Anthropic Prices IPO at Up to $1.4 Trillion

Anthropic Prices IPO at Up to $1.4 Trillion

While Nvidia was imploding, Anthropic was finalizing the biggest tech IPO in memory. The company set a price range of $70-$78 per share for 825 million shares, targeting a fully diluted valuation up to $1.404 trillion and raising as much as $64.4B [1]. A separate $26B private placement includes checks from Google ($10B), Nvidia, AMD, Broadcom, Micron, Microsoft, and TSMC — basically every major chip and cloud player wanting a seat at the table [2].

Trading is expected to open on Nasdaq under "ANT" around November 19, with run-rate revenue reportedly around $81B as of October 31 and the offering multiple times oversubscribed [3]. That's a staggering revenue number for a company barely out of its research-lab phase a few years ago.

The timing is almost comedic: Anthropic pricing a trillion-dollar IPO the same day Nvidia's CEO resigns in scandal. It's a reminder that the AI trade right now has two speeds — infrastructure (volatile, concentrated, now leadership-fragile) and model layer (ascendant, cash-flowing, investor-starved for allocation). Watch whether Nvidia's wobble cools appetite for ANT before the 19th.

Claude Fable 5.2 Targets Long-Horizon Agentic Work

Anthropic didn't just file paperwork this week — it shipped. Claude Fable 5.2 (plus a Mythos 5.2 variant for verified orgs) landed November 9 as Anthropic's most capable model yet for coding, reasoning, and vision, with up to 30% faster outputs and a 1M-token context window [1]. Pricing actually dropped — $8/$40 per million input/output tokens standard, $16/$80 for fast mode — which is notable given how much infrastructure cost usually moves in the opposite direction [2].

It's available across Claude's own platform, AWS Bedrock, Google Cloud, and Microsoft Foundry on day one — maximum distribution, minimum friction. The pitch is squarely about multi-step autonomous tasks and agent orchestration, not chat.

Early users are reporting strong results in multi-agent trading and coding setups, and it's quickly being positioned as the model of choice for complex agentic workflows [3]. Combined with Anthropic's IPO timing, this looks like a company stacking proof points right before investors start asking hard questions about durability.

Simular Launches Box: A Shared Computer for AI Agents

Simular's Box, launched November 12, is a small story next to Nvidia and Anthropic but points at where the real product work is happening. It's a cloud-based Ubuntu desktop built specifically for AI agents — ChatGPT, Claude, Codex, Gemini — to share files, apps, and logins, hand off tasks between each other, and keep working after you close your laptop [1].

The free tier gives one desktop per account with 1,000 daily credits and auto-sleep; paid tiers start around $50/month for more instances and credits, with MCP integration for easy connection into existing agent stacks [2].

This is the unglamorous infrastructure layer that multi-agent orchestration actually needs — a persistent, shared environment where agents don't just talk to each other, they work together on the same machine. Expect more of this category as "agent runs locally in your terminal" stops being good enough.