2 Artificial Intelligence (AI) Stocks That Won in 2025 Are Losing in 2026. Why This Is a Buying Opportunity

Palantir gained 135% in 2025 but is down ~18% in Q1 2026; revenue growth has accelerated for 10 consecutive quarters with 70% sales growth in the most recent quarter, and its Foundry AI platform plus strong government/Maven deployments support a durable growth runway. Broadcom rose 49% in 2025 and is down ~10% in Q1 2026, yet it projects about $100 billion in custom AI chip sales in fiscal 2027 and claims AI hardware margins will remain stable, underpinning a multi-year revenue and data-center networking growth opportunity.

Analysis

Palantir’s most durable asset is not its models but the data plumbing it wires into customers’ operations. That creates a multi-year, contract-level annuity where marginal revenue growth compounds through additional module sales and higher data ingestion, not just model re-training — a structural moat that increases customer lifetime value and raises the cost of displacement. Expect this to mechanically lift gross retention and make short-term churn metrics less informative as a valuation input.

Broadcom’s custom silicon model creates downstream lock‑in across datacenter stacks: once a hyperscaler accepts a multi‑generation co‑design cadence, they internalize integration risk, software toolchains, and supply relationships (TSMC, packaging partners), which raises switching costs and creates embedded backlog that spans multiple fiscal years. That dynamic can cap pricing power of general-purpose accelerators in selected accounts and convert what looks like cyclical capex into recurring roadmap revenue for Broadcom and its foundry partners.

Key catalysts and risks are asymmetric in time: over 0–3 months, guidance season and hyperscaler spend commentary will drive volatility; 6–18 months will reveal multi‑gen chip commitments and government program renewals that validate structural lock‑in; 2–5 years is when embedded network effects become primary value drivers. Tail risks: a coordinated hyperscaler push for open accelerator standards, rapid model commoditization, or export controls could compress premium margins quickly and re-rate multiples across the group. The market’s pullback is a timing, not a correctness, signal — the question is pace of conversion from contracts to predictable cash flow.

Key Decisions for Investors

• Long PLTR (size 1–2% NAV) via a 12‑month call spread: buy 12‑month calls ~30% OTM and sell 12‑month calls ~80% OTM to cap cost. Rationale: capture multi‑quarter contract cadence and retention re‑rating; target 2.5x premium if execution accelerates, max loss = premium paid.
• Long AVGO via 18‑month LEAPs (size 1–1.5% NAV) financed by selling 3–6 month covered calls (5–10% OTM). Rationale: capture multi‑gen custom chip backlog while monetizing near‑term carry; target 30–60% upside in 12–24 months, downside = premium/stock move against position plus opportunity cost if guided shipments slip.
• Pair trade for event risk hedging: long PLTR equity (0.75% NAV) / buy protection on AVGO (6‑month puts ~10–15% OTM, 0.5% NAV). Rationale: protects portfolio if a macro/AI capex pullback knocks Broadcom while retaining exposure to Palantir’s secular data‑ops thesis.
• Short a near‑term NVDA call spread (3–6 months, narrow) as a hedge against AI‑capex peak narrative (size to offset 25–50% of options vega). Rationale: if hyperscaler commentary shows spend moderation, NVDA will reprice fastest; this trade offers limited risk with positive carry if implied vol falls.