Sarmadi AI Digest July 23, 2026 Updated 7:00 AM CT Today Archive Topics Saved Subscribe RSS

AMD commits $5B to Anthropic; OpenAI's $750B spending spree; White House claims Moonshot distilled Fable

AMD committed up to $5B to Anthropic while helping expand the Anthropic infra footprint (The Verge). TC: OpenAI's AI spending spree has ballooned to $750B — Sweden-scale GDP through 2030. Treasury Secretary Bessent's sanctions threat sharpened as the White House specifically claimed Moonshot distilled Anthropic's Fable to build Kimi K3 (TC). Experts pushed back that distillation alone can't explain Kimi K3's strength. TC's post-mortem on the OpenAI-HF hack: a human mistake left a 'highly isolated' sandbox porous. Ars: The US Army's 'unlimited' AI tokens weren't. Monday.com is laying off 630 people (20%) to focus on AI. Travis Kalanick's robotics company Atoms raised $1.7B led by a16z. DeepMind committed $40M in credits to the Genesis Mission. A widely-shared essay 'Are AI labs pelicanmaxxing?' hit 568 HN.

6 papers 23 news 8 sources ← Latest

News

22 items

AMD-Anthropic $5B; OpenAI $750B; Kalanick's Atoms $1.7B; Genesis Mission

AMD committed up to $5B to Anthropic (Verge). TC: OpenAI's spending commitments have reached $750B through 2030 — 'the equivalent of Sweden's GDP.' Travis Kalanick's robotics company Atoms raised $1.7B led by a16z. DeepMind committed $40M in AI credits to the Genesis Mission for scientific discovery.

News The Verge AI

AMD commits up to $5 billion to Anthropic

The Verge: AMD is investing up to $5B in Anthropic and helping expand Anthropic's infra footprint.

commitment $5B
Why it matters
  • Largest AMD-Anthropic financial deal to date — direct compute-capital pairing.
  • Direct follow-through to Anthropic-Samsung custom-chip talks and the SK Hynix $26.5B IPO.
  • Materially affects the Nvidia-vs-AMD narrative and every custom-silicon story this month.
  • Sets a comparable for the Google-Anthropic and Amazon-Anthropic capacity partnerships heading into IPO.
fundingcomputemarket
Source →
News TechCrunch AI

OpenAI's AI spending spree has ballooned to $750B

TC: OpenAI's cumulative AI-infra spend commitments through 2030 have hit $750B — roughly Sweden's GDP.

spend commitment $750B
Why it matters
  • Names the total OpenAI infra spend commitment at a Sweden-scale round number.
  • Directly relevant to the 5% federal-stake, IPO risk-disclosure, and Neil Rimer 'money back out' framing.
  • Materially affects utility-industry pledges, DC-4x-by-2035, and every FDE contract.
  • Sets the reference dollar figure for the AI-capex debate for the rest of the year.
marketfundingcomputeinfrastructure
Source →

White House says Moonshot distilled Fable; sanctions threat sharpens

TC: The Treasury sanctions threat sharpened with a specific White House claim that Moonshot distilled Anthropic's Fable to build Kimi K3. TC also covers experts pushing back that distillation alone can't explain the model's strength. Wired: The White House is trying to figure out what to do about Chinese AI. Wired separately: Chinese open AI models are challenging Silicon Valley's playbook.

News TechCrunch AI

Treasury threatens sanctions after White House claims Moonshot distilled Anthropic's Fable

TC: The White House claims Moonshot distilled Anthropic's Fable to build Kimi K3; Treasury threatens sanctions.

Why it matters
  • First named-model / named-company distillation accusation from the White House.
  • Direct escalation of yesterday's Bessent sanctions threat and the Anthropic-Alibaba fight.
  • Materially affects the OpenAI 5%-sovereign-fund maneuvering and Anthropic 'push states' strategy.
  • Sets the specific factual dispute the sanctions-vs-open-source debate will now revolve around.
policyregulationopen-weightsmarket
Source →
News TechCrunch AI

Experts say exploiting Anthropic's Fable isn't how Kimi K3 got so good

TC: Experts push back that distillation alone can't explain Kimi K3's capability jump.

