Brighter with Herbert Recap: Brian Wang on SpaceX's Orbital AI Compute Plan, Starship Flight 15 and Raptor 4

The latest full video on the official "Brighter with Herbert" YouTube channel, "Every SpaceX Investor Needs To See This Plan", was posted on November 14, 2026 and runs about 48 minutes. Host Herbert talks with Brian Wang, a futurist and science blogger who runs the Next Big Future blog and YouTube channel, about SpaceX's plans to put AI computing in orbit. They recorded on Friday, November 13, the day Starship Flight 15 was scheduled to launch. Topics include a reported order of "space-ready" AI racks, how many Starship launches the plan would need, Nvidia's next chips, cooling and solar power in space, a more powerful Raptor engine, SpaceX's pitch for an Arizona data center, how fast AI is improving, and what AI agents could mean for jobs. Both speakers are strong SpaceX and Tesla supporters, so please read this as one point of view.

This is an independent summary written by GNOUP. It is not affiliated with, sponsored by or endorsed by Brighter with Herbert, Herbert, Brian Wang, SpaceX, Tesla, Nvidia or anyone else named in the video. The views, forecasts and claims described below belong to the speakers, not to GNOUP. Nothing here is investment, financial, legal or tax advice, and nothing here is a recommendation to buy or sell any security. The channel's own disclaimer says its content is for discussion only and that the host is long Tesla and other stocks. Please consult a licensed financial professional before making financial decisions.

Key takeaways