Goldman Sachs flags 3 chip stocks to buy and 3 to avoid into Q1 earnings

Goldman Sachs sees broad fundamental upside across most semiconductor sub-sectors heading into Q1 earnings, naming Teradyne as its highest-conviction buy and flagging Applied Materials and AMD for upside. The bank highlights compute, memory, storage and capital equipment strength and notes AMAT's roughly 60% exposure to etch and deposition as a re-rating catalyst. Goldman expects hyperscaler capex to tilt upward supporting server CPU and ASIC demand, favors HDD/NAND ecosystems and analog semis with industrial/defense/datacenter exposure, and warns KLA, Onsemi and Arm (Sell) may lag due to DRAM skew, automotive/CIS/SiC headwinds, and smartphone weakness respectively.

Analysis

The current tape is setting up as a classic equipment/automation-led re-rating rather than another pure software or OEM multiple expansion. As hyperscalers and a handful of large memory fabs pull orders forward, expect acute capacity-driven margin upside for testers and process-tool vendors because utilization gains are non-linear — a 5-10ppt rise in fab utilization typically translates into 20-40% incremental profit flow-through at the equipment level, not a 1:1 revenue lift. That dynamic creates convexity into Q1 results: beats will compress required capex visibility risk and force multiple expansions faster than revenue growth justifies.

Second-order winners are suppliers tied to lithography/etch/deposition consumables and test IP flows; their lead times shorten margin normalization for platform equipment vendors and raise switching costs for customers. Conversely, systems integrators and boutique OEMs with concentrated exposure to single hyperscalers face sharper downside if spending reverts — dispersion will widen across the capex/value chain. For semi names tied to both server AI and legacy end-markets, the near-term beat-or-miss run is driven more by book-to-bill and backlog conversion than by end-demand trends alone.

Key risks: a step-down in hyperscaler cadence (policy-driven or inventory digestion) or a surprise memory supply ramp would invert the thesis quickly — these are measurable within 30–90 days via vendor backlog disclosures and WFE (wafer fab equipment) billings. Time horizons: tradeable signals around Q1 prints (days–weeks), fundamental re-rating over 3–12 months if utilization remains elevated, structural AI-driven demand playing out over multiple years. Monitor book-to-bill, fab utilization, and lead-time inflation as stop/start indicators.

Consensus blind spot: the market underprices asymmetric upside in capital equipment/T&M where order pull-ins create earnings beats with limited incremental capex needed. That favors convex option-like exposure to tester and process-tool names while shorting execution-sensitive integrators that are already priced for perfection.

Key Decisions for Investors

• Long TER (Teradyne): Buy shares sized 3–5% of equity risk and overlay a 3–6 month call spread (buy 1x near-the-money call, sell 1x higher strike) to cap cost. Rationale: asymmetric upside if tester order flow confirms; target 25–40% upside in 3–6 months, max loss = premium + share drawdown; stop-loss on shares at -12%.
• Long AMAT (Applied Materials): Initiate on weakness or on a sub-5% pullback; target 20–30% total return over 6–12 months as etch/deposition utilization re-rates multiples. Hedge by selling out-of-the-money calls (covered call) for income; set a protective put if shares gap down >15% into earnings.
• Pair trade — Long AMD / Short SMCI (notional matched): Allocate equal notional long AMD (3% portfolio) and short SMCI (3%) to capture dispersion between a diversified silicon vendor and a high-beta systems integrator. Expect the spread to widen 15–30% in 3–6 months if tool demand outperforms systems revenue; hard stop if spread compresses >10% against position within 30 days.
• Event hedge: Buy 3–6 month puts on a semiconductor equipment ETF or on AMAT/TER (cheap liquidity permitting) sized to cover 20–30% of portfolio concentrated weight. Cost = insurance premium; payoff protects against a 30–50% drawdown triggered by hyperscaler capex retraction or sudden memory oversupply.