Investment Memo
Formal investment memos plus older company and thesis memos. Memos are a format; Software, Industrials, Chips and TSLA are the subject buckets.
Bloom's solid-oxide fuel cells turn gas into continuous onsite power. This memo examines data-center demand, Q1 2026 earnings, Brookfield financing and its accounting effects, and competition from turbines and reciprocating engines. It also asks what BE must earn to justify a $69B market cap.
An AI memory thesis covering HBM, server DRAM, LPDDR, flash, networking, KV-cache offload, and custom HBM. It separates near-term shortages from the longer-term cHBM opportunity, ranks potential public-market beneficiaries, and defines cycle risks and exit signals.
Vertiv supplies data-center power, cooling, and critical infrastructure. Strong fundamentals come at roughly 50x FY2026 guided adjusted EPS midpoint, 9.0x sales, and 56x adjusted FCF. Base-case fair value is around $340/share versus the recent $315.71 quote.
A review of Leopold Aschenbrenner's Situational Awareness 13F sequence and the post-quarter Nebius 13G. It examines the shift from chips toward power, delivered clusters, goodput, and AI-cloud unit economics. CoreWeave, Nebius, and IREN are compared on live clusters, owned power, financing risk, customer quality, and valuation.
A review of Meta after Q1 2026. Family of Apps generates strong advertising cash flow, while Reality Labs remains loss-making. The $125-145B 2026 capex guide is central to valuation. Includes scenarios, AI revenue paths, supplier implications, and public references from Jefferies, Evercore, JPMorgan, and Damodaran.
60–80% of tech equity value sits in terminal value, which depends on durable competitive advantages. The thesis is that AI threatens software durability first. The 2025 trade favored long silicon (AVGO, SKH) and short SaaS (Duolingo −67%, Adobe −25%). The memo examines implications for growth equity, private credit, and $300–500B/yr AI capex.
Samsung reports ₩57T/quarter in memory operating profit, up 8x YoY. AI inference demand for DDR5, NVIDIA SOCAMM/LPDDR5X demand (a 13x jump from GB200 to VR300), and HBM4 qualification with NVIDIA support the DRAM cycle. The memo views a valuation of 2.2x annualized operating profit as a large Korea discount.
XLI offers concentrated exposure to defense rearmament after Hormuz, $2T+ of infrastructure legislation entering peak disbursement, and AI-driven electrification. At 27x forward P/E, valuation is high. The thesis favors scaling in through cash-secured puts as these spending cycles develop.
The thesis is that Tesla's valuation underestimates autonomy and robotics as EV sales decline. Robotaxis are live in Austin, Cybercab production has started, and Optimus Gen 3 is expected in summer 2026. A possible SpaceX $1.75T IPO raises speculation about a $3T+ merger. Sum-of-parts valuation ranges from $285B bear to $1.5T+ bull.
Palantir's Ontology raises switching costs. AIP Bootcamps convert at 75% in 5 days, and Maven is now a Program of Record. At 60-80x revenue, the stock requires exceptional execution. The memo favors selling puts to enter at a lower price.
NVIDIA's $20B Groq deal supports its shift toward inference. The memo examines its 92% AI training share, 18-year CUDA advantage, and a projected $255B inference market by 2030.
Google trades 29% below peers. Its researchers invented transformers, and TPUs save $3B/yr. GCP is growing at 44% CAGR. The memo estimates a 40% annualized yield from cash-secured puts.
ASML has 100% of the EUV lithography market. Advanced chips that require EUV depend on its machines. AI demand supports equipment spending, with High-NA EUV the next technology transition.
Memory supply constrains AI capacity. Three companies control all HBM supply. The memo projects growth from a $4B market to $91B by 2030 at 56% CAGR and cites 118% utilization. Tight supply favors producers, with MU the preferred investment.
Power availability constrains AI deployment. ERCOT has 78 GW requested and 1.1 GW approved, a 70:1 ratio. The memo estimates that onsite power can create a $5.2B timing advantage over 36 months.
Unitree shipped 10,000 robots at $16K each, 10x cheaper than Boston Dynamics. The memo projects a $145B market by 2030 and argues that China's rare earth supply-chain control strengthens its cost advantage.