February 24, 2026
Snacks vs Servers: The Energy Fight
Atlantic: Sam Altman Is Losing His Grip on Humanity
Altman compares brains to bots; commenters call spin, bias, and a god complex
TLDR: At an India AI summit, OpenAI’s Sam Altman said “training humans” takes huge energy and claimed chatbots may already be more efficient per answer. Commenters split between calling it PR spin amid climate worries, mocking the “AI vs humans” framing, and accusing the article of lazy anti-Altman bias
Sam Altman told an AI summit that “it takes a lot of energy to train a human,” arguing chatbots might already be more efficient per answer. The internet detonated. The loudest chorus: Why are we even comparing people to machines? One top voice asked if we should spend scarce energy raising problem‑solving humans—or building tools that make humans redundant. Others saw the quote as a tell: casually putting people and bots on equal footing gives off big techno‑god vibes.
Climate‑worried readers torched the analogy as PR smoke while data centers crank up new gas plants. “OpenAI is burning cash and fuel,” one critic fumed, calling the line a distraction from emissions and those massive “Stargate” server farms. Meme makers had a day with it: “20 years of snacks vs 20 megawatts,” “brain = banana, chatbot = power plant,” and “my mom trained me with chicken nuggets.” Chaos, but funny.
Not everyone loved The Atlantic’s framing, either. Skeptics yawned at what felt like a “Sam Altman bad” rerun, and one commenter dropped a cheeky ownership pic as a wink at media ties link. Meanwhile, folks side‑eyed Anthropic for letting its bot “feel distress,” calling it creepy marketing. The split: race to nuclear/renewables fast—or hit pause on AI’s power binge
Key Points
- •Sam Altman, at an AI summit in India, compared per-question energy use of trained AI models to that of humans, claiming AI may have caught up in energy efficiency.
- •The Atlantic asserts the brain’s energy use is far lower than AI models for simple queries and emphasizes AI’s contribution to climate change as the core concern.
- •OpenAI’s Stargate data centers are reportedly setting up combustion turbines; other data centers are building gas-fired power plants or extending coal plants, with emissions potentially comparable to dozens of major US cities.
- •Altman stated superintelligence could be a few years away and urged rapid shifts to nuclear, wind, and solar due to growing AI use; the article suggests waiting is another option.
- •Anthropic’s CEO Dario Amodei made similar human-AI training analogies, and Anthropic is exploring Claude’s potential consciousness and allows it to end abusive conversations citing “model welfare.”