Pondero AI logo

Pondero AI

Archives
Log in
Subscribe
August 25, 2026

Pondero Brief: Hugging Face at $13B puts every open model under one owner

Pondero Brief - AUGUST 25TH, 2026

Plus Nvidia circles Perplexity at $30B, and 79% of enterprises blame governance.
pondero. BRIEF · AUG 25, 2026

Hugging Face at $13B: open models just got an owner

The Hub hosts 2 million models. A hyperscaler acquisition could restructure all of that.

Hugging Face has engaged a bank to field buyer interest above $13 billion, per TechCrunch, which confirmed the Business Insider report on Aug 24. That is nearly 3x its 2023 Series D of $4.5 billion, and 73% richer than Microsoft's $7.5 billion GitHub buy. Our take.

Also in today's brief

  • Hugging Face may end open model access as we know it
  • Nvidia is circling Perplexity at a $30B valuation
  • XPeng raised $900M to put humanoid robots in stores by year-end
  • Alibaba raised $10B and shipped a video model in the same week
  • 79% of enterprises say governance is blocking their AI rollout
  • GitHub Copilot killed its own lock-in argument this month
 
Money & Moves
Nvidia and Perplexity logos

Nvidia is circling Perplexity at $30B-plus.

Nvidia is in equity talks at a $30 billion-plus valuation, over 50% above the last round, per The Decoder citing The Information. Perplexity's annualized revenue tripled to $750 million-plus. This is the GPU-landlord playbook: an equity stake ahead of a large compute purchase. Our take.

 
XPeng IRON humanoid robot

XPeng raised $900M for its IRON humanoid.

XPeng's robotics arm closed a $900 million Series A at a $6.3 billion valuation, per PRNewswire. Alibaba and Tencent are both investors. IRON runs 3 in-house Turing chips at 2,250 TOPS on-device, with mass production targeted by end of 2026. Two rival giants backing the same physical-AI bet in one week is the real signal. Read it.

 
Alibaba share sale and Wan3.0 model

Alibaba raised $10.2B and shipped Wan3.0 the same week.

Alibaba priced a record $10.2 billion Hong Kong follow-on, 3x oversubscribed, per Bloomberg, with 100% of proceeds going to its AI full-stack push. Days later it shipped Wan3.0, a video model doing 30-second 1080p clips that accepts PDFs and slides as creative inputs. The raise and the release are one strategy, not two headlines. Our take.

 
Policy & Legal
Enterprise AI governance survey

79% of enterprises say governance is the top barrier to scaling AI.

In a Google Cloud survey of 1,402 IT leaders, 79% named security, governance, and MLOps the number-one blocker, ahead of compute cost. 83% say their current infrastructure needs an upgrade for agentic AI. The root problem Google names is "agent sprawl," and the fix is scoped permissions plus human approval gates. Read it.

 
Quick Hits
• Make. If the governance survey hit home, this is where you build the approval workflow: human-in-the-loop gates and audit trails for the AI agents you are already running. Try Make →
• Buttondown. The newsletter platform that sends this email. Clean API, real referral program, no AMP nonsense. Good if you are building an audience of your own. Try Buttondown →
 
From the Pondero Stack
GitHub Copilot Agent Plugins 1.0 review

GitHub Copilot Agent Plugins 1.0 is GA.

GitHub Copilot shipped Agent Plugins 1.0 GA, per GitHub's changelog, and the ecosystem lock-in objection is gone. One plugin.json now runs across VS Code, Cursor's client, and Copilot CLI, so the "redundant second seat" case against it for Cursor-first teams no longer holds. Pricing held flat at $19 per user for Business. Cursor still edits cleaner on whole-repo multi-file refactors, which is why we landed at 4.5, not higher. Read the review. Try Copilot.

 

How was today's brief?

★★★★★ Nailed it  |  ★★★ Solid  |  ★ Missed

Jonathan Hildebrandt Jonathan Hildebrandt
Co-founder and primary operator of Pondero. Writes the Pondero Brief.

Affiliate disclosure  ·  Unsubscribe  ·  Manage preferences

Pondero earns commissions on some links. This does not affect our editorial picks.

Don't miss what's next. Subscribe to Pondero AI:
← Newer Pondero Brief: NVIDIA Groq 3 LPX in production at 3,400 tokens per second Older → Pondero Brief: An anonymous model outscored Claude Fable 5, and forensics named the builder
pondero.ai
Bluesky
LinkedIn
Twitter
LinkedIn
Powered by Buttondown, the easiest way to start and grow your newsletter.