TL;DR: AI data centers often make headlines in news reports about soaring electricity bills. But the reality? Well, it’s complicated. Volatile natural gas prices and neglected grid infrastructure upgrades are among the real culprits, but that doesn’t mean that data centers won’t ultimately contribute toward your eye-wateringly high electricity bill. It could all come down to how data center operators and power providers manage the US’ aging electrical grid.

What happened: Data centers catch flak for elevating electricity rates for consumers, but current rising power prices have more to do with surging natural gas prices, BloombergNEF’s head of US power, Helen Kou, tells Tech Brew. We’re relying more on gas for power due to the retiring of US coal plants, which makes gas markets more turbulent.

Another big factor is grid upgrades: If utilities shell out cash to modernize grid infrastructure, those costs are usually passed on to ratepayers, Kou says. But some utility companies are pushing for data centers to absorb those costs to insulate consumers. If this happens, data centers might even bring energy prices down in the short term.

In the future, though, Kou says the proliferation of data centers could increase demand, and therefore energy prices, unless more investments are made in power generation.

Blame game: So why are data centers often accused of raising rates? Jason Ferrara, the managing partner at data center advisory firm Callosum, tells Tech Brew that they’re an “easy target” and a “misunderstood topic.”

“They’re massive facilities,” he says. “People that live nearby these things see them and wonder how they’re impacting their daily lives. It’s easy to point to; it’s a ginormous box. It’s not like it’s an underground facility that you don’t see.”

Ferrara also said that the Trump administration’s ratepayer protection pledge—which calls on large tech companies like OpenAI and Microsoft to pony up for any extra power their data centers need—shows that server farms “are not out there to harm your communities.” The pledge, which tech companies signed earlier this month, is nonbinding, though, which has been a cause of concern for legislators and energy experts.

Bottom line: Data centers might be making headlines as the reason for ballooning electricity bills, but pricey grid upgrades and rising natural gas prices are also pushing costs higher—and that pressure may not ease anytime soon. —TC

Airplane mode, but for AI

Tech companies keep racing to add AI to every product imaginable (even ones no one asked for, like this refrigerator with no handle that we told you about a few weeks ago). And frustratingly, not every tool lets you disable its AI features. But for Tech Brew reader Michael from Idaho, the dream isn’t more AI—it’s less:

Why can't I just have the option to turn off AI? And I mean AI for all apps and services across the board. It is constantly getting in my way.

Doesn’t seem like too much to ask—just a universal “no thanks, I’ll do it myself” button. Call it airplane mode for artificial intelligence. Because sometimes the smartest feature is letting humans opt out. —CM

Silicon Valley tends to go through buzzwords like a college freshman goes through majors—whether it's “Web3,” “blockchain,” or the “metaverse.” Now, according to a New Yorker piece, the word du jour is “taste”—the thing that might separate the true visionaries from the vibe coders in a world where AI is automating new tasks every day. The New Yorker rounds up examples of tech figures who’ve ruminated on the subject recently: Venture capitalist Marc Andreessen claimed that, in the AI era, the art of picking good investments might be “one of the last remaining fields” done by people. Y Combinator co-founder Paul Graham thinks taste will only become “more important” in the future. A former ByteDance engineer even called it the human “moat”—in VC speak, the thing competitors can’t copy.

The New Yorker argues Silicon Valley could also be obsessed with taste because AI doesn’t exactly have a perception for being “cool” and tasteful—even as people use it to discern between countless everyday choices, like what tone to strike in an email, what kind of wine to bring to a housewarming, which project to prioritize at work, and what neighborhood to look for apartments in (not to mention all the AI-driven recommendations features included in popular apps today).

AI companies’ pitch appears to be that AI can and should help you taste-maxx—but to some, that’s like hiring a focus group to pick your favorite movie. LLMs produce, by design, a weighted average of their training data—a USC study published this month even found evidence that AI might be homogenizing human writing, reasoning, and perspective. Taste, meanwhile, is by definition not a consensus, but one person’s judgment. In fact, it might be the one thing you can’t squeeze out of a machine that’s been fed the sum of all human knowledge. —WK

Chaos Brewing Meter: /5