2 Ridiculously Cheap AI Stocks That Could Turn $1,000 Into at Least $5,000 by 2028

Micron reported blowout fiscal Q2 2026 results and strong guidance but its stock is down ~30% since the print; it trades at ~8.2x sales and just under 20x trailing EPS (forward ~7.6x) and the article cites a $98.91 forward EPS scenario implying a potential $2,275 share price (~6x current). CoreWeave is down ~39% over six months, trading at ~7x sales, with capex slated to jump from $14.9B (2025) to $30–35B, revenue backlog up 342% YoY to $66.8B and revenue up 168% to $5.1B; the piece models $34.5B revenue in 2028 implying a ~$241.5B market cap (~5.6x today). Both names are presented as undervalued AI-exposure plays given tight HBM supply and robust data-center demand through 2030.

Analysis

Micron and CoreWeave sit on different parts of the same bottleneck: HBM/GPU compute intensity. Micron’s profit leverage is not just product mix but gross-margin convexity — each dollar of HBM price strength drops through at a much higher incremental margin because process-node and wafer-utilization fixed costs are already sunk; that implies upside accelerates as supply remains constrained or as Nvidia/others pull-in orders. CoreWeave’s value is optionality in converting contracted backlog into revenue, but the value is binary on multi-month to multi-quarter build milestones (rack-to-revenue conversion) and on access to GPU supply and grid/power capacity, which can create step-function revenue realizations rather than smooth growth.

Key second-order winners: substrate/test/OSAT vendors, specialty chemical suppliers, and regional utilities where new megawatt clusters land (they get long-term demand and PPA pricing power). Losers include legacy hyperscalers that choose to internalize AI capacity instead of renting (that would slow third‑party colo growth), and chipmakers with large capital intensity but lower HBM exposure (higher capex sensitivity if memory pricing normalizes).

Time horizons matter: in days to months, headline GPU availability, trade-policy headlines, and quarterly capex cadence drive volatility; in 6–36 months, supply additions (Samsung, SK Hynix capex) and model-efficiency improvements (reducing HBM/TFLOP) are the structural reversers. Tail risks include a rapid supply response from wafer capacity expansion, velocity of model optimization that reduces HBM intensity per workload, or customer concentration-driven churn if a large customer internalizes capacity — any of which would compress valuations quickly.

For portfolio construction, prefer asymmetric exposures that monetize the convexity: concentrated, hedged MU equity positions to capture margin upside, and optionality-style exposure to CRWV’s backlog through LEAPs staged to capacity-conversion milestones. Maintain liquidity buffers and explicit funding plans for potential equity raises at CRWV and set objective go/no-go readouts tied to MW online and realized gross margins.

Key Decisions for Investors

• Trade A — MU core long (12–36 months): Buy MU to 2–3% NAV. Hedge with a 12‑month 25% OTM put sized at 30% of notional and finance by selling a 12‑month 60% OTM call for a collar; target 2.5–4x upside if HBM pricing stays >current consensus, stop-loss on equity if puts breach implied vol >+50% from trade entry.
• Trade B — CRWV optionality (18–36 months): Buy CRWV Jan‑2028 LEAPs 30% OTM (initial 50% position). Add the second 50% tranche only after the company reports sequential quarterly capacity online >=20% of guidance or backlog conversion of >=10% realized revenue. Hedge tail risk with 2027 25% OTM puts sized at 20% of position to limit dilution/execution risk.
• Trade C — Relative-value pair (12–24 months): Long MU / Short INTC, 1.5:1 dollar ratio to neutralize semi-cycle beta; rationale is MU’s HBM convexity vs Intel’s broader capex-bearing IDM exposure. Exit when MU:INTC relative performance >150% or if memory spot pricing declines >30% YoY.
• Risk controls & alerts: Reduce CRWV exposure by 50% if company announces an equity raise >5% of market cap or if net cash burn guidance requires capital markets access; trim MU if HBM spot prices decline >20% sequentially or if SK Hynix/Samsung announce immediate additional HBM wafer fabs with <18‑month lead times.