RESEARCH / START HERE
Follow a question, not a pile of posts.
Each article starts with a short brief. Read the full analysis when you want the evidence, and use the paths below to follow the next question.
Compute · Power & payback · Capital & policy · Other themes · Latest
01 / COMPUTE
Beyond the GPU
Start with the hardware thesis. Follow it into memory and programmable chips.
- Where the AI hardware bottleneck moves →
A framework for the next AI hardware bottleneck: connectivity, custom silicon and power, with Marvell as a case study and valuation as a key risk. - AI memory beyond HBM →
AI agents need more than HBM. Explore the case for CPU memory, CXL and context storage—and the revenue signals that would validate it. - The overlooked FPGA bottleneck →
Why programmable chips matter to AI networking, control and testing. A framework for Lattice, AMD and Microchip, with clear risks to the shortage thesis.
02 / PHYSICAL INFRASTRUCTURE
From power to payback
Follow the chain from electricity and siting to operating returns.
- AI power: the value of arriving sooner →
Why time-to-power matters for AI data centers. Explore Bloom Energy, Doosan Fuel Cell and the execution tests behind an onsite power thesis. - Can data centers move offshore? →
Floating data centers could ease siting constraints. The key tests for Korean shipbuilders are paying customers, viable economics and repeat orders. - Oracle: contracts must become cash flow →
Oracle’s AI backlog is only the starting point. Examine capacity delivery, utilization, financing and the path from contracted demand to cash flow.
03 / CAPITAL & POLICY
Follow the money
A three-part framework: productive capacity, its financing and the investment payoff.
- 1. What does the capital actually build? →
Part I of America’s Y Strategy: how incentives and private capital might become productive capacity, and why announcements are not enough. - 2. Who finances the waiting period? →
Part II of America’s Y Strategy: stablecoins, Treasury funding and the refinancing risks that remain while long-term projects are built. - 3. Who captures the investment return? →
Part III of America’s Y Strategy: distinguish industrial winners from attractive investments using cash flow, valuation and explicit failure signals.
Counterarguments & other themes
How to read the research
Distinguish reported facts, the author’s interpretation and forward-looking scenarios. Company guidance and vendor benchmarks are not independent forecasts. Check each article’s as-of date before using market data.
A theme becomes an investment case only when it connects to revenue, cash flow and valuation. Read the failure case as carefully as the thesis. Where original calculations or direct sources are missing, the conclusion is less reproducible.
Latest research
Agentic AI Changes the Memory Trade: HBM Is Only the Hot Tier
Read the brief →: Agentic AI Changes the Memory Trade: HBM Is Only the Hot TierAI agents need more than HBM. Explore the case for CPU memory, CXL and context storage—and the revenue signals that would validate it.
The FPGA Shortage: AI’s Next Hidden Infrastructure Bottleneck
Read the brief →: The FPGA Shortage: AI’s Next Hidden Infrastructure BottleneckWhy programmable chips matter to AI networking, control and testing. A framework for Lattice, AMD and Microchip, with clear risks to the shortage thesis.
AI Has a Siting Problem: Why Floating Data Centers Could Become Korea’s Next AI Infrastructure Opportunity
Read the brief →: AI Has a Siting Problem: Why Floating Data Centers Could Become Korea’s Next AI Infrastructure OpportunityFloating data centers could ease siting constraints. The key tests for Korean shipbuilders are paying customers, viable economics and repeat orders.
AI Leaders Are Saying “Slow Down.” Does That Break the AI Infrastructure Trade?
Read the brief →: AI Leaders Are Saying “Slow Down.” Does That Break the AI Infrastructure Trade?Would slower AI development end the infrastructure trade? Separate training risk from inference demand, and watch spending, orders and utilization.
Oracle (ORCL): From Funding Burden to Monetization Proof
Read the brief →: Oracle (ORCL): From Funding Burden to Monetization ProofOracle’s AI backlog is only the starting point. Examine capacity delivery, utilization, financing and the path from contracted demand to cash flow.
Part I: America’s Y Strategy — How the U.S. Builds Productive Capacity Without State Capitalism
Read the brief →: Part I: America’s Y Strategy — How the U.S. Builds Productive Capacity Without State CapitalismPart I of America’s Y Strategy: how incentives and private capital might become productive capacity, and why announcements are not enough.
Part II: Funding the Y Strategy — Stablecoins, Treasury Engineering and the Bridge Regime
Read the brief →: Part II: Funding the Y Strategy — Stablecoins, Treasury Engineering and the Bridge RegimePart II of America’s Y Strategy: stablecoins, Treasury funding and the refinancing risks that remain while long-term projects are built.
Part III: Investing in America’s Y Strategy — Winners, Risks and Signals to Watch
Read the brief →: Part III: Investing in America’s Y Strategy — Winners, Risks and Signals to WatchPart III of America’s Y Strategy: distinguish industrial winners from attractive investments using cash flow, valuation and explicit failure signals.
Astra and the Next AI Hardware Rally: Why Marvell Fits the New Bottleneck
Read the brief →: Astra and the Next AI Hardware Rally: Why Marvell Fits the New BottleneckA framework for the next AI hardware bottleneck: connectivity, custom silicon and power, with Marvell as a case study and valuation as a key risk.
AI Has a Power Problem: Why Bloom Energy’s Success Could Re-Rate Doosan Fuel Cell
Read the brief →: AI Has a Power Problem: Why Bloom Energy’s Success Could Re-Rate Doosan Fuel CellWhy time-to-power matters for AI data centers. Explore Bloom Energy, Doosan Fuel Cell and the execution tests behind an onsite power thesis.
Why AI Could Make Japanese Gaming IP More Valuable
Read the brief →: Why AI Could Make Japanese Gaming IP More ValuableAI may make games easier to produce while making attention harder to win. Explore Japanese gaming IP, productivity, catalysts and valuation risks.
Long Agribusiness, Short Packaged Foods: A 6–12 Month Agricultural Inflation Trade
Read the brief →: Long Agribusiness, Short Packaged Foods: A 6–12 Month Agricultural Inflation TradeHow food inflation can divide agribusiness and packaged foods. A relative-value framework with supply, pricing-power and short-position risks.