You Can't Buy Anthropic Stock Yet, but You Can Buy These 4 AI Stocks Instead

Anthropic said annualized revenue has roughly tripled since the end of 2025 to over $30 billion in early 2026, underscoring surging demand for Claude and AI agents. The article highlights four key infrastructure beneficiaries—Amazon, Alphabet, Broadcom, and Nvidia—linked to hundreds of billions of dollars of compute and chip spending, including Amazon's $100 billion+ AWS commitment over 10 years and Alphabet's up to 1 million TPUs. Anthropic's draft IPO filing and rapid AI infrastructure build-out are constructive for the named suppliers and the broader AI supply chain.

Analysis

The key second-order read is that this is no longer just an AI demand story; it is a capacity-prepayment story. When a frontier model provider signs multi-year commitments across AWS, GCP, custom ASICs, and GPUs, the economic value shifts upstream to whoever can lock scarce power, networking, and chip supply before the next wave of agentic workloads hits. That favors the infrastructure names with the best mix of pricing power and supply-chain control, while also implying a rising barrier to entry for smaller cloud and chip vendors that cannot reserve enough data-center capacity.

The most important contrast is between monetization certainty and valuation dispersion. AMZN and GOOGL likely get the cleanest duration benefit because cloud revenue visibility improves as compute commitments become more contractual, but AVGO may be the underappreciated torque play if TPU/networking build-outs remain on schedule into 2027. NVDA remains the highest-beta expression, but its upside here depends less on raw demand and more on whether it can defend share in agentic inference versus cheaper custom silicon; that makes it the most sensitive to any evidence of workload migration away from GPUs.

A contrarian risk is that the market may be overpricing linear demand conversion from model usage into supplier earnings. The build-out can be delayed by power interconnects, permitting, and deployment cadence, so stock reaction could outrun actual revenue recognition by 2-4 quarters. Another hidden risk is customer concentration: if the largest model labs increasingly negotiate on price as they scale, margins at the infrastructure layer can compress even while volumes surge.

Near term, the catalyst stack is favorable into the IPO window and any follow-on disclosures about compute expansion, but the better trade may be to own the picks-and-shovels with longer backlog visibility rather than the model layer itself. The cleanest setup is to express the thesis through relative value: long infrastructure beneficiaries with contracted visibility versus a basket of software names still being valued on optionality rather than cash-flow conversion.

Key Decisions for Investors

• Long AMZN / long GOOGL into the next 1-3 quarters: best risk-adjusted exposure to contracted AI spend; favor AMZN on AWS re-acceleration, GOOGL on TPU monetization and cloud backlog optionality.
• Add AVGO on pullbacks for a 6-12 month hold: highest convexity to custom silicon and AI networking demand, with the market still underappreciating the 2027 TPU ramp.
• Hold NVDA as a tactical long only, not a core pair leg: use as a 1-2 quarter momentum trade, but cap size because custom-ASIC adoption can pressure GPU share in inference workloads.
• Pair trade: long AMZN or GOOGL / short a broad software basket over the next 6 months: thesis is that infrastructure monetizes first while application-layer AI spending remains more speculative.
• Use call spreads instead of outright longs on NVDA into catalyst events: upside remains strong, but implied volatility and competitive-share risk make defined-risk structures more attractive.