Anthropic IPO buyers see valuation as low as $1.1T

Anthropic’s planned IPO, once floated at up to $2T, is now expected to price as “low” as $1.1T, prompting sharp debate over whether even a trillion‑dollar valuation is remotely justified. Commenters question the company’s moat in an increasingly commoditized AI market, pointing to fast‑improving open‑weight and Chinese models, slowing revenue growth, heavy capital commitments for compute, and mounting regulatory and political headwinds. Some see Anthropic and OpenAI as potential long‑term giants if they truly achieve superior “digital intelligence,” while others warn that a realistic repricing into the mere hundreds of billions could puncture the broader AI financing bubble.

Valuation and IPO Pricing

  • Many see even $1.1T as wildly overvalued; suggested “fair” ranges cluster around $200–400B, with some arguing sub‑$100B.
  • Others argue that if Anthropic really sells “intelligence” at near‑human or superhuman levels, $1T could be cheap, though they acknowledge that’s a big “if.”
  • Commenters note the target keeps sliding downward ($2T → $1.5T → $1.1–1.3T) and speculate the IPO could be delayed to avoid forcing large investors to recognize losses versus private marks.
  • Several explain IPO mechanics: initial valuation mainly sets how much cash Anthropic raises and where early stakeholders exit; post‑IPO price moves don’t change the money the company already took in.

Financials and Growth

  • Reported ARR figures (e.g., $80B annualized) are contested: some call them “bullshit non‑standard metrics,” pointing out Q3 growth slowed and losses remain massive.
  • One side cites ~85% gross margins on inference and claims subscriptions are likely subsidized by expensive API usage.
  • Others counter that “profitability” is often defined by excluding training costs, stock‑based comp, or other major expenses, making the business structurally unprofitable so far.

Moat, Competition, and Commoditization

  • Strong skepticism about moat: models are close substitutes; switching costs are low; open‑weight and Chinese models (DeepSeek, Qwen, Kimi, etc.) are catching up quickly and are far cheaper.
  • Some think regulation and “AI safety” lobbying are aimed at regulatory capture to protect margins; others note that emerging regulation may instead slow data center build‑outs and hurt Anthropic.
  • Several argue that in a competitive market with many capable models, fat long‑term margins on tokens are unlikely.

Local and Open Models vs. Claude

  • Multiple engineers report moving much of their day‑to‑day work to strong local/open models (e.g., Qwen3.8, DeepSeek) and planning to cancel Claude subscriptions; others say these models still lag far behind frontier Claude/GPT for complex coding and long, agentic workflows.
  • Hardware constraints are key: good local performance generally requires 32–128GB+ of VRAM/unified memory, which is currently expensive but may normalize as hardware and efficiency improve.

Macro, Labor, and Hype Cycle

  • Some think 2026 wasn’t peak AI hype and that demand will keep rising; others compare this to tulip/railroad/dot‑com bubbles where the tech persists but many high‑flyers die.
  • There’s visible anxiety about AI‑driven layoffs and “Layoffpalooza 2027,” but disagreement on how broad the impact is today.
  • Several worry that if Anthropic or peers enter major indices, ordinary retirement savers will bear the risk of an AI valuation crash.