Dan Niles: 'Nowhere Near the Bottom' in Software

Dan Niles says tech stocks have fallen since the outset of the war in Iran and that software is 'nowhere near the bottom,' signaling near-term downside and continued pressure on software valuations. He expects agentic AI to drive the next phase higher for tech, implying medium-term upside for AI-exposed companies despite current caution.

Analysis

Macro headlines have amplified a positioning-driven unwind in software, but the mechanism is important: flows out of long-duration, high-valuation software into cash/defensives forces multiple compression that can persist well after headlines fade. Over the next 2–12 weeks expect headline-driven volatility (CTA rebalances, fund redemptions), while the medium-term (6–24 months) story will be decided by who converts AI R&D into recurring, high-margin revenue and who suffers elongated sales cycles.

The next phase of upside will be highly concentrated: GPU compute, hyperscaler cloud services, and a small set of agentic-AI application vendors that build proprietary data moats benefit disproportionately. Second-order winners include data-center power & cooling suppliers, TSMC/ASML exposure via specialty nodes, and enterprise software that becomes a workflow layer for AI agents; losers are mid/small-cap SaaS with weak net retention, professional services relying on discretionary cloud transformation spend, and legacy software that cannot embed agentic workflows without heavy re-architecture.

Key catalysts that could reverse the current derating are: (1) demonstrable, monetizable agentic-AI deployments (productivity metrics published by large customers) within 3–9 months; (2) clearer GPU supply/demand dynamics or pricing that sustain enterprise capex; and (3) a durable de-escalation in geopolitical risk that allows risk-on flows to return. Tail risks that would extend the drawdown include escalation that disrupts energy or chip supply lines, or a broader liquidity shock from rapid Fed repricing over the next 1–3 quarters.

Contrarian read: the market is treating every software name as equally fragile when agentic AI is likely to re-concentrate economics into a narrow set of infra and workflow owners — that makes concentrated, asymmetric bets attractive. The optimal portfolio tilts long infrastructure + high-Quality AI-native franchises, hedged by short-duration bets on cyclical/discretionary SaaS exposure, with volatility hedges around macro headlines.

Key Decisions for Investors

• Long NVDA (NVDA) via a 12–18 month call spread: buy 2027 Jan 400C / sell 2027 Jan 700C to express asymmetric exposure to GPU-driven agentic AI adoption. Rationale: captures upside if agentic deployments accelerate; max loss = premium paid, target >2x payoff if NVDA sustains leadership in datacenter GPUs within 12–24 months.
• Core cloud tilt: buy MSFT and GOOG (equal-weight) size for 6–18 months to own software + hyperscaler AI stack. Risk/reward: expected lower volatility than single-name semis, captures recurring revenue upside as customers shift to cloud-native agentic services; hedge with 1–3% portfolio put protection if headline risk spikes.
• Short small-/mid-cap SaaS exposure via IGV (software ETF) or a basket of high-valuation SaaS (e.g., cash-funded short futures/options) for 3–9 months. Rationale: captures continued multiple compression and elongated deal cycles; target 20–40% downside in stressed scenario, cap risk via stop-loss or limited-duration options.
• Macro hedge: buy short-dated VIX calls or put spreads on SPX (30–90 day) ahead of major geopolitical/earnings windows. Rationale: protects asymmetric long infra/AI positions from headline-driven liquidity shocks where most downside occurs in compressed windows.