A McDonald’s AI Christmas ad sparked outrage, becoming a flashpoint for a broader backlash. With billions being spent and serious ethical questions unanswered, is the AI revolution in trouble before it even really starts?
The AI chip market is on a tear, with combined sales from major players topping $400 billion in 2025. Next year is projected to be even bigger, but the blistering growth is creating serious bottlenecks and raising questions about sustainability.
Researchers have unveiled TurboDiffusion, a new AI video generation method that’s up to 200 times faster than current tech. It can create a 5-second HD clip in just 24 seconds, a task that used to take nearly 80 minutes. This leap brings real-time AI video generation much closer to reality.
In a major strategic shift, Nvidia has reportedly scooped up nearly all of AI chip startup Groq’s workforce. This move targets the critical AI inference market, where Nvidia seeks to extend its dominance beyond just training models.
OpenAI is offering over half a million dollars for a new “head of preparedness” role focused on mitigating AI dangers. The job comes as companies cite AI-related reputational risks 46% more often than last year. CEO Sam Altman says it will be a “stressful” job diving into the deep end immediately.
To keep AI systems running, data center developers are bypassing the overloaded electric grid. They’re installing power sources like aeroderivative turbines from old 747s and massive diesel generators, creating a major environmental and economic trade-off.
In 2025, AI tools like Google’s Gemini and NotebookLM became essential for research and coding. But features bolted onto existing apps like Gmail and Word feel more like a tax than a benefit. The good stuff is getting really good, but the overpromising hasn’t stopped.
The author, who pioneered machine teaching at Microsoft, says enterprises are stuck in “pilot purgatory” because AI agents aren’t given room to practice. True expertise, like on a basketball team, comes from repetition and defined roles in realistic environments.
The race for AI compute is breaking all the old rules. With Moore’s Law fading, companies like Cerebras are building radically different hardware, like a 30cm chip with 90,000 cores, to power the next generation.
As 2025 ends, business leaders are planning for a 2026 shaped by AI’s continued boom, shrinking DEI initiatives, and stricter return-to-office rules. The article argues against an AI bubble, emphasizing a seismic shift in human-machine interaction. Success will hinge on insulating against turbulence
