- Author rejects selling SK hynix because Samsung sampled hybrid bonding first: sampling enables validation, while product, customer, timing and process requirements remain unclear.
- Samsung is the author’s highest-upside pick because memory, foundry and packaging under one roof could shorten development cycles and give customers more system-optimization options.
- The author expects strong prelims, with DRAM, NAND, memory prices, AI infrastructure and expanding HBM4 supporting results; AMD’s MI455X collaboration and potential foundry work provide evidence of renewed customer engagement.
Investment pitches
filter:
63 72 - Author expects no pure meme to exceed $1B market cap this cycle: launchpad proliferation fragments liquidity and gives new tokens repeated chances to siphon attention; they cite a 90% chance of a competing “vamp.”
- Above a certain size, pure memes lack an outside bid: degens have bought, funds cannot underwrite fundamentals, and capital is divided across many alternatives.
- The author favors tokens with innovation plus value accrual—usage-linked burns, revenue-funded buybacks, or growing reserves. RWA memes could bridge attention and utility, but the stance is to take profits earlier on pure-meme runners.
75 - AMZN’s lag reflects a real threat: AI agents could control discovery and purchasing, weakening Amazon’s retail and advertising moat.
- The author expects gradual adoption and Amazon adaptation through Project Moonraker or partnerships; AWS, Trainium, Graviton and other AI infrastructure could outweigh advertising dollars at risk.
- Anthropic concerns and higher diesel costs add pressure, but the author points to strong Anthropic growth, an intact AWS commitment, eventual fuel-cost pass-through and fulfillment efficiencies—and remains willing to own AMZN if the bull case is more right.
63 - Ultra-fast inference can expand capability, not just reduce waiting: sequential agent calls compound latency, while saved time enables more testing, verification and reasoning.
- Premium pricing can be rational when human time or iteration speed is scarce: Anthropic charged 6× standard prices for up to 2.5× faster inference.
- The Cerebras bull case rests on scarce frontier-level, low-latency capacity; the author cites reported $200M/MW pricing and a disclosed 750MW, $20B-plus agreement.
- A speculative $1T scenario requires $100B revenue, 25% FCF margins and a 40× multiple; it depends on scaling capacity, retaining performance and avoiding commoditizing competition.
83 - Ibiden’s high-end ABF capacity is tightening: tier-1 and tier-2 suppliers are full, lead times reach 48–52 weeks, and 2027 capacity grows only 10–15% versus a projected shortfall from 27Q1.
- AI silicon consumes disproportionate capacity: author forecasts 2027 Intel/AMD server CPU shipments up 36%/62%, with new substrates 56–63% larger; Rubin is estimated at 1.8× Blackwell’s area, while TPU V9 and Trainium 4 yield only four substrates per panel.
- The thesis depends on higher 27Q1 contract-renewal pricing; specific quotes are not yet visible. The FY27Q2 results call is the near-term catalyst, while AMD substrate revenue temporarily declines as capacity shifts to Intel.
65 - KNOTS fell from a $52m peak to $2.5m as STONK and crypto sold off; its beta and yield narrative amplified weakness when declining volatility reduced the yield.
- Author sees no KNOTS-specific failure: STONK, Solana, BTC, and onchain markets weakened together, while STONK revenue declined only commensurately with onchain activity.
- At $3.8m, author is holding and awaiting a sustained STONK uptrend, with $5.5m as a bullish-structure threshold; the thesis fails if onchain activity pauses or STONK revenue collapses while broader volume recovers.
61 - Author argues SONY gets roughly a 30% cut of GTA 6 sales across its 70%+ console share, yet forecasts price in no GTA 6 earnings growth this year.
- Gaming contributes about 40% of conglomerate earnings, entirely from digital sales, add-ons, microtransactions, and PSN subscriptions—not hardware—making the exposure high margin.
- Launch sales through March 2027 are the near-term uplift; GTA Online in 2027 could extend it through add-on content and PSN subscriber growth, with GTA V still a top PS5 game.
- At 17x forward P/E and roughly 17% projected earnings growth, the author expects a mid-20s platform multiple versus 25x in November 2025, supporting a new ADR high and aggressive low-IV options positioning.
66 - Weak NFP (+29K vs. 90K expected; July negative, August revised 162K to 133K) leaves a hiking Fed vulnerable, while the dollar fell 0.23%.
- Despite BTC fading from $86.6K to $83.8K, desks bought and rolled upside: $914K in November $90K calls, $4.0M in deep-ITM October $70K calls, and $1.86M rolled into March $88K calls.
- Structures target $92K–$96K by October 30, with a $301M Derive call condor and onchain block-flow share rising from 2.5% to 15% supporting the venue-shift theme.
- The October 27–28 Fed meeting is the key date; below $82K, the author says these positions become downside fuel. ETH remains a weaker beta trade than BTC.
77 - Bernstein initiates FPS at Outperform with a $48 target, citing a peer discount and forecasting 70% EPS CAGR through 2030 versus roughly 60% consensus.
- As modular data centers rise from 40% to 60% of construction by 2030, integrated systems could expand FPS’s content opportunity 55% to $2.8M/MW.
- FPS’s 30–50% shorter lead times, available capacity and vertical integration support share gains; non-hyperscaler share is 7% versus 4% overall, alongside a first hyperscaler win.
- Bernstein forecasts EBITDA margin expanding 500 bps to 28% by 2030 through utilization, SG&A leverage and mix; positive revisions are the proposed catalyst.
63 - SDEV’s concentrated ownership—R01 and Framework hold 87.6% of disclosed shares, leaving ~6.4M outside blocks versus ~1.2M short—can amplify the squeeze.
- At the October 1 close, SDEV traded at ~3.7x warrant-adjusted NAV; its $84.5M remaining ATM could provide the author’s estimated ~$30M for SKY purchases.
- Using $7.3–12.6M daily volume, 10% participation would deploy $30M over 24–41 days, buying ~333M SKY at $0.09 before slippage and sustaining a bid.
- SKY’s August net revenue rose ~50% to $15.75M, while higher USDS-backing yields could support earnings; confirmation requires ATM proceeds, purchases, and a sustained daily close above $0.09.