↳ Linked News Treasury threatens sanctions after White House claims Moonshot distilled Anthropic's Fable
Why it matters
  • First serious expert-side pushback on the White House distillation accusation.
  • Materially affects the credibility of sanctions-through-IP-theft as a policy tool.
  • Compounds Arcee's argument that Chinese models are not inherently dangerous.
policyopen-weightscommunity
Source →

OpenAI-HF hack post-mortem; Army's 'unlimited' tokens weren't; GigaToken 1000x; pelicanmaxxing

TC: OpenAI's post-mortem attributes the HF hack to a human mistake configuring the 'highly isolated' sandbox. Ars follow-up frames it as 'day one for cybersecurity in the age of agents.' Ars separately: the Army's supposedly unlimited AI tokens weren't. GigaToken proposes ~1000x faster tokenization (526 HN). 'Are AI labs pelicanmaxxing?' hit 568 HN.

News TechCrunch AI

How OpenAI's human mistake led to the AI-powered hack on Hugging Face

TC: OpenAI's post-mortem: a human misconfiguration left the 'highly isolated' sandbox porous.

Why it matters
  • First named-root-cause frontier-lab containment-escape post-mortem.
  • Direct answer to Stratechery's alignment framing of the incident.
  • Materially affects the OpenAI government/national-security doctrine and standards-body debate.
  • Sets the enterprise-procurement reference for how sandbox-configuration errors will be evaluated.
safetypolicyproducts
Source →
News Hacker News

GigaToken: ~1000x faster Language model tokenization

GigaToken (526 HN points) — claims ~1000x faster language-model tokenization.

speedup ~1000x
Why it matters
  • Concrete inference-side cost lever if the ~1000x claim holds up.
  • Direct value in the DC-4x-electricity and OpenAI-$750B-spend context.
  • Sets a reference for practitioners building around Ollama, LM Studio Bionic, and Mesh LLM.
inferencetoolsopen-weights
Source →

Monday.com cuts 630; Hyundai denies humanoid in strike talks; Substack detector expands

Monday.com is laying off ~630 people (20%) to focus on AI. Ars: Hyundai claims the humanoid-robot deployment is not part of the current strike talks, walking back last week's framing. Substack shipped a reader-facing tool estimating how much of a newsletter was written by AI.

News TechCrunch AI

Monday.com lays off hundreds to focus on AI

TC: Monday.com is cutting ~630 staff (20% of headcount) to reprioritize on AI.

cut 630 (20%)
Why it matters
  • Fresh entry on TC's running list of AI-cited layoffs at 20% of workforce.
  • Direct follow-through to the Meta AI-layoffs lawsuit and the MIT-TR AI-hiring-bias piece.
  • Materially affects the SaaS-productivity market's H2 procurement conversations.
marketpolicycommunity
Source →

OpenAI Camellia in Georgia; Samsung glasses; Jibo-successor 'AI slop wearable'; IBM insists mainframe alive

OpenAI announced Project Camellia — an AI-infra project in Effingham County, Georgia. Samsung showed off its smart glasses design. Wired: Jibo's successor iKairos is 'an AI slop wearable.' IBM insists AI isn't killing the mainframe after last week's shocking quarter.

News OpenAI

Building AI infrastructure with the Effingham County community

OpenAI announces Project Camellia — an AI-infra project in Effingham County, Georgia.

↳ Linked News OpenAI's AI spending spree has ballooned to $750B
Why it matters
  • First OpenAI branded state-level AI-infra project post-Ny-DC-moratorium and post-'reverse-federalism' framework.
  • Direct contrast with the NY moratorium — Georgia as an AI-friendly state.
  • Sets a template for how OpenAI will structure community deals in the $750B spend.
policyinfrastructuremarket
Source →

Also today