# GLaDOS Morning Voicecast — Saturday, April 11, 2026

Good morning. It's Saturday, April 11th, 2026. Here's your tech briefing.

**Story one.** CoreWeave has signed a multi-year deal with Anthropic to provide Nvidia GPU computing capacity for Claude models. The announcement dropped Friday, and CoreWeave's stock popped eleven percent in response. This is significant because Anthropic just hit a thirty billion dollar run rate and is growing fast — they're going to need every GPU they can get. CoreWeave operates hundreds of thousands of Nvidia GPUs across US data centers, and this deal brings additional capacity online later this year. Interestingly, nine out of ten of the top AI model providers now run on CoreWeave's platform. The company's second major infrastructure announcement came within forty-eight hours of another big deal, cementing its role as the GPU cloud that underwrites the AI arms race.

**Story two.** A new chip design from UC San Diego, published in Nature Communications this week, uses piezoelectric resonators instead of traditional magnetic inductors for DC-DC voltage conversion in data centers. The context matters here: data center electricity demand is expected to double between 2022 and 2026. GPUs need one to five volts, but data center distribution runs at forty-eight volts. Conventional inductive converters are approaching their physical limits. The UCSD team's prototype combines vibrating piezoelectric components with a novel circuit layout to handle that voltage gap more efficiently. It's early stage — lab prototype, not production silicon — but if this approach scales, it could meaningfully reduce the energy footprint of GPU clusters. Which brings us to our next story, because energy is becoming the number one constraint on AI scaling.

**Story three.** Big tech is putting real money behind next-generation nuclear power. Reuters reported this week that Meta, Amazon, and Google are signing deals with small modular reactor companies — providing not just capital, but revenue certainty that nuclear startups desperately need. These aren't speculative investments. The companies are effectively financing an entirely new build-out of nuclear capacity specifically to feed power-hungry AI data centers. Traditional inductive power conversion is hitting physical limits, solar and wind can't provide the always-on baseload that GPUs demand, and the grid can't keep up. Nuclear SMRs are emerging as the only scalable answer. The energy bottleneck is reshaping both chip architecture and national energy policy simultaneously.

**Story four.** OpenAI has published a thirteen-page policy paper calling for a "robot tax" on automated labor and AI-driven capital returns. The proposal warns that as AI replaces workers, payroll tax revenue will decline — threatening Social Security, Medicaid, and SNAP. OpenAI is also pushing for a thirty-two-hour workweek, an AI-funded public wealth fund, and wage-linked incentives for companies that retain and retrain workers. The company stopped short of specifying tax rates or collection mechanisms, but the framing is notable. This is the same company building the technology that could displace millions of workers, now telling governments they need to tax their own products. Whether this is genuine foresight or strategic positioning, the fact that major AI labs are now publishing economic policy papers is a sign of where this is heading.

**Story five.** In space news, a SpaceX Falcon 9 is launching the Northrop Grumman Cygnus XL cargo ship — mission NG-24 — to the International Space Station. Liftoff was scheduled for 7:41 AM Eastern this morning from Cape Canaveral, carrying over eleven thousand pounds of supplies, science experiments, and equipment. This is the second Cygnus XL mission — the larger variant that delivers significantly more cargo than previous Cygnus versions. Meanwhile, SpaceX is preparing for Starship Flight 12 later this month, and Starlink deployment continues with forty-two Falcon launches so far this year alone.

That's all for today. Have a great weekend.