Back to News
Market Impact: 0.4

2 Growth Stocks That Could Go Parabolic

NVDAINTCNFLX
Artificial IntelligenceTechnology & InnovationCompany FundamentalsCorporate EarningsPrivate Markets & VentureInvestor Sentiment & Positioning

Sandisk reported fiscal Q2 revenue growth of 61% YoY and net income jumping from $104M to $803M, with data-center revenue up 76% and the stock up ~179% in 2026 (P/E ~15). CoreWeave has locked-in committed revenue via long-term contracts, operates 43 data centers (11 added in 2025), saw revenue ~300% higher since IPO, and secured a $2B strategic investment from Nvidia, but remains unprofitable with high debt and trades at ~6.7x trailing 12-month sales.

Analysis

Nvidia is the hidden fulcrum in this micro-ecosystem: its ability to allocate GPU supply and embed software/driver lock‑ins creates optionality that benefits suppliers of compute while selectively disadvantaging independent renters who can’t secure preferential inventory. A supplier that enjoys privileged access to chips can compress unit costs and accelerate capacity growth without the same financing drag, which amplifies share gains for the preferred partner and raises customer acquisition costs for rivals. Storage vendors with high gross margins (and flexible NAND sourcing) are second‑order beneficiaries because lower $/GB materially reduces marginal cost of training and inference at scale, improving data-center economics and expanding addressable compute demand. Conversely, companies that are capital intensive and levered to growth capex face a double squeeze: rising opex for energy/cooling plus higher borrowing costs if capital markets tighten, elevating dilution risk on a 6–18 month horizon. Near-term catalysts to watch are (1) GPU allocation announcements and batch shipments over the next 90 days, (2) NAND pricing signals from contract renewals in the next 1–2 quarters, and (3) any financing or covenant events from capital‑hungry infra providers in the coming 6–12 months that could force dilutive equity or asset sales. Tail risks include an abrupt shift to more efficient model architectures that reduce GPU hours per training run and an export/regulatory shock restricting high-end accelerator shipments, either of which would compress revenue growth across the stack. The consensus skews simplistic: it treats compute demand as linear and ignores supplier capture of margin through preferential chip allocation and software lock‑in. That asymmetry means owning upstream, higher‑margin suppliers (and hedging renters with capital exposure) is a cleaner way to play the AI infrastructure cycle than owning high‑growth, balance‑sheet‑dependent renters outright.