Long term · 5–10 year horizon
For Roth IRA money you won't touch until retirement. Built around tax-free compounding.
Markets are messy short-term, predictable long-term. Structural shifts (demographics, AI, energy transition) play out over decades.
Hit rate 70%+ on quality businesses, even higher on indices. The long arc bends toward growth. The hard part is not selling during 30% drawdowns.
In a Roth, every dollar of gain is yours — no tax on growth, no tax on withdrawal. Compounding tax-free for 30+ years is the single biggest math edge retail investors get.
Roth portfolio buckets
Foundation (60–80% of Roth)
Boring on purpose. Broad index funds you'll thank yourself for at 35.
Total US market + international + bonds. Set, forget, rebalance yearly. Roth = tax-free compounding forever.
Thematic — Power & AI infrastructure
Energy + data center buildout is a 20-year story.
Hyperscalers need power. Utilities, nuclear operators, electrical infrastructure all benefit.
Thematic — Semis picks-and-shovels
Whoever wins AI, they all need chips.
ASML has a literal monopoly on EUV. TSM is the only one fabbing leading-edge at scale.
Thematic — Nuclear renaissance
First new build cycle in 40 years. Long lead times = long-term thesis.
Uranium miner + reactor builder + operator + enricher. Four different exposures to one thesis.
Speculative bucket (≤5%)
Money you can afford to lose. Quantum, junior miners, frontier biotech.
Treat like venture capital. Expect 80% drawdowns. The winners pay for the losers.
Long-term theses
US semiconductor reshoring is irreversible
CHIPS Act is just the down payment. Geopolitics + AI demand = decade of fab buildout in US.
Nuclear renaissance: first new build cycle in 40 years
AI data centers need 24/7 baseload power. Nuclear is the only carbon-free option that delivers it.
Critical minerals decoupling from China
US wants supply chain independence. Rare earths + lithium + uranium are the chokepoints.
Quantum compute: the speculative moonshot
Real revenue is 5+ years out. Stocks trade on roadmap milestones, not earnings. Treat as venture.
Industrial robotics + humanoid moonshot
Reshoring needs automation. Aging population needs labor substitutes. Humanoid is 5+ years out but real.