Amazon seen as undervalued despite AI capex concerns: Jefferies

Jefferies says Amazon may offer investors meaningful upside despite near-term headwinds. Analysts note concerns around outsized AI-related capital expenditures and AWS growth trailing peers have damped sentiment, leaving the stock a frustrating laggard but with potential upside per the firm.

Analysis

The capital-intensive buildout for large-scale AI workloads creates a non-linear demand profile across the supply chain: networking and switch vendors (Arista, Broadcom) and data-hall integrators (Switch, Equinix) should see near-term order visibility and margin tailwinds, while Amazon’s continued investment in custom silicon (Trainium/Inferentia/Graviton evolution) is a latent offset to third-party GPU spend — plausibly shaving 10–30% of AWS inference GPU dollar demand over 18–36 months if adoption accelerates. That dynamic also compresses the addressable upside for pure-play GPU beneficiaries (shorter-cycle NVDA exposure) while improving Amazon’s long-run unit economics, setting up a classic trade between short-term FCF compression and durable margin optionality.

Risk segmentation matters by horizon. Over days–weeks, stock movement will be dictated by guidance and positioning flows; knockouts come from an earnings print that materially raises or cuts capex guidance. Over 3–12 months, key catalysts are AI instance utilization, customer case studies quantifying $/inference or $/token economics, and capex cadence — a shift to utilization >60% across AI clusters would materially re-rate margins. Tail risks include prolonged weak enterprise AI adoption, sudden energy cost spikes or regulatory constraints on data-center buildouts; any of these could keep capex elevated while utilization remains sub-par, extending the valuation drag.

The consensus seems to over-penalize capex as a permanent margin problem rather than a timing issue. If Amazon is buying durable cost advantage (lower $/token via custom silicon + scale), the upside is asymmetric: a sustained re-acceleration in AWS AI monetization could deliver 30–50% upside in 12–24 months while the downside, if capex proves wasteful, is a slower drawdown confined to FCF compression. That asymmetric payoff favors optionality via long-dated, limited‑downside structures and relative-value plays that capture re-rating while hedging macro/capex risk.

Key Decisions for Investors

• Buy-dated LEAP call (12–24 month) on AMZN to capture asymmetric upside from an AWS AI monetization inflection — size to 1–3% of portfolio; reward scenario: 30–100%+ if utilization and pricing improve; risk: full premium loss if capex fails to convert to demand.
• Pair trade — Long AMZN / Short MSFT (equal notional, 6–12 month horizon) to express idiosyncratic AWS recovery vs broader enterprise AI exposure. Target relative return: +30–50% if AWS re-accelerates; stop-loss if AMZN underperforms MSFT by 10% to cap drawdown.
• Tactical supplier long: Buy Arista (ANET) or Broadcom (AVGO) exposure (3–9 months) to play near-term networking and ASIC demand from hyperscale AI builds. Reward: beating consensus order guides and margin leverage; risk: Amazon vertical integration incremental spend shifting away from vendors.
• Hedge and staging: If initiating outright AMZN exposure, buy 6–9 month put protection for ~20–30% downside coverage or sell 3–6 month covered calls to finance premium. This preserves upside optionality while limiting short-term capex/positioning shocks.
• Event trigger watchlist: enter/add to positions on one of two triggers — (A) earnings or guidance that lowers future capex trajectory by >15% YoY, or (B) a customer adoption signal (public win or utilization >60% across AI instance families) — each should prompt a 25–50% increase in position size depending on risk tolerance.