relative support.