Anthropic Has Already Raised $130 Billion Ahead of Its IPO. Here's What Potential Investors Need to Know.

Anthropic, the Claude developer, is reportedly preparing a late-Sep/Oct IPO and could seek as much as $100B, which would be the biggest IPO ever and potentially value the company around $2T. Revenue run-rate was $65B in July—about 6x year-over-year—after multiple large compute deals (e.g., 5GW with Amazon and 5GW with Google/Broadcom, plus reported $45B with Nscale and $35B with Lambda). The combination of rapid growth, substantial capital raising, and AI infrastructure buildout suggests a bullish outlook for AI equity appetite.

Analysis

This is less a single-company event than a signaling mechanism for the entire AI capex stack. A mega-IPO from a frontier-model lab would validate that model training/inference is still being financed off a very large future TAM, which should keep hyperscaler/cloud and custom-silicon ordering power elevated for another 12-18 months. The cleanest beneficiaries are the infrastructure toll collectors: AMZN and GOOGL for capacity monetization, and AVGO for TPU economics and design-wins; they monetize the build-out whether or not the end application economics are fully proven.

The more nuanced read is on NVDA: it benefits from the broader compute spend, but the disclosed mix of TPU and non-NVIDIA capacity suggests some demand is being arbitraged away from GPUs toward purpose-built silicon. That matters for relative performance more than absolute fundamentals. If investors start underwriting Anthropic as a quasi-public growth vehicle, the market may also re-rate private AI peers downward because scarcity value in late-stage rounds weakens once a marquee name has a public price tag.

Risk is two-sided. Near term, the IPO window can stay hot for days to weeks if the prospectus shows accelerating revenue and manageable dilution, but any hint that revenue is subsidy-driven, concentration-heavy, or capex-intensive will compress the multiple quickly. Over 6-18 months, the falsifier is simple: if compute commitments fail to convert into visible gross-margin leverage or if customer growth slows, the story shifts from growth compounding to financing dependence.

Key Decisions for Investors

• Favor AMZN and GOOGL over NVDA into the filing/roadshow window; the former monetize demand for compute immediately, while NVDA’s upside is more dependent on how much of Anthropic’s capacity mix stays GPU-heavy.
• Add AVGO on pullbacks as the cleaner custom-silicon beneficiary; if the IPO narrative extends the AI capex cycle, AVGO has a more durable claim on next-order design wins than pure-capex beta.
• Relative-value idea: long AMZN / short NVDA for 1-3 months if the prospectus confirms meaningful TPU and cloud diversification; thesis fails if GPU commitments expand materially or NVDA guidance re-accelerates.
• Watch the IPO filing for gross margin, revenue concentration, and remaining contractual compute obligations; if implied revenue multiple is extreme and margins are still negative, fade the first-day pop rather than chase it.
• If the deal prices on headline TAM rather than audited economics, consider short-dated put spreads on the most valuation-sensitive AI proxy basket rather than the name itself.