YB new stock pitches (Tue, Aug 25)

Hunterbrook Media uncovered what appears to be Modine's ($MOD, an industrial chiller, air-handling, and cooling equipment maker) internal data center database, accidentally uploaded to public GitHub about a month ago (without password or payment required) during an apparent database migration from Microsoft Power BI to Google Cloud Cortex by a software engineer with a Modine email address; the database—over 100,000 rows across 41 tables covering April 2025 through the early 2030s—reveals customer names, factory planning, prices, margins, supplier constraints, and engineering problems, corroborated by public statements and independent trade data. Key findings: the previously unnamed $4 billion 'strategic data center customer' (announced May 26 with a $165 million upfront payment for 3-megawatt Airedale chillers from 2027-2029) is Google—matching project 'gCUB4.3' (3,512 units at $1.2M each = $4,214,400,000, first-site delivery January 2027, though ~$500M is marked closed/lost leaving $3.7B live, 80% firm); Modine also has prior 2-megawatt Google chiller projects worth $3.55B and total live Google gross demand of ~$8.16B (~$6.07B probability-weighted). The biggest opportunity is Amazon Web Services (also never publicly named) at ~$12.38B gross demand through 2033, mostly air handlers ($11.42B, DAHUs and larger TAHUs) plus ~$1B chillers, with $741M already closed/won and ~$5.6B probability-weighted—surpassing the Google deal—reflecting a hands-on engineering partnership. Beyond these, 20+ other customers (including Crusoe's Abilene/Stargate project for OpenAI/Oracle, CyrusOne, Anthropic, Related Digital) total over $2.3B in pipeline (~$1B firm, $640M likely). Modine applies conversion weights of 100% (firm), 80% (likely), and 40% (early-stage), and total modeled demand is nearly $23B. Probability-adjusted, the September fiscal quarter pipeline implies $540M (~20% above guidance), and FY2027 implies $2.3B (>100% YoY growth, beating the guided 60-80%). Modine is also pursuing certification of a 1-megawatt coolant distribution unit (CDU) against Nvidia's liquid-cooling requirements (no dollar value, no confirmed certification or order), which would move it closer to silicon like rival Vertiv ($VRT). The central risk is supply/capacity, not demand: production is modeled to fall short of demand in most months (especially 2-megawatt chillers), no plant has yet made a 3-megawatt chiller (Dallas and Jefferson City delayed by capital-equipment lead times), and an engineering staff shortage paused at least $149M of projects (CyrusOne, Crusoe) as engineers were reallocated. Critical supplier bottlenecks include flooded evaporators from Wieland ($25.6M impacted), axial fans from ebm-papst ($114M gated; ebm-papst is being acquired by Madison Air Solutions for $5.4B and is expanding capacity), and coils from Cancoil (which admits demand outpaces capacity, needing 18 months-3 years for new factories); Modine is mitigating by qualifying new suppliers such as Canadian Coil Technologies and Delta Electronics, the latter confirmed by trade data showing 75 tons of 'backward blower' shipments to Modine's Grenada plant between July 21 and August 12.

Read the full article here. Read time: 16 min

Natural Gas Services (NGS), a compressor rental company operating critical infrastructure with ~20-year asset lives on long-term contracts, presents a simple investment thesis: they buy compressors and rent them out, with compression being essential—if the compressor fails, everything fails. Demand is increasing due to natural gas-powered data centers, LNG, and the natural progression of oil wells becoming gassier over time (requiring more compression). The investor believes NGS is on track for $180mm-$210mm EBITDA run-rate by end of 2030, arguing the street is under-modeling the mark-to-market of existing contracts when they come due and giving no credit for potential M&A. At that point the stock could command a 7x-9x multiple with debt held approximately constant (currently $329mm); at the low end the share price would roughly double, and at the high end more than triple. The investor recently added back nearly all shares trimmed earlier this year, bringing the position to over 6% of the portfolio.

Read the full article here. Read time: 1 min

Fugazi Research is short Aeva Technologies (AEVA), a 2021 despac-survivor lidar startup trading at a ~$1.32 billion market cap ($18.87/share as of August 20, 2026) against just $18.1 million in FY2025 revenue—roughly 72x trailing sales, 53x annualized H1 2026 run-rate, or 39x its own FY2026 guidance midpoint—versus Oracle at 6x and Coreweave at 7-10x sales; it also trades at ~48x its thin $27.5 million book value, behind which sits an $871.9 million accumulated deficit. The company describes itself as an early-stage business that has only sold prototypes and non-recurring engineering services with immaterial commercial deliveries, posting net losses of $145.4M (2025), $152.3M (2024), and $149.3M (2023)—~$447M cumulative over three years—and losing ~$9.24 per $1 of revenue in H1 2026 (net loss of $114.6M on $12.4M revenue). Aeva survives via dilution and capital markets rather than operations, having raised $115M in a June 5, 2026 offering (5,168,539 shares at $22.25), maintaining a $125M Sylebra standby facility through November 8, 2026, and carrying $100M of 4.375% convertible senior notes due 2032, with shares outstanding rising 10.7% (to 69.7M) in under five months and R&D of $85.4M exceeding revenue 4.5x. The bear case centers on 'valuation by adjacency'—a decade-long treadmill of hype-driven partnership announcements (NVIDIA, Daimler/Torc, Audi, ZF, Denso, SICK, Railergy, NASA, LG Innotek, and now an August 5, 2026 AI data-center optical-connectivity JDA targeting late-2027 deployment and 2028 production ramp with no committed volume) that pump the stock without producing scaled revenue; a misplaced bet on FMCW lidar, an architecture Tesla rejected entirely (camera-based vision) and Mobileye abandoned in 2024; extreme customer concentration (64% from three customers in 2025, 72% in 2024, 59% in H1 2026); a misleading 'Order Book' that is merely a forward-looking non-binding billing estimate; a $14M Delaware settlement over misleading merger proxy materials; and insiders selling ~$11.9M into the August 2026 announcement spike (CEO Salehian and CTO Rezk selling 405,099 shares for ~$9.79M at $23.81-$25.69, departing CFO Sinha selling 70,000 shares for ~$1.48M) with no open-market purchases. Fugazi concludes AEVA is speculative, equity-funded commercialization drift masquerading as an AI story, and considers it uninvestable at any price above zero.

Read the full article here. Read time: 14 min