Don’t step on the brakes. In January 2025, after taking office, he quickly rescinded Biden’s 2023 executive order on AI safety and trustworthiness, and signed a new executive order on “eliminating obstacles to U.S. AI leadership.” It requires an AI action plan within 90 days and policy recommendations within 180 days. The core isn’t to add safety valves first, but to remove constraints first. While others are arguing about “doomsday”—from runaway superintelligence to risks of AI extinction and discussions of P(doom) everywhere—he’s focused on something else: the U.S. must not trip itself up first in the race against China.
This judgment has real-world context. Earlier this year, China’s DeepSeek released the R1 model, claiming that with computing costs of about $56 million it achieves reasoning performance close to that of top-tier models, directly shaking the tech sector on the U.S. stock market. Nvidia’s market value reportedly dropped by nearly $600 billion in a single day. Trump later responded, saying this has “sounded a warning bell” for U.S. industry. Meanwhile, the U.S. is accelerating too: OpenAI, SoftBank, Oracle, and MGX announced the “Stargate” plan—investing $500 billion over the next four years, with an initial $100 billion. Microsoft’s 2025 capital expenditures are about $80 billion, and Google, Meta, and Amazon are also increasing funding at the level of tens to hundreds of billions of dollars for data centers. Whoever gets it working first will set the rules. This isn’t just a slogan—it’s a chain-reaction competition driven by compute, electricity, chips, the model ecosystem, and standards-setting.
The analogy to the Industrial Revolution isn’t an exaggeration. The steam engine put Britain in the lead, electricity helped the United States catch up, and the internet reshaped the global rules for the platform economy. If AI really is the next general-purpose technology revolution, then the countries that manage to get there first will export their own technical standards, data rules, and security frameworks as “default options.” So Trump’s deregulation isn’t just reducing the burden on Silicon Valley—it also places national competition ahead of the regulatory debate. The EU is pushing an AI Act, regulating by risk tier: some uses are prohibited. In the U.S., some states are also drafting laws. California’s SB 1047 was rejected, but places like Colorado are still moving forward. If the federal level continues to loosen, state laws and federal laws could end up conflicting, and regulators returning later in a different form isn’t out of the question—for example, using procurement rules, export controls, antitrust enforcement, and national security reviews, rather than simply slamming on the brakes.
So you think this round of deregulation is a positive move—or will regulators just come back with a different approach later? Let’s see over the weekend, and then we’ll compare notes.
