THE REVIEW
When the Machines Got Thirsty
A special edition examining the collision between artificial intelligence and physical reality
Dear reader, we present to you an unusual newspaper.
For decades, the technology industry has operated on a simple faith: that computing power would continue to grow exponentially, unbound by earthly concerns. The silicon would shrink, the models would expand, and the future would arrive on schedule. Then came 2025—the year the world’s most valuable technology companies discovered they had a physics problem.
This special edition of The Review examines what might be called the Great Reckoning: the moment when artificial intelligence, that most ethereal of technologies, collided with the stubbornly material world. The stories that follow trace this collision across six continents, from Japanese glass looms to orbital satellites.
You will read about a 589 billion from a single company’s value. About nuclear reactors being resurrected from the dead to power chatbots. About the Japanese cloth that has become more strategically important than any semiconductor.
These stories are connected by a single thread: the realization that software must ultimately run on hardware—and hardware is made of atoms. Welcome to the age of material computing.
The $6 Million Earthquake
How a Chinese startup trained a world-class AI model for the price of a modest Manhattan apartment—and terrified Wall Street.
A PDF posted to the internet cost Nvidia $589 billion in a single day.
On Jan. 27, 2025—exactly one year ago today—the Chinese artificial intelligence company DeepSeek released a technical report alongside its R1 reasoning model. The document claimed the final training run had cost approximately $5.6 million. American laboratories had spent hundreds of millions, sometimes billions, reaching similar capabilities.
Constraint became catalyst. DeepSeek developed techniques—multi-head latent attention, auxiliary-loss-free load balancing—that allowed their 671 billion parameter model to train with a fraction of the compute budget. They eliminated the expensive “Critic” model from their reinforcement learning pipeline. They open-sourced everything.
“We had no incentive to find the efficiency frontier when investors were providing billions to find the capability frontier instead,” a senior researcher at a major American AI lab admitted anonymously.
“The US export controls designed to hobble Chinese AI instead produced the AI equivalent of high-altitude training.”
For Further Reading: Perspectives
PRO: "DeepSeek's Latest Breakthrough Is Redefining the AI Race"
Source: CSIS Analysis (Dec 2025) Argues the efficiency gains signal the end of the “winner-takes-all” assumption in AI and open new paths for smaller nations.
CON: "Sabotage Fears Outpace Evidence"
Source: EU Institute for Security Studies (July 2025) Notes DeepSeek still relied on foundational U.S. research and Nvidia hardware, limiting claims of true independence.
The Nuclear Renaissance Nobody Expected
Big Tech discovers that chatbots need power plants—and the only ones available are the kind that split atoms.
The machines that will think for us must first be fed, and what they eat is power.
American data centers consumed 183 terawatt-hours of electricity in 2024. By 2030, this will likely double. In Virginia’s Loudoun County, server farms consume more electricity than all homes combined. The response? Nuclear power.
In late 2025, Microsoft announced a 20-year agreement to restart Three Mile Island Unit 1. Google signed agreements for 500 MW of small modular reactors. Amazon invested over $20 billion in nuclear projects. They need “baseload” power—constant output that solar and wind cannot provide without massive battery storage.
But a timing problem looms. AI demand doubles every year; nuclear reactors take a decade to license. Until 2030, the greenest tech companies may be keeping coal plants alive just to keep the lights on.
For Further Reading: Perspectives
PRO: "2026: The Year Nuclear Power Reclaims Relevance"
Source: Carbon Credits (Dec 2025) Analyzes how grid constraints and AI demand are reshaping nuclear’s role as essential low-carbon baseload.
CON: "Microsoft Wants to Resurrect Three Mile Island. It Will Never Happen."
Source: The Hill (Jan 2026) Former FERC chair argues regulatory, material, and logistical hurdles make nuclear restarts unrealistic.
The Trillion-Dollar Loom
The entire AI industry is waiting on output from a few specialized looms in Japan.
The most important material in artificial intelligence is not silicon. It is a specialized fiberglass cloth called T-Glass.
When a massive AI chip heats up, organic packaging materials warp. This snaps the microscopic wires. T-Glass prevents this, but nearly all of it comes from one Japanese company: Nitto Boseki. They are sold out through 2027.
The industry is now racing toward “Glass Substrates”—replacing the cloth with solid glass sheets. Intel, Samsung, and SK are investing billions. The nation that masters glass handling first will control the physical foundation of the next decade of compute.
For Further Reading: Perspectives
PRO: "Glass Substrates: The Breakthrough Material"
Source: Financial Content (Jan 2026) Analysis of glass as a critical enabler for the next decade of AI computing.
CON: "The Race To Glass Substrates"
Source: SemiEngineering (Aug 2025) Notes brittleness challenges, handling issues, and uncertainty about Intel’s internal commitment.
EDITORIAL: When Software Met Atoms
There is a famous line in technology: “Software is eating the world.”
The artificial intelligence era represents something different. Software has eaten enough of the world that it is now bumping into the world’s physical constraints. DeepSeek proved efficiency is mandatory. The nuclear rush proves reliable power beats theoretical capacity. The cooling crisis proves thermodynamics cannot be disrupted.
The era of software eating the world is giving way to something more complex: a dialectic between digital ambition and material constraint. The world is eating back.
Production Note: This edition was produced on Jan 27, 2026. All factual claims are sourced from reputable publications (Bloomberg, IEA, Reuters). Your skepticism remains appropriate and encouraged. Coming Next Week: The Talent War—examining the shortage of cooling engineers. © 2026 The Review. All rights reserved.