Meta's $12B EU Fine and AI in Your Living Room

Here's a number worth sitting with: Meta could owe the European Union up to $12 billion — not for a data breach, not for a privacy scandal, but essentially for making its apps too good at keeping you glued to your phone.

The European Commission has issued a preliminary finding that Meta violated the EU's Digital Services Act through the design of Instagram and Facebook. The specific culprits? Infinite scroll, autoplay video, and personalized recommendation algorithms — the kind of features most of us interact with dozens of times a day without thinking twice. Regulators say these tools deliberately push users into a kind of digital autopilot, one that Meta never properly evaluated for its impact on mental and physical health.

That last part is important. The Commission isn't just saying the features are annoying or manipulative in a vague sense. It's saying Meta failed to conduct adequate risk assessments — especially when it comes to minors and vulnerable adults. That's a process failure as much as a product one, and it matters because it suggests Meta didn't just build something harmful, it didn't even do the homework to find out whether it might be.

The Commission also took a swipe at the tools Meta offers users who want to cut back on screen time. Time management features are too easy to dismiss. Parental controls require a level of technical effort most parents simply aren't going to put in. And mental health awareness prompts? Largely cosmetic, according to the report. The overall picture is of a company that built an off-ramp and made sure it was nearly impossible to use.

If the preliminary finding becomes a final decision, Meta could face a fine equal to 6 percent of its global annual revenue. With 2025 revenue sitting at just under $201 billion, that math lands around $12 billion. The Commission is also pushing for real product changes — think autoplay and infinite scroll turned off by default, meaningful screen-time breaks baked into the experience, and recommendation systems that optimize for something other than maximum engagement.

Meta will have the chance to respond before anything is finalized, so this isn't over. But the pressure is coming from multiple directions at once. Separately, the Commission is investigating Meta's age verification practices and content protections for younger users. And this August, Meta heads to trial in the US, where four states are pursuing penalties that, in a worst-case scenario, could reach into the trillions.

The timing is also notable. The EU is currently weighing a continent-wide ban on social media access for under-16s, with a report due next week. Europe is clearly in a mood to act, and Meta is standing squarely in its crosshairs. Whether or not the fine sticks at its maximum level, the era of treating addictive design as a purely internal product decision appears to be ending — at least on one side of the Atlantic.

More than 70 percent of Americans say they don't want a data center built near them. Sunrun thinks it has figured out the workaround: build the data center inside their houses instead.

The solar and home battery company is launching a pilot program it's calling distributed AI compute. The idea is straightforward in concept, genuinely strange in practice. Sunrun will install small computing nodes inside homes that already run its solar panels and battery storage systems. Those nodes will collectively function as a kind of decentralized data center, and Sunrun will sell the processing power to enterprise clients — AI companies, most likely — who need it. Homeowners who participate will get paid for the privilege of hosting.

It's a creative solution to a problem that's becoming harder to ignore. The AI industry's appetite for compute is enormous, and the traditional answer — build a massive data center somewhere with cheap land and power — is running into serious resistance. Communities across the country are pushing back against the noise, water consumption, and energy demands that come with industrial-scale facilities. Sunrun's pitch is essentially: what if we distributed all of that across a million suburban rooftops?

The company says it already ran a proof-of-concept test that went well, though it hasn't shared specifics on what that actually means in terms of performance or reliability. That vagueness is worth flagging. Running AI inference workloads across geographically scattered, consumer-grade hardware is a genuinely hard engineering problem. Latency, consistency, and coordination across nodes are all real challenges that don't disappear just because the business model is clever.

Sunrun has about 1.1 million customers, and those with existing solar and battery setups can now join the waitlist for the pilot. The battery piece matters here — these compute nodes would presumably draw on stored solar energy, which fits the sustainability angle Sunrun will almost certainly lean into when marketing this. The company says it plans to evaluate results before deciding whether to roll the program out more broadly.

The broader concept isn't entirely new. Distributed computing projects have existed for decades, from early scientific grid projects to crypto mining's more chaotic version of the idea. But pitching it as a turnkey product for homeowners, with AI companies as the paying customers on the other end, is a different kind of proposition.

Whether it works at scale is still an open question. But the underlying logic is sound enough that it deserves to be taken seriously. The AI industry needs compute. Homeowners want passive income. Communities don't want giant server farms next door. If Sunrun can actually thread that needle, it won't be the last company to try this model — it'll just be the first.