← AM·PM Brief

Jay Goldberg: Anthropic's Compute Build Could Push It Back to Losses | AI Daily Brief

· Evening brief · 9 news · 7:56

Audio in Mandarin Chinese · English transcript below

⚡ Jay Goldberg won't rate Anthropic: no stock price, key blanks in IPO filing. Estimates ~$800B infrastructure commitments by 2033; Anthropic warns profit turns to loss as compute shifts from cloud leases to owned infra after just 1-2 quarters.

Anthropic's path to profitability may prove to be merely a fleeting respite rather than a sustainable breakthrough, as the company now faces its most consequential test: a dramatic strategic pivot from cloud leasing toward unprecedented capital-intensive self-built infrastructure. According to Jay Goldberg, the firm's infrastructure commitments through 2033 substantially exceed what has been publicly disclosed, while critical gaps in its prospectus render long-term profitability essentially unassessable. Taken together, these developments illuminate a troubling capital paradox now ensnaring leading LLM developers collectively: the more advanced their models become, the heavier the infrastructure burden they must shoulder, and the more elusive sustainable profitability grows.

Jay Goldberg: Anthropic's Compute Build Could Push It Back to Losses | AI Daily Brief

Today's Top 3 Headlines

  1. Opinion

    Jay Goldberg: Anthropic may return to losses after building own compute

    Seaport's Jay Goldberg won't rate Anthropic: no stock price yet, key blanks remain in the prospectus. He estimates ~$800B infrastructure commitments by 2033; Anthropic warns profits will flip back to losses within 1-2 quarters as it shifts from cloud rental to owned compute.

    Source ↗
  2. Related News

    📰 Anthropic IPO to Adopt Founder-Control Structure: WSJ

    WSJ examines the "imperial CEO" phenomenon using Anthropic: like other hot tech firms going public, it adopted a share structure letting founders retain control, spotlighting the link between founder control and IPO arrangements.

    Source ↗
  3. Opinion

    All-In: Cybersecurity fears cap enterprise AI DAU at 10-20%

    All-In cites Gallup: only 10-20% of workers regularly use AI at work. Big firms block access over cybersecurity fears, banning email and knowledge base integration. Daily active usage could jump to 60-80% once risks are mitigated; Uber has already bought ChatGPT for all employees.

    Source ↗

+6 more headlines

  • Johannes Schaback: Design for manufacturability early, don't wait for prototype completion
  • Robert Caro: Every Era Said Impossible, Moses Still Built the Roads
  • David Tisch: Macro calls won't save early-stage investing; wins & losses are in specific companies
  • Ethan Mollick: Max Thinking is Codex's top tier, still trails Pro
  • Zoubin Ghahramani: One Prompt, Gemini Agent Builds Robotic Pumpkin
  • Vinod Khosla: Recruiting top talent isn't a numbers game, you need reciprocity first
Unlock all 9 headlines + deep analysis →Free 7-day trial · cancel anytime
Browse all past briefings →