Anthropic is in advanced discussions for a funding round that could close within two weeks and value the San Francisco-based AI safety firm above $900 billion, according to sources familiar with the negotiations. The proposed valuation would place Anthropic behind only SpaceX and ByteDance among private companies globally, and ahead of every publicly traded firm outside the Magnificent Seven save Berkshire Hathaway. The round's structure remains undisclosed, though participants are expected to include sovereign wealth vehicles and at least one major cloud hyperscaler with existing compute commitments.

The timing compresses what was, until recently, a six-month expected negotiation window into a fortnight sprint. Anthropic's last disclosed round, a $7.3 billion Series C led by Menlo Ventures and involving Amazon, Alphabet, and Salesforce Ventures, closed in March 2024 at a $18.4 billion post-money valuation. That marked a 4.9x step-up from its prior $3.7 billion valuation ten months earlier. The prospective $900 billion figure—pending final term sheets—implies a 48.9x appreciation in thirteen months, a pace unseen in venture secondaries outside cryptocurrency manias and the 1999 telecoms buildup.

Three factors converge here. First, Anthropic's Claude 3.5 Sonnet has captured enterprise wallet share from OpenAI in regulated verticals—financial services, healthcare systems, and government contractors—where constitutional AI and interpretability matter for procurement committees. Second, the frontier model market is bifurcating into OpenAI and everyone-else, with Anthropic emerging as the defensible alternative for institutions unwilling to sole-source from Microsoft's effective subsidiary. Third, compute leverage is inverting: Anthropic's reported $2.7 billion annualized revenue run rate, if accurate, implies a 333x revenue multiple at $900 billion—a figure that only holds if allocators believe the firm's training efficiency and inference cost curves will compress faster than OpenAI's, turning today's model margin into tomorrow's operating leverage.

The valuation also signals the end of AI's brief discount window. Twelve months ago, crossover funds treated foundation model builders as capital-intensive science projects with uncertain monetization. Now they are priced as inevitable quasi-utilities. The shift matters for two reasons. Anthropic's constitutional AI framework—its core technical differentiation—relies on recursive model oversight, which demands 30-40% more training compute than standard RLHF at equivalent capability. If investors are willing to fund that cost structure at near-trillion-dollar valuations, the market has decided AI safety is no longer a margin tax but a moat. That assumption will be tested when Anthropic's next model generation ships, likely in Q3 2025, and enterprises compare its price-per-token against GPT-5 or Gemini Ultra's successors.

Allocators should track three follow-on events. First, whether Amazon or Alphabet exercise pro-rata rights or allow dilution—both hold board seats and compute agreements that hinge on ownership thresholds. Second, the composition of new capital: if sovereign funds dominate, Anthropic gains geopolitical optionality but invites the export-control scrutiny that has already touched OpenAI and DeepMind. Third, Anthropic's hiring velocity in the next 90 days—particularly whether it pulls senior researchers from OpenAI, Google, or Meta, signaling confidence in its technical roadmap or desperation to justify the valuation.

The round's two-week close timeline, if it holds, means term sheets are already circulating and lead economics settled. What remains is syndicate assembly and regulatory pre-clearance in jurisdictions where $900 billion private valuations trigger merger-review thresholds even absent a change-of-control. The speed suggests Anthropic is not testing the market but executing a pre-negotiated plan, likely shaped by compute procurement deadlines and competitive pressure from OpenAI's own rumored raise. The frontier model arms race is now a capital race, and the entry price just moved.