Nonrival

Nonrival

Archives
Log in
Subscribe
November 10, 2026

Data centers, Anthropic's IPO risk, and who's winning AI

Nonrival

Expert analysis from think tanks and academia, on public policy, economics, and technology.
Human experts. AI summaries.


Voters Turned Against Data Centers in 2026, and Democrats Rode the Backlash to a Historic Midterm Win

the_conversation

  • In the 2026 midterms, 40 out of 41 local ballot measures restricted data center development, and exit polls showed 56% of voters called data centers a major factor in their congressional vote — suggesting the AI infrastructure boom has become a genuine political liability.
  • Opposition is unusually bipartisan: majorities of both Democrats (69%) and Republicans (54%) say they don't want data centers in their communities, united by concerns over electricity costs, water use, noise, and land consumption.
  • The author cautions that the 'data centers cost Democrats' narrative may be overstated — we don't yet know where losing candidates stood — but argues the narrative itself will shape politics regardless, potentially making data center skepticism a litmus test in the 2028 primaries.

Paraguay's 'Dubai of South America' Myth Collapses Along With Its Biggest Bank

the_conversation

  • Paraguay's largest digital bank, Ueno, failed in November 2026 after regulators discovered massive undercollateralized loans to its controlling shareholder — taking down President Santiago Peña, who had deep financial ties to the bank, within days.
  • The crisis revealed how thoroughly Paraguay's financial system had become entangled with political elites: the central bank chief, the president, and the social security system (which had parked $355 million in Ueno) were all connected through a web of revolving-door appointments and shared business interests.
  • Paraguay's strong growth figures and new investment-grade credit ratings masked the same weaknesses — regulatory capture, connected lending, weak supervision — that triggered banking collapses in the 1990s and 2000s, raising doubts about whether the country's economic model is as solid as advertised.

The US Is Winning the AI Race, But China May Be Running a Different Race Entirely

cset

  • The US leads China in semiconductors, frontier AI models, and computing power, but measuring the competition solely by large language model benchmarks may miss how China actually defines and pursues artificial general intelligence.
  • China places greater emphasis on 'embodied' AI — intelligence grounded in physical interaction with the world — along with brain-inspired computing and robotics, giving it potential advantages if those paths prove more productive than scaling today's models.
  • A forthcoming CSET report argues US strategy needs to be robust to multiple possible AI futures, not just the one where frontier language models lead to AGI.

In the News

Anthropic's IPO Could Be the AI Industry's Defining Moment — and Nobody's Ready for What Happens If It Goes Wrong

roosevelt

  • Anthropic is heading toward a $1.5 trillion IPO, but policy experts warn its own S-1 filing admits its products could cause 'catastrophic or existential risks' — while there's still no federal law governing AI and the Trump administration favors self-regulation.
  • The IPO's circular financing structure — with Amazon and Google as investors, customers, and compute suppliers simultaneously — raises fears of a 2008-style financial crisis, with pension funds and retirement accounts exposed to the fallout.
  • Experts argue that without structural reforms like employee board seats, Glass-Steagall-style separation of AI and cloud giants, and mandatory safety regulation, the wealth from AI will flow almost entirely to founders and investors while workers bear the risks.

Analysis

Two Economists Find That Simply Telling Unemployed Workers About Benefits Nearly Doubles Enrollment

upjohn

  • Eligible unemployed workers often don't claim unemployment insurance due to administrative barriers and lack of information — not lack of interest.
  • A randomized experiment with 50,000 workers in Washington State found that sending informational letters nearly doubled application rates, especially among low-wage workers.
  • The research also finds that expanding UI eligibility generates $2.57 in social value for every $1 of net government spending, suggesting the program is significantly underutilized.

Also Worth a Look

  • The U.S. Military Is Plugging Holes in Drone Defense That Law Enforcement Can't Fill — and That's a Problem (csis)
  • South Korea Wants to Cap Food Delivery App Fees to Help Restaurants — but the Evidence Suggests It Could Backfire (itif)
  • The Supreme Court's 'History and Tradition' Method Has a Problem: History Contains Contradictory Traditions (scotusblog)
  • Democrats Just Had a Historic Midterm Wave — but 2028 Is Anyone's Guess (brookings)
  • Serbia's new reformist government risks letting its best chance at energy independence slip away (atlantic_council)
  • Tariffs, a War, and the AI Boom Are All Fueling Inflation — and the New Congress Has Few Easy Fixes (uva_darden)
  • The U.S. Is Getting Tougher on Export Control Enforcement While Making It Harder for Companies to Know What the Rules Are (csis)
Don't miss what's next. Subscribe to Nonrival:
← Newer Ozempic's blind spot, missile shortfalls, and rogue AI swarms Older → AI exposure data, polling misses, and office job globalization
Powered by Buttondown, the easiest way to start and grow your newsletter.