The Iran war is forcing Gulf states to confront a question that had long been deferred but rarely answered openly: what exactly are they receiving in return for anchoring so much of their financial, trade and security architecture to the United States?

The old bargain of American protection in exchange for dollar-priced oil, petrodollar recycling and strategic alignment now looks more fragile because the Gulf monarchies have taken much of the direct economic pain from retaliation while the credibility of U.S. protection has come under visible strain.

Most of the companies at the forefront of artificial intelligence are racing to build systems that don’t just autocomplete text, but emulate something closer to human-style thinking: breaking problems into steps, checking intermediate results, and revising answers. Those “reasoning” models are becoming the new standard for hard tasks in science, coding and maths. But fresh research shows that this leap in capability comes with a steep hidden cost: energy.

When researchers put these models under the microscope, they found that reasoning systems burned, on average, around one hundred times more electricity to answer a thousand text prompts than simpler alternatives that either lack explicit reasoning features or have them switched off. 

U.S. President Donald Trump is expected to sign a suite of executive orders as early as Friday aimed at fast-tracking the development of the domestic nuclear energy sector, leveraging emergency powers to overhaul regulatory frameworks, expand fuel production capacity, and accelerate reactor deployment.

The initiative—driven by surging electricity demand from the rapid expansion of artificial intelligence infrastructure—marks the most aggressive federal intervention in nuclear energy policy in decades. Trump, who declared a national energy emergency on his first day back in office, is framing the effort as a new energy-industrial mobilization, likening it to a modern “Manhattan Project.”