- CCXI offers Agility exposure: about 100 Digits operate at nine customer sites, logging 65,000+ hours; its Robots-as-a-Service model sells robotic labor with software and maintenance, not merely hardware.
- Demand includes a claimed $300m+ multi-year Digit v5 order for roughly 1,000 three-year rentals, but it is one commitment, includes deployment-linked warrants, and is not recognized revenue.
- Agility’s illustrative $100,000 annual RaaS revenue per robot implies $1bn at 10,000 deployed versus roughly 100 today; meanwhile cash burn exceeds $100m annually and revenue remains undisclosed.
- Author says CCXI is worth owning near ~$10 trust because redemption rights roughly protect cash, but post-close ~$3.4bn valuation leaves scale and competition unresolved; size it as venture capital and watch redemptions plus v5’s early-2027 launch.
Investment pitches
filter:
82 78 - The thesis targets repeated, scarce joint hardware: Goldman scores harmonic gears 16 and planetary roller screws 14, making reducers the highest-barrier, highest-content components.
- Brains, batteries, cameras and lidar are weaker humanoid exposures because demand is shared with other markets; dexterous hands and actuator assembly score 15 and 14.
- Author’s one robot-maker exception is Agility via CCXI: Digit has logged 65,000+ hours and carries more than $300m of milestone-contingent orders; the merger is expected in 2026, after which the ticker becomes AGLT.
- Author calls it a speculative sleeve, not core; roughly $10/share in trust is the stated pre-close floor, while dilution, redemptions, customer concentration and reducing BOM from about $125,000 to $30,000 remain risks.
87 - After rising >1,000% to ~$1,048 and crossing $1T market cap, MU trades at ~9.1x forward earnings; the author sees peak-cycle pricing despite Q3 revenue of $41.5B versus $33.5B guided and EPS of $25.11 versus ~$19.
- AI makes HBM a bottleneck: Micron’s 2026 supply is sold out, while 16 five-year take-or-pay agreements cover ~20% of DRAM and one-third of NAND volume, potentially placing half or more of revenue under contract.
- Q3 price increases—not volume—lifted gross margin to 84.9%; Q4 is guided to ~$50B revenue and ~86% margin. The author recommends buying pullbacks, but rising capex, rival overbuilding, AI/customer concentration, and Taiwan-China exposure could expose the cycle.
79 - AAOI is positioned in the scale-out interconnect bottleneck: management guides for 2026 revenue above $1.1B versus $455.7M in 2025, with demand exceeding supply through at least mid-2027.
- In-house EML lasers and MOCVD capacity support 800G/1.6T ramps; management targets 40%+ gross margin versus 29.1% today, potentially converting volume growth into earnings leverage.
- At roughly $170, AAOI trades near 22x trailing sales and remains GAAP-loss-making; the author says BUY, with scenarios of ~$100 bear, ~$200 neutral, and $350 bull.
- The thesis depends on execution: Digicomm and Microsoft provide 81.9% of revenue, shares rose 49% in a year, and CPO/SiPh could challenge the moat; watch 800G volume and margins next earnings.
75 - SNDK is the author’s BUY-with-discipline bet that AI inference makes NAND storage a structural bottleneck: Datacenter revenue rose from $197M to $1.467B year over year, up 645%, before its 128TB SSD launch.
- NBM contracts with five hyperscalers reportedly provide $42B of contracted minimum revenue, shifting NAND toward more predictable economics; TTM gross margin reached 78.4%.
- HBF is unmodeled optionality: SNDK and SK Hynix published an OCP specification, but samples target H2 2026, inference devices early 2027, and commercialization may be closer to 2030.
- At 32.3x forward P/E versus roughly 7–19x for peers, the thesis requires margins and 2027 earnings to hold; cyclicality, customer and Kioxia concentration, and disruption leave substantial downside.
84 - CRDO reported $479M revenue, up 114.7%, $1.20 EPS versus $1.17, and Q2 guidance of $525–535M versus $514.7M expectations, yet shares fell 20%.
- The repricing reflects deceleration: Q2 implies roughly 96% growth, FY27 guidance exceeds 85%, and the second half slows further. At $165, author calculates roughly 12x forward sales at 68% gross margin.
- Credo’s retimers and active electrical cables bridge copper and optics; faster co-packaged-optics adoption in the 2027 architecture cycle could design out those products, with customer concentration another risk.
- Author says CRDO is a trade rather than a long-term buy, wants below $160, and considers $140 genuinely cheap; the thesis remains intact and the stock is more attractive after the drop.
85 - NBIS may pass through Nvidia’s >15% server-cost increase: mid/short-term contracts, uncommitted 2027 capacity, and new pricing above $20m/MW preserve repricing power.
- About 70% of Q2 bookings were prepaid, covering half of capex; short-term deals reportedly exceed $40m.
- Author’s $353 base case implies roughly 68% upside at ~$210, but ARR more than triples while multiple compression absorbs much of the growth.
- The $5.75B converts and 15.8m Class A issuance put about 34m shares in play versus ~280m; dilution is certain, but the author remains bullish on improved entry and business quality.