Training AI models in space sounds like science fiction, but Nvidia Corp. (NVDA)-backed Starcloud just made it reality. The Washington-based startup has successfully trained artificial intelligence models from orbit, marking what could be the opening chapter in a race to build massive data centers beyond Earth's atmosphere.

Computing Power 100 Times Stronger Than Anything Before

Here's what makes this interesting: Starcloud launched Nvidia's (NVDA) H-100 GPU into space, and according to CNBC, this chip packs 100 times the computing muscle of anything previously sent into orbit. That's not a marginal improvement. That's a completely different ballgame.

The company has been putting this hardware to work training Alphabet Inc. (GOOGL) (GOOG) Google's open-source language model Gemma from space via its Starcloud-1 satellite. But they didn't stop there. Starcloud also trained an LLM called NanoGPT, created by former Tesla Inc. (TSLA) AI lead Andrej Karpathy, on Shakespeare's works. So yes, there's now a satellite orbiting Earth that's been studying Elizabethan drama.

Starcloud isn't just experimenting for fun. The company describes its mission as building "large data centers in space that are much more rapidly scalable and cost competitive" than traditional Earth-based facilities. Their white paper outlines plans for an orbital data center that's 2.4 miles tall, 2.4 miles wide, and powered by 5 gigawatts of solar energy, complete with cooling panels designed to operate in the vacuum of space.

Why This Actually Matters

Starcloud CEO and co-founder Philip Johnston celebrated the achievement on social media platform X on Wednesday, calling it a "significant step on the road to moving almost all compute to space" and reducing the drain on Earth's resources. The pitch makes some sense: space offers unlimited solar power, natural cooling through radiation, and no concerns about power grid constraints or water usage for cooling systems.

Elon Musk Wants In On This Too

The timing here is noteworthy because SpaceX CEO Elon Musk has been talking up his own ambitions for solar-powered AI satellites. Musk argues these space-based systems would be more cost-effective than traditional data centers anchored to Earth's infrastructure.

The billionaire recently shared that SpaceX is targeting 1 megaton per year in AI satellite launches, which according to his math would generate 100 gigawatts of AI computing power in space. He's even floated the idea of building factories on the moon to facilitate launching AI satellites into deep space. Because why stop at orbit when you can go full sci-fi?

Separately, Musk confirmed that SpaceX is preparing for an IPO next year with a target valuation of $1.5 trillion. He had hinted at this possibility during Tesla's annual shareholder meeting last month.

Michael Burry Questions Nvidia's GPU Claims

While Starcloud celebrates its orbital achievement, "The Big Short" investor Michael Burry has raised eyebrows about Nvidia (NVDA) CEO Jensen Huang's statements regarding the company's Blackwell GPU shipments. Burry is asking people to share pictures showing these GPUs stored in "large quantities" within the U.S. and overseas. It's an unusual public challenge that suggests skepticism about production and delivery claims.

Nvidia (NVDA) demonstrates strong momentum indicators alongside satisfactory quality and growth metrics, though value metrics remain weak. The stock shows favorable price trends across short, medium, and long-term timeframes.

Price Action: NVDA shares declined 1.92% in after-hours trading to $180.25.