Software stocks selloff: PLTR, MSFT drop on Anthropic’s ‘Mythos’ model fears

Palantir shares fell ~7% and the iShares Expanded Tech-Software ETF dropped ~3.7% after new AI product launches from Meta and Anthropic and a critical note from short seller Michael Burry. Anthropic’s Mythos model reportedly delivered a 17 percentage-point improvement on Terminal Bench 2.0 and a 13-point gain on SWE benchmarks versus Opus 4.6, raising near- to medium-term disruption risks for IT/software incumbents; Burry warned Anthropic is threatening Palantir’s commercial opportunity (citing Anthropic’s ARR jump from ~$9B to ~$30B). Analyst commentary and the sector-wide selloff signal elevated competitive and sentiment pressure on software names.

Analysis

The market reaction is trading a near-term narrative — that a step-change in model capability commoditizes parts of enterprise software — while underweighting two frictions that slow real-world displacement: integration+ops cost and regulated/government workflows. Incumbents that monetize through bespoke deployments and security certifications (high-touch sales, on-prem integrations, Fed/DoD accreditations) can defend value capture for 12–36 months even if baseline model capability improves rapidly.

Second-order winners are outfits that own the capital-intensive part of the stack (data-center OEMs, custom server builders, and systems integrators that can retrofit agentic workflows into legacy estates). Conversely, vendors whose revenue mix is concentrated in small-to-medium commercial automation products with high price elasticity are at elevated risk of low-double-digit share loss within 12–24 months absent rapid margin reengineering.

Catalysts that will re-rate names are measurable and time-bound: enterprise pilot conversions and ARR renewal retention rates over the next 2–8 quarters, public benchmarks of production stability/latency under customer loads, and any regulatory or procurement changes for classified workloads. The prudent path is trading around these binary readouts — favor optionality and defined-loss structures rather than naked directional exposure until real adoption curves are visible.

Key Decisions for Investors

• Pair trade (tactical, 3–9 months): Short PLTR (size 0.5–1% NAV) vs Long SMCI (size 0.5–1% NAV). Rationale: capture downside from commercial-displacement risk while owning hardware demand for AI ops. Use a max-drawdown stop of 20% on either leg; target asymmetric payoff of 2:1 if PLTR falls ~30% while SMCI rallies 20–40.
• Defined-risk short on PLTR (conservative): Buy 3–6 month put spread (sell higher strike, buy lower strike) sized to risk 0.25–0.5% NAV. This limits loss if the market misprices government resiliency and preserves upside if negative sentiment continues; target 2–3x return if implied vol normalizes and shares gap down on successive weak renewals.
• Long META (selective, 6–12 months): Buy equity or 9–12 month calls (small size 0.5–1% NAV) to play higher-margin ad/engagement leverage and balance-sheet optionality. Tail risk: ad slowdown; hedge with a 0.5% NAV short in cyclic ad-exposed names if macro data softens. Target 20–40% upside if ad mix and ARPU beat expectations.
• Rebalance exposure to incumbents with large government revenue buckets (risk-manage): Avoid naked long exposure in names where >30% revenue is commercial SaaS without embedded security certifications. If owning, prefer structured protection (collars or put spreads) across a 6–18 month horizon to protect against rapid margin compression.