Sarmadi AI Digest June 19, 2026 Updated 6:55 AM CT Today Archive Topics Saved Subscribe RSS

Sanders unveils $7T AI wealth fund; Amazon to sell its AI chips against Nvidia

Bernie Sanders proposed a $7 trillion plan to put Americans 'in control of the AI industry' via a sovereign AI wealth fund — the first major US politician to put a number that big on AI public ownership. Amazon is in talks to sell its in-house AI chips directly to other data centers, Andy Jassy framing it as a multi-hundred-billion-dollar opportunity against Nvidia. FERC ordered grid operators to give AI data centers a fast lane for interconnections — formal federal preference for AI capacity over other load. Three Amazon software engineers are now under investigation by their employer for testifying at Seattle City Council in support of the data-center moratorium. OpenAI hired Transformer co-inventor Noam Shazeer back from Google ahead of its IPO, the same day enterprise-AI head Barret Zoph left for the second time. Inference startup Baseten is reportedly raising $1.5B at $13B; Snap spun off its AI video unit due to cost; Match found 47% of US singles view AI in dating negatively.

6 papers 21 news 6 sources ← Latest

News

22 items

Sanders' $7T plan and the AI political-economy showdown

Bernie Sanders proposed a $7T national AI wealth fund. A tech-worker-backed PAC (Guardrails) is countering Big Tech's lobbying with small-donor activism. Amazon is under fire for retaliating against engineers who testified for Seattle's data-center moratorium. The political-economy story has matured from headlines to specific dollar amounts and named conflicts.

News Ars Technica AI

Bernie Sanders unveils $7 trillion plan to give Americans control of AI industry

Sanders proposes a $7T plan that includes a sovereign AI wealth fund, framing public ownership as the answer to the AI-IPO concentration of wealth.

plan size $7T
Why it matters
  • First major US politician to put a trillion-dollar number on AI public ownership.
  • Lands as Pew shows two-thirds of Americans say AI is advancing too quickly — politically viable terrain.
  • Forces the lab and hyperscaler political teams to take public-ownership framing seriously.
  • Sets the upper anchor for any compromise legislation in 2027.
policyregulationmarket
Source →
News Wired AI

3 Amazon Workers Say They're Under Investigation for Speaking Out About Data Centers

Three Amazon engineers filed a Seattle civil-rights complaint after the company opened a disciplinary investigation against them for testifying at City Council on the data-center moratorium.

Why it matters
  • First well-documented internal retaliation by a hyperscaler against employees backing local DC limits.
  • Directly tests labor-law protections around climate and infrastructure speech.
  • Strongly compatible with the Sanders framing — employees vs employer over AI capacity.
policyregulationinfrastructure
Source →

Amazon takes Nvidia head-on; FERC gives AI a grid fast lane

Amazon is in talks to sell its in-house AI chips to other data centers — Andy Jassy framing it as a multi-hundred-billion-dollar opportunity. FERC ordered grid operators to fast-lane AI-DC interconnections. ASML denies the US suggestion that its top chip tool is now in China. Taiwan is boosting drone production for both its defense and the US military.

News TechCrunch AI

Amazon hopes to challenge Nvidia more directly by selling its AI chips

AWS is in talks to sell its in-house AI chips externally, with CEO Andy Jassy calling it a multi-hundred-billion-dollar opportunity against Nvidia.

Why it matters
  • First hyperscaler attempting to commercialize its in-house AI silicon at scale.
  • Directly attacks Nvidia's pricing power on the supply side, in tandem with the Cerebras/Groq/XCENA wave.
  • Pairs with the rumored Anthropic-Fable security-research role for Amazon — Amazon competing on chips, clouds, and safety simultaneously.
computemarketinfrastructure
Source →
News TechCrunch AI

AI data centers just got a government-mandated fast lane to the grid

FERC told grid operators to give AI data centers priority for interconnection — federal preference for AI load over other applicants.

Why it matters
  • Explicit federal preference for AI capacity over other commercial and residential interconnections.
  • Directly addresses the binding bottleneck behind the $130B in blocked DC projects YTD.
  • Major political risk: locks in a 'whose grid is this for?' fight at the state PUC level.
regulationinfrastructurepolicy
Source →

OpenAI bulks up for IPO; capital flowing into infra and world models

OpenAI hired back Transformer co-inventor Noam Shazeer from Google for IPO prep — the same day Barret Zoph left enterprise AI again. Baseten is reportedly raising $1.5B at $13B for inference infrastructure. General Intuition is in talks for $300M @ $2B for an embodied-AI world model trained on Medal's 2B-video dataset. Elastic agreed to buy DeductiveAI for up to $85M.

News TechCrunch AI

OpenAI is bringing on some big guns in the lead-up to its IPO

OpenAI poaches Transformer co-inventor Noam Shazeer back from Google as part of its IPO-prep talent build.

Why it matters
  • Shazeer is one of the most senior individual technical-talent hires possible — IPO-prep narrative signal.
  • Direct hit on Google DeepMind's senior-researcher bench at a politically sensitive moment.
  • Pairs with the leaked losses story — OpenAI buying credibility ahead of the S-1.
marketproducts
Source →

Anthropic Fable: White House making up the rules in real time

Anthropic still cannot distribute Mythos or Fable 5. Wired's reporting is now explicit: the White House is improvising regulatory standards live. The Verge's Decoder devoted a full episode to 'who decides when AI is too dangerous,' and a sharp paper, MosaicLeaks, asks whether your research agent can keep a secret in the first place.

News Wired AI

The White House Is Making Up Its Rules for AI in Real Time

Wired argues the White House is improvising AI-export rules live; Anthropic still can't distribute Mythos or Fable 5.

Why it matters
  • Explicit framing of post-EO process as ad-hoc — the predictability frontier labs were promised is gone.
  • Will encourage other labs to slow public capability disclosure further.
  • Strengthens the Sanders-framing of capture by political process, not market or science.
policyregulationproducts
Source →
News Hugging Face

MosaicLeaks: Can your research agent keep a secret?

ServiceNow research: agents tasked with multi-source research routinely leak secrets they were told to protect.

Why it matters
  • Names a concrete enterprise-deployment failure mode — relevant to every legal, financial, or HR research agent.
  • Pairs with the RedAct capability-trace work as a real procurement question.
agentssafetyevaluation
Source →

Consumer-AI sentiment hardens; Snap spins off, Adobe goes all-in

Match found 47% of US singles view AI in dating negatively. Snap spun off its AI video team into a new company, Dotmo, citing cost. Adobe shipped AI assistants across Photoshop, Premiere, and Illustrator, plus a redesigned Firefly studio. OpenAI added enterprise spend controls and a health-intelligence update.

Also today