AWS boss explains why investing billions in both Anthropic and OpenAI is an OK conflict

Amazon/AWS made a reported $50 billion investment in OpenAI (after ~ $8 billion invested in Anthropic), which AWS CEO Matt Garman described as a manageable conflict of interest given AWS’s long experience competing with partners. Garman argued multi-model routing and model specialization (planning, reasoning, code completion) will let cloud vendors mix partner and first‑party models, a strategic necessity because these models were available on rival Microsoft Azure. Anthropic’s recent $30 billion round included multiple investors also backing OpenAI, underscoring cross‑ownership among major AI backers and competitive dynamics in the cloud/AI market.

Analysis

AWS’s emergent role as the control point for model routing creates a durable margin opportunity that is often underappreciated: even a low single-digit take-rate on customers’ AI inference spend converts into high-margin, recurring revenue because CPU/GPU, storage and networking are already fixed-cost buckets for hyperscalers. If global enterprise AI inference grows into the mid‑to‑low tens of billions in the next 24 months, a 5–8% routing/management fee implies several billion of incremental annual revenue for whoever anchors routing — a lever that compounds faster than new IaaS capacity buildouts.

Second-order winners extend beyond the hyperscalers: inference-optimized silicon vendors and managed-service integrators capture outsized unit economics as customers buy “models + ops” bundles rather than raw compute hours; conversely, independent model vendors that lack deep commercial partnerships may get squeezed into commodity pricing or relegated to niche verticals. The ability of a cloud provider to preferentially route traffic to its own or preferred models is the single biggest latent moat inside existing contracts — it magnifies switching costs without changing sticker prices and can materially alter lifetime customer value over 12–36 months.

Key reversal risks are regulatory and enterprise governance: antitrust or procurement rules that mandate model neutrality, large public-sector contracts demanding open-source stacks, or a high-profile model failure/exfiltration event would force rapid de‑risking and re-pricing of routing services. Shorter-term catalysts to watch are productization milestones (paid routing SLAs, per-inference billing), major enterprise migrations, and GPU supply shocks; these will move relative share quickly and are actionable within quarterly to annual windows.

Key Decisions for Investors

• Overweight AMZN (12–24 months): establish a 2–4% NAV position via a call‑spread LEAP to express AWS routing monetization (buy 12–18 month 20–30% OTM calls, sell 12–18 month 60–80% OTM calls). Rationale: asymmetric upside if routing take‑rates scale; cap premium outlay to limit downside to ~2–4% NAV, target 2–4x payoff if AWS captures material routing share within 18 months.
• Relative pair — long AMZN / short MSFT (6–12 months): implement a 1.0–1.2x notional pair (e.g., long AMZN equity or calls, short MSFT equity or buy puts) sized to 1–2% NAV net. Rationale: arbitrage execution/monetization optionality at AWS vs. Microsoft’s bundled model strategy; payoff if market re‑rates AWS for routing economics. Risk: keeps exposure low because MSFT retains durable cloud cash flows; stop-loss at 15% adverse move.
• Tail hedge / regulatory rotation — long ORCL (9–12 months): small 0.5–1% NAV position via calls or equity. Rationale: Oracle is a natural beneficiary if customers de‑risk hyperscaler lock‑in or accelerate on‑prem/DB licensing; acts as a defensive hedge to hyperscaler regulatory pain with asymmetric upside and limited cost.
• Operational triggers & risk controls: set alerts for three catalysts — announcement of paid routing SLAs/pricing, any major enterprise contract wins referencing multi‑model routing, and published regulatory inquiries into model favoritism. Rebalance options positions to trim 25–50% of upside exposure upon receipt of confirmed routing revenue guidance or large enterprise adoption signals.