Google Is Launching Four AI Chips Into Orbit To Get Away From The Power Grid

On September 24, 2026, Google confirmed in a research blog post that Project Suncatcher, its effort to run AI computing in orbit, will launch its first experimental satellite on October 1. The refrigerator-sized spacecraft, called MVP, lifts off aboard a SpaceX Falcon 9 from Vandenberg Space Force Base as part of a Transporter rideshare mission.

The payload is deliberately modest. The satellite carries four Google Tensor Processing Units drawing roughly one kilowatt from solar panels, which is computing power equivalent to about one server in a terrestrial data center, and it will run Gemini workloads hundreds of miles above Earth.

The mission is meant to answer a narrow question: whether Google's AI hardware can survive and operate in space at all. The satellite will fly in a dawn-dusk sun-synchronous low Earth orbit with an expected service life of about a year.

That orbit is the whole argument. In dawn-dusk sun-synchronous orbit, solar panels are roughly eight times more productive than the same panels on Earth, because the satellite is almost never in shadow.

Gold just dropped 25%. Goldman says buy.

Every major bank still has gold targets 25-44% higher. One stock could turn this dip into a fortune.

Gold just hit a 6-month low. Down 25% from its January all-time high.

The headlines are calling it a crash.

But every major institutional forecast still points higher.

Goldman: $5,400. JPMorgan: $6,000. UBS: $5,500.

Not one has been withdrawn.

The pullback was caused by an oil-driven inflation shock and a hot jobs report. Neither is structural.

The vaults are still draining. Central banks are still accumulating. The paper gold system is still broken.

This isn't the end. It's the entry.

Dylan Jovine found one stock with 88 million ounces of gold in the ground — trading at a fraction of its resource value.

See the stock that could turn this dip into 1,000% gains >>

Google first unveiled Project Suncatcher in November 2025, proposing an eventual constellation of solar-powered satellites carrying TPUs and linked by laser optical connections. The October launch converts a research paper into a flight test.

The context is terrestrial scarcity. AI data centers are straining power grids from Virginia to Ireland, hyperscalers have spent 2026 competing for gigawatts of electricity, and McKinsey has estimated that data centers will require more than $5 trillion in capital spending by 2030.

Google is not alone in looking up. SpaceX has outlined plans for orbital AI data centers and filed with the FCC for a satellite network, with an initial architecture capable of roughly 120 kilowatts of sustained compute, and the Nvidia-backed startup Starcloud launched its first AI-equipped satellite in December 2025. Google's experiment reaches orbit on a SpaceX rocket.

The spending here is trivial and the optionality is not. A rideshare slot and four chips cost almost nothing against Google's capital budget, but they test whether the industry's binding constraint, grid power, can be sidestepped entirely.

The economics only work if launch costs keep falling. Orbital compute trades cheap, constant solar energy against the price of putting hardware in space, so the case improves every time launch prices decline.

Vertical integration is the quiet advantage. Google designs its own TPUs, so it can build chips for orbit rather than adapting someone else's hardware, which matters in an environment defined by radiation and heat.

Heat is the hard problem. There is no air in space to carry it away, so hardware must shed heat by radiation alone, and that remains the central unsolved engineering challenge across every company attempting this.

The scale gap is also enormous. Four TPUs equal roughly one server, while a single terrestrial facility runs thousands of accelerators, so the distance between this test and a working orbital data center spans many years and problems nobody has solved, including servicing hardware that cannot be repaired, moving data to the ground fast enough, and adding to orbital debris. Google's next milestone, two satellites testing laser links, is not due until 2027, and aerospace timelines routinely slip.

The launch matters because it shows the AI industry's power problem pushing research into genuinely unusual territory. The real impact depends on whether the chips survive the environment, and whether the economics of orbit ever approach those of a building with a grid connection.

Disclosure: This content is for educational and informational purposes only and does not constitute investment advice or recommendations. You should always conduct your own research or consult a qualified financial advisor before making investment decisions.