AI news · Saturday, August 8, 2026
Amazon plans a massive gas power plant for its Texas data center
Amazon is facing backlash over a new data center in Pecos County, Texas, that will be powered by an on-site natural gas plant permitted to emit 33 million tons of carbon dioxide annually. To put that in perspective, the plant would emit more greenhouse gases than any other power facility in the United States, effectively undermining the company’s pledge to reach carbon neutrality by 2040. While Amazon claims the site won't increase electricity costs for local families, the decision highlights how the massive energy appetite of AI is forcing tech giants to bypass the grid and build their own high-emission infrastructure. This tension between corporate sustainability goals and the reality of AI’s hardware requirements is becoming a recurring theme as companies like Meta and Google follow similar paths.
In the talent and tools space, OpenAI has quietly absorbed the presentation startup NextSlide, which specialized in turning documents into polished slide decks. The move is part of an ongoing trend of labs snapping up small teams to improve how their models present information and handle complex work flows. Meanwhile, workers are increasingly preparing for the career instability caused by this rapid automation; some are now saving up dedicated 'burnout funds' to cover long breaks as the pace of work accelerates. McKinsey partners are even warning junior staff against producing what they call 'AI slop'—content that looks professional on the surface but fails under any real scrutiny.
In the competition for image generation, xAI’s new Imagine 2.0 model has landed in second place globally in Arena benchmarks, trailing only OpenAI’s GPT-Image-2. The model is being marketed with a suite of professional editing tools like background removal and 'Magic Wand' adjustments, signaling that the focus is shifting from simple image generation to complex, multi-step production workflows. As these models become more capable, enterprise leaders are increasingly hesitant about which systems to trust, weighing the high costs and potential safety guardrails of frontier labs against the security risks of open-source models. This uncertainty has created a lucrative market for evaluators like Arena, which is now selling data on how models perform when pitted against each other in 'battle royale' tests. For those feeling overwhelmed by the constant presence of these systems in their daily software, there are increasingly manual ways to opt out, such as disabling Google’s smart features or using browser extensions to scrub AI bars from productivity apps.
The quick hits
- Amazon is building a private, gas-burning power plant in Texas that would become the largest source of climate pollution in the U.S. — highlighting the environmental price tag of AI.
- xAI's new image model, Imagine 2.0, is now ranked second globally behind OpenAI’s leading model — proving the gap between frontier AI labs is narrowing fast.
- McKinsey partners are warning junior consultants to stop producing 'AI slop' — a blunt reminder that generic, AI-generated work is becoming easy to spot and dismiss.
- Workers are starting to save 'burnout funds' to afford career breaks as workplace stress increases — showing the human cost of current AI-driven productivity demands.