Evercore sees Anthropic S-1 filing as positive for cloud partners

Evercore said Anthropic’s S-1 disclosures offer a positive read-through for the hyperscale cloud providers that supply and invest in it. Anthropic reported revenue rising from $41 million in 2023 to $386 million in 2024 and $4.6 billion in 2025, then reaching $16.3 billion in the first half of 2026; it said demand exceeds available compute capacity. The filing also cites $200 billion in NVIDIA-based compute commitments, while capacity constraints and take-or-pay commitments remain risks to Anthropic’s expansion.

Analysis

Anthropic’s capacity constraint is a demand signal, but not yet proof of attractive returns for its suppliers: the key transmission is whether cloud and chip spending converts into durable, profitable workloads rather than expensive capacity commitments. Alphabet has a potential two-sided benefit: more AI workloads can support cloud demand, while Claude’s Workspace integration may improve Google’s enterprise AI offering—and also gives customers another model option, intensifying competition for Google’s own AI products. NVIDIA is exposed to the near-term buildout, but large commitments make utilization and customer concentration important alongside demand growth.

Near term, the read-through is modestly positive for GOOG and NVDA; do not extrapolate Anthropic’s reported growth into supplier earnings without contract economics or evidence of realized deployments. Over 1–3 months, watch cloud capex commentary, capacity delivery, and whether Anthropic’s commitments translate into recognized supplier revenue. Over 6–18 months, the central risk is overbuild: if AI demand or monetization disappoints, take-or-pay obligations and underused infrastructure could pressure returns. The China data allegations add a separate, potentially escalating regulatory and access risk, but the reported claims do not establish broad market impact.

Contrarian point: constrained supply can support pricing and workload allocation, but it can also cap customer growth; the market may be treating backlog-like demand as guaranteed revenue. The thesis weakens if cloud providers flag lower AI utilization or returns, Anthropic revises commitments, or capacity additions outrun monetizable demand.

Key Decisions for Investors

• Treat the filing as a positive watch item for Alphabet (GOOG), not a standalone buy signal. Add only on weakness if Google Cloud commentary confirms incremental AI workload conversion; reassess if capex rises without corresponding cloud growth or improved returns.
• For NVIDIA (NVDA), avoid chasing the demand narrative. Monitor supplier revenue conversion and hyperscaler utilization; a slowdown in orders or evidence of excess capacity would falsify the near-term positive read-through.
• Track Anthropic’s disclosed compute commitments against actual capacity delivered, utilization, and customer monetization. Those details are needed to distinguish durable supplier demand from contingent or economically unattractive commitments.
• Watch for China regulatory findings or restrictions that affect access to AI models or cross-border data flows. The current allegations are a risk alert, not evidence of a sector-wide disruption.