Space news: Project Hail Mary, Artemis, data centers

One-liner: Short Wave host Regina Barber launches a new segment, "Spacing Out with Gina," with science correspondent Geoff Brumfiel and All Things Considered host Scott Detrow, covering three space stories: how scientifically accurate the movie Project Hail Mary really is; Elon Musk's push to put AI data centers in orbit; and the latest on the Artemis II moon mission after a launch delay.

How Scientifically Accurate Is Project Hail Mary?

Key findings: The film is adapted from the novel of the same name by Andy Weir — also the author of The Martian — and is set in the near future: scientists discover alien microbes breeding on Venus that begin "eating" the sun. Per Barber's summary of the plot, within roughly 30 years the microbes would dim the sun enough to kill crops, animals, and eventually people; the same microbes are consuming other stars too, so world leaders send a mission to a nearby star that hasn't been affected, hoping to find a way to save the sun. One of the astronauts, scientist-turned-middle-school-teacher Ryland Grace (played by Ryan Gosling), cracks: "I put the 'not' in 'astronaut.' I've never done anything. I've never done a spacewalk. I can't even moonwalk." Barber went in "completely blind" — no book, no trailer — and loved it, especially the jokes, the puppetry, and the visuals; when she let slip that the film includes puppets, it caught Detrow off guard — he'd asked earlier, "Please do not spoil anything" — and he blurted, "Did you just say puppets?" Barber's only reply: "That's all I'm going to say." Astrobiologist Michael Wong of Carnegie Science, who read the book, saw the movie, and loved it, told the show the science mostly holds up: the film correctly nods to special relativity — traveling near light speed causes time dilation, so Gosling's character ages differently than people back on Earth. It does take creative liberties with the biology, though: the alien microbes are water-based and live on the sun's surface, a premise Barber isn't sold on ("I don't know about that"). Wong also flagged what Barber calls an "accidental scientific mistake": the two exoplanets featured in the movie were real candidate exoplanets when the book was written, but in the few years between the book and the film, astronomers found those planets may not actually exist. Barber's take: that's not really a mistake — the science just moved on.

Background: Under special relativity, the closer an object travels to the speed of light, the slower time passes for it relative to a stationary observer — the one hard-science element the show's expert vouched for outright. Whether a given exoplanet exists is itself a moving target: early candidates get inferred from indirect evidence, and more precise follow-up observations can later overturn them — which is exactly what Wong says happened here.

Putting Data Centers in Space: Elon Musk's Orbital AI Compute Bet

Key findings: Brumfiel says launching AI infrastructure into orbit is being taken seriously by some, most aggressively by Elon Musk, who's made it the central mission of SpaceX — his space launch company, which just merged with his AI company, xAI. Speaking Saturday (relative to taping) while unveiling a massive new microchip factory built to supply chips for space-based data centers, Musk said: "I actually think that the cost of deploying AI in space will drop below the cost of terrestrial AI much sooner than most people expect. I think it may be only two or three years." Barber, noting Short Wave's past coverage of how much water and electricity ground-based data centers consume, asked whether space is the fix. Brumfiel says the case for it comes down to power: ground data centers face real limits on generating enough electricity, while in space "you got the sun" — free, plentiful, and the back-of-napkin math looks good. But independent experts he consulted were far more skeptical. The biggest "computery" thing currently in orbit, the International Space Station, generates only around 100 kilowatts from its solar panels; for comparison, xAI's Colossus computing facility in Memphis, Tennessee consumes around 150 megawatts — Barber's response: "Wow. I don't want to." Brumfiel had MIT aerospace professor Olivier de Weck run the numbers, and the verdict was blunt: "You would have to build, just for the power in photovoltaically, a facility that's 1,000 times the size of the ISS." Beyond power, there's an even bigger back-end problem — heat. Space is cold, but it's also a vacuum, so there's no medium to carry heat away from the chips; you'd need massive radiators that use fluid to gather heat and conduct it out into space, nearly as large as the solar arrays themselves. Weighing Musk's "two or three years" against the expert's "1,000 times the ISS," Brumfiel's read is that space computing will grow, but likely starts small and distributed across satellites; giant data centers with kilometer-long solar arrays, he thinks, are well beyond "two or three years" off: "But we'll see. You know, Elon loves to give a tight deadline."

Background: Solar power in space benefits from sunlight undimmed by the atmosphere, making output more consistent. But a vacuum has no air or water to carry away waste heat the way convection does on the ground — heat has to be gathered by circulating fluid and radiated away through dedicated radiators, which is why "generating power is the easy part; getting rid of the heat is the hard part" is the episode's recurring refrain.

Artemis II: A Hydrogen Leak Delays Launch

Key findings: Barber explains that the US is set to send astronauts back into orbit around the moon — the first crewed lunar mission since the early 1970s — with the launch window originally opening soon, on April 1. Brumfiel adds that the mission's rocket has run into technical trouble: a hydrogen leak forced NASA to push the launch back. As of taping, he says, "anything could happen. We'll be watching."

Rich Details & Trivia

• This episode is the long-workshopped debut of a new segment, with the hosts settling on the name "Spacing Out with Gina" live on air; Detrow says: "I'm excited this has gone from something we have talked excitedly about around the coffee machine at work to something into microphones."
• The trio trades origin stories for their space obsession: Barber grew up watching Star Trek: The Next Generation with her dad; Detrow says the sheer scale of space sometimes triggers a kind of "personal paralysis" that's "kind of uncomfortable, but in a cool way"; Brumfiel finds space comforting instead — "I've always found it kind of comforting" — adding that "space is a reminder that we are but on a tiny little dot, and at the end of the day, there's much larger things afoot"; Barber agrees, "me too."
• Detrow had planned to finish Andy Weir's novel before watching the movie; Barber talks him out of it on air: "No, no. Put the book down. Go see the movie."
• When the conversation turns to star-eating alien microbes, Detrow works the bit into the very next story, quipping "Unless it's eaten" about the sun-powered data-center pitch.
• The trio signs off hoping to make the segment a regular thing, with Barber joking that every future "Spacing Out" will probably end with "what's happening to Artemis right now"; Detrow's deadpan reply: "Already has."
• Produced by Rachel Carlson and Kai McNamee; edited by Christopher Intagliata, Amina Khan, and showrunner Rebecca Ramirez.

Cross-Domain Connections

The space-data-center story rhymes with, but doesn't need a hard link to, the acquired domain's NVIDIA trilogy (especially acquired-nvidia-3): that episode's "the data center is the computer" narrative and its figures on global data-center capex ($1T in installed assets, $250B in new annual capex) are the exact ground-based backdrop behind Barber's question about water and power use here. This episode does its own napkin math from the science side instead (ISS's 100 kilowatts vs. xAI Colossus's 150 megawatts, "1,000 times the size of the ISS") — the two pages can point at each other without a forced link. The acquired domain also holds a Tier 0 mirror of an episode on SpaceX (sources/acquired/2020-05-26_spacex_en.md) that hasn't been extracted to Tier 1 yet, so there's no page yet to hang this episode's mention of the SpaceX-xAI merger on. Within science, this episode's Artemis II delay update echoes the already-extracted shortwave-moon-robots episode (which covered Artemis III crew member Andre Douglas's training) — same mission family, but the two episodes stand as separate pages rather than being merged.

Pages to Build

• artemis-program — NASA's return-to-the-moon program; this episode covers Artemis II (first crewed lunar orbit) delayed by a hydrogen leak, part of the same mission thread as shortwave-moonwalks-artemis (Artemis II prep) and shortwave-moon-robots (Artemis III crew training) — best consolidated into one page tracking the program's missions across episodes.
• data-centers-in-space — the feasibility fight over orbital AI data centers (Musk's optimistic timeline vs. an MIT professor's "1,000 times the ISS" estimate, the heat-rejection bottleneck), an emerging AI-infrastructure-meets-space topic likely to recur on Short Wave.

Source · shortwave