Cameron Otsuka

Soft Wars for Hard Infrastructure

In the 90s, Americans were anxiously seated in front of their TVs watching Operation Desert Storm unfold, leading to an oil shock, amidst “white-collar” recession fears due in part to computers. In 2026, Americans are anxiously scrolling on their phones watching Operation Epic Fury unfold, leading to an oil shock, amidst “white-collar” recession fears due in part to AI.

Just as computers before, it is easy to point to data centers as the physical manifestation of socioeconomic AI anxiety. This anxiety has led to a rash of headlines, from students booing their graduation speakers, moratoriums on building data centers, to political upheavals. Placed within the context of great power competition, data center anxiety becomes geopolitically useful.

A recent report from Bitcoin Policy Institute (BPI) argues this is the case:

This report demonstrates that there is a documented CCP-aligned US nonprofit ecosystem producing coordinated, transnationally distributed content opposing US AI infrastructure, AI labs, and AI export controls…

But to what end?

Chinese AI models, by regulation, must pass an ideological review and censor information that challenges the Communist Party — and the same models are being exported globally.

It is well-documented at this point that LLMs generate different amounts of political bias depending upon which language is being used. Bias is even “viewable” in model weights.

“What do such machines really do? They increase the number of things we can do without thinking. Things we do without thinking - there’s the real danger.” — Frank Herbert, Dune

If certain models with political biases become the global plurality, their outputs become the de facto “truth.” Just as the great power competition is overtly seen in the fight for geographical choke points and critical minerals, there is the covert game over local policy and public opinion that is often difficult to monitor and attribute.

Across the ocean, China’s populace doesn’t seem to have the same data center anxieties. Elementary schools are integrating AI and robotics into their classrooms:

Starting in the fall, every student in elementary and middle school in Beijing, and several other districts, began learning about AI. Third graders learn the basics. Fourth graders focus on data and coding. By fifth grade, students are learning about “intelligent agents” and algorithms.

And the pace of electrical grid build-outs are vastly outnumbering the US:

I am interested to know whether there are other think tanks with counter-research to BPI’s piece from the foreign perspective. If you know of anything, send it my way.

Mine Print Hash

Matt Dines and I go deep into warning signs from the SOFR futures market, how it impacts money markets and the underpinnings of leveraged finance, and round out with a discussion on the newly announced compute futures, which may unlock financing for the next wave of AI compute construction.

Open Threads

Choke points and resources:

• Bitcoin is money for enemies: Iran starts Bitcoin-backed ship insurance. Link
• Iran wants to bottleneck the internet. Link
  • And a related article about where the undersea cables emerge from the sea. Link
• UAE tries to bypass Hormuz with a new pipeline. Link
• Australia tells China to divest from their critical minerals. Link
• EU wants in on Pax Silica¹. Link

Sovereign debt is screaming:

• 30-year US Treasury yield highest level in 19 years: Link
• Japanese 10-year bond yields at multi-decade highs: Link
• UK 30-year gilt yield highest since 1998: Link