'Big Short' investor Michael Burry says Anthropic is 'eating Palantir's lunch' as enterprise AI spending shifts

Anthropic's enterprise traction is highlighted as ARR jumped from $9B to $30B in months per Ramp/Ara Kharazian analysis, with Anthropic taking ~73% of new enterprise AI spending and nearly 1 in 4 Ramp businesses now paying for Claude (vs 1 in 25 a year ago). Michael Burry says Anthropic is "eating Palantir's lunch," noting Palantir took ~20 years to reach ~$5B while Burry holds a short in PLTR; short-seller Andrew Left has also recently shorted Palantir. The piece signals a shift in enterprise adoption toward plug-and-play AI (positive for Anthropic/peers like Databricks) and negative investor sentiment toward Palantir.

Analysis

The market dynamic now favors API-first, low-friction model suppliers because procurement friction is the primary limiter of C-suite AI budget allocation — reducing integration time from quarters to weeks materially shortens payback on customer acquisition and multiplies ARR growth without proportional sales headcount. That structural advantage cascades: higher gross retention and lower CAC for a leading plug-and-play vendor translates into far faster revenue scaling than legacy incumbent bundles that depend on bespoke services, making valuation multiples a function of go-to-market efficiency rather than pure model quality.

Second-order winners include cloud infra and inference-capacity providers (where incremental enterprise API spend lands) and vendors selling endpoint security, data governance, and vector search; losers include system integrators and bespoke integration revenue streams that get compressed as standard connectors and SDKs proliferate. This also creates concentrated counterparty risk: platform outages, pricing power shifts, or a single-signature vendor commitment from large enterprises would transmit revenue swings across many suppliers and MSPs.

Key catalysts and reversals are primarily operational: meaningful price increases, model regressions, or enterprise procurement reversals (renewal/ARPU step-downs) can unwind adoption narratives in 3–9 months, while regulatory or national-security reviews could extend impact to multi-year horizons. Near-term volatility will be amplified by activist/short positioning in public analogues; absent a product-level failure, the adoption trend still looks like a multi-quarter story rather than a binary overnight shift.

The consensus underestimates concentration and re-contracting risk: if one private provider captures a dominant share, incumbent vendors can still claw back margin via feature parity, bespoke data advantages, or exclusive integrations with large enterprise ecosystems. Watch monthly active paid logos, ARPU per paying account, and cloud egress/inference spend per customer as 30–90 day leading indicators of durable enterprise lock-in.

Key Decisions for Investors

• Short PLTR via put spreads (3–12 month expiries): buy a 3–9 month put and sell a lower strike to finance premium. Position size: 1–3% net portfolio risk; target 30–50% downside in 6–12 months, max loss = premium paid.
• Pair trade — short PLTR / long MSFT or NVDA (6–24 months): overweight cloud/inference exposure vs government-centric analytics. Risk/reward: asymmetric; cap potential drawdown on the long leg with covered calls if NVDA/MSFT run >25% in 3 months.
• Long RAMP (ticker: RAMP) or buy 6–12 month call spreads: thematic bet on a data/telemetry provider monetizing AI adoption signals. Target 2–3x upside if enterprise telemetry becomes a standard procurement signal; cap losses to option premium.
• Directional infra play — long NVDA (options) or buy 6–12 month call spreads on major cloud names (MSFT/AMZN/GOOG): captures lift from rising inference API spend. Hedge with a small short put leg to improve cost basis but monitor earnings-driven gamma risk closely.