Google books Blue Origin's New Glenn for orbital AI compute in 2029

Share
Google books Blue Origin's New Glenn for orbital AI compute in 2029

Google has hired a rocket for its data centers in the sky. Blue Origin's New Glenn will fly Project Suncatcher's first dedicated heavy-lift mission in the first quarter of 2029, carrying several satellites into sun-synchronous orbit — each with computing capacity "roughly at the scale of a data center rack," Google said on Thursday.

The launch services agreement is the project's first commitment to a specific vehicle and a specific year. That matters more than the chip milestone everyone celebrated in October, because the chip was never the hard part of putting AI in orbit. The ride is.

What Google actually bought

New Glenn is a heavy-lift vehicle: 45 tonnes to low Earth orbit, a seven-meter payload fairing with about twice the volume of a five-meter class, and a reusable first stage designed for at least 25 flights, per Blue Origin. The fairing volume is the relevant number. Google is not buying a seat; it is buying a stack — room to fly multiple compute satellites at once, into a plane it chooses, on a schedule it controls.

That is a real step up from October 1, when the project's first satellite rode a SpaceX Falcon 9 rideshare out of Vandenberg alongside more than a hundred other payloads. A refrigerator-sized spacecraft sharing a bus is how you prove a TPU survives launch. It is not how you build a cluster.

Blue Origin declined to disclose the contract value, and Google did not say how many satellites the 2029 mission carries. Both omissions matter, because they are the two numbers that would tell you whether this is a research line item or the first order of a fleet.

The date is the honest part

Suncatcher's public roadmap now reads: one satellite in orbit in 2026, two satellites talking to each other over laser links in 2027, a multi-satellite deployment on New Glenn in 2029. Read against Google's own economics paper, that sequence is a tell.

The paper published in Joule models the cost of power in orbit — the dominant cost being the launch that puts the solar panels there — at roughly $14,700 per kilowatt-year at today's launch prices, against $570 to $3,000 for a US data center on the ground. It projects launch falling to about $200 per kilogram by the mid-2030s, the threshold at which orbital power lands inside the terrestrial range. Today's baseline in that model is around $3,600 per kilogram.

So the 2029 launch is booked for a date at which the arithmetic still says no. Google is buying flight heritage: a deployment sequence, a thermal model validated in vacuum, and a dataset on how sets of TPUs behave when they are moving relative to each other at 7.6 kilometers per second. That is a defensible thing to buy. It is just not the thing the phrase "data centers in space" implies to a reader.

The scale is worth stating plainly, because the coverage tends to skip it. Several satellites at roughly one rack each is a handful of racks. Even the paper's illustrative cluster — 81 satellites flying within a one-kilometer radius — is a server room in the sky, not a data hall. A single hyperscale campus draws tens of megawatts. Google says low Earth orbit offers up to eight times the solar energy per year of a panel at mid-latitude, and it does. It also offers a fraction of the compute per launch.

Who wins

Blue Origin wins the headline it needed. New Glenn has had a rough year: a hotfire anomaly at Launch Complex 36 on May 28 damaged the pad badly enough that Blue Origin abandoned the original pad configuration, and the company is still working through the return to flight, with a redesigned main oxygen valve only qualified on November 10, according to its own updates. A named, marquee, multi-satellite customer for 2029 is credibility at exactly the moment the company is asking the market to believe in its cadence.

Google wins optionality. It is the only hyperscaler with a signed launch contract for AI compute in orbit, and it is spending what is, by Alphabet standards, a rounding error to keep the option alive while the terrestrial build-out runs into interconnection queues and local consent. If launch prices fall the way the paper projects, the company that already knows how to deploy and operate TPUs in orbit is a year or two ahead of the one that starts in 2032. If they don't, Google has bought a research program and some very good engineering data.

SpaceX loses a payload and keeps the relationship — it flew the prototype. That is the useful reminder that this market has no settled supplier. Google is deliberately spreading its launches across providers, and the same instinct that got Blue Origin this mission is why the 2027 laser demo could fly on anyone.

The skeptic's read

The physics does not care about the announcement. In vacuum there is no air to carry heat away, so every watt a TPU burns has to leave through a radiator, and radiator area scales with power — the same trade-off that pushed terrestrial operators to liquid cooling for AI data centers, only with no chiller at the end of the loop. The October prototype runs its TPUs in bursts of roughly 15 minutes and then lets the panels catch up — a duty cycle, not a workload. Google's own paper lists thermal management, high-bandwidth ground links and on-orbit repair as unsolved.

Radiation is the second wall. Google's Trillium chips survived a total ionizing dose equivalent to a five-year mission, but they still flip about one bit per million operations, which is fine for inference and, in the words of project lead Travis Beals, "problematic" for a training run with thousands of chips going for months. Suncatcher is a serving fleet, not a training cluster, and that limits what it can ever be.

Then there is the commercial case. AWS's chief executive has called space data centers "just not economical," and nothing announced Thursday changes the arithmetic — it schedules a test. When we wrote about the first orbital TPU in October, the interesting number was the launch cadence the economics require: roughly 1,800 Starship flights' worth of payload over a decade. A single New Glenn contract does not move that. It is one data point on the curve Google is betting will bend.

What to watch

Three things, in order. The 2027 two-satellite laser test, which is where close-formation flight either works or doesn't — the paper's requirement is pointing precision comparable to hitting a coin from miles away while both ends move. New Glenn's actual return to flight, because a 2029 target set by a rocket that has not flown since May is a target, not a date. And whether Google publishes the thermal duty cycle from the current prototype: if the bursts get longer, the radiators are working, and the whole argument shifts from physics to price.

Until then, the honest summary is that Google has bought a launch for a test, and the test is whether the cooling and the lasers work. That is a better story than the haiku, and a smaller one than the headline.

If orbital compute stays a research program for a decade, is that money well spent or a distraction from the grid problem on the ground? Write to us via our contact page.

Get SiliconNoon in your inbox

One email a day with the AI stories that matter — what happened, the numbers, and why it matters. No spam, unsubscribe in one click.

Read more

Deep Dive — OpenAI's $30B round fell to $20B at a $1.25T valuation

Deep Dive — OpenAI's $30B round fell to $20B at a $1.25T valuation

In September, OpenAI went to investors asking for $30 billion at a $1.4 trillion valuation. On Thursday night the Financial Times reported the target is now about $20 billion at $1.25 trillion, because some investors cut the size of their commitments amid the AI selloff, according to the FT. Hours earlier, the Wall Street Journal had put a number on how far the round has actually got: roughly $11 billion committed, less than half the original target, the Journal reported. That is the single…

Deep Dive — AI neoclouds hit a credit wall as SpaceX borrows $40B

Deep Dive — AI neoclouds hit a credit wall as SpaceX borrows $40B

On Sunday evening, Elon Musk announced on X that SpaceX had closed $40 billion of debt to buy Nvidia GPUs, and used the announcement to deliver a verdict on his competitors: the debt markets have "almost entirely shut off access to money for any company that doesn't have substantial revenue independent of AI data centers," and "the 200+ neoclouds are going to see guppy-level death and consolidation," he wrote. The $40 billion is the number everyone will quote.

Deep Dive — Anthropic's IPO opened at $111.50 and closed at $97.17

Deep Dive — Anthropic's IPO opened at $111.50 and closed at $97.17

Three prices explain Anthropic's first day as a public company, and only one of them is the number the company chose. The offering priced at $83 on Wednesday night. The first trade printed at $111.50 on Thursday morning, 34.3 percent higher. The close was $97.17, up 17.1 percent, and by then the stock had surrendered half of everything it gained at the bell — $14.33 of a $28.50 pop. That $28.50 is the standard measure of what the issuer left behind.

AI 101 — What is liquid cooling for AI data centers?

AI 101 — What is liquid cooling for AI data centers?

Liquid cooling means the heat from AI chips is carried away by fluid running through metal plates clamped to the processors, instead of by air blown past them. It is the plumbing change behind almost every new AI data center, and the reason a modern AI hall looks less like a server room and more like a boiler room. For about thirty years a data center was a windy place: rows of servers, fans at the front, cold air pushed up through a raised floor, hot air pulled out the back.

Deep Dive — Baidu Token Exchange: verified models, uninsured buyers

Deep Dive — Baidu Token Exchange: verified models, uninsured buyers

Baidu spent the past month quietly building a marketplace for other companies' models, then told the world about it this week. Token Exchange — 词元所, Ciyuansuo — went live at Baidu World 2026 on November 12, roughly a month after it began taking suppliers. Robin Li introduced it as "an exchange for large language models," and the analogy he reached for was a stock market: a stock exchange attracts good companies to list so investors can find and back them; Token Exchange, he said, gathers good…

Jensen Huang resigns: Colette Kress is Nvidia's interim CEO

Jensen Huang resigns: Colette Kress is Nvidia's interim CEO

Nvidia's board announced Thursday that Jensen Huang has resigned as president, chief executive and a member of the board, effective immediately, the company said in a statement. Colette Kress, chief financial officer since 2013, becomes interim CEO and keeps the CFO job. Harvey C. Jones, the former Synopsys chief executive, was appointed the first chair of the board in Nvidia's 33-year history. The statement carries two quotes: one from Jones, one from Kress. Neither is from Huang.

Nebius Q3: revenue up 571%, but 2027 capex jumps to $40-45 billion

Nebius Q3: revenue up 571%, but 2027 capex jumps to $40-45 billion

Nebius just posted the best quarter of its life as a public company and gave back 10% of its share price for it. The numbers beat. The plan for 2027 is what the market refused to pay for. The quarter was a genuine beat Nebius reported third-quarter revenue of $980.8 million, up 571% from $146.1 million a year earlier, ahead of the $907.5 million LSEG consensus, according to the company's results release and CNBC's report on the print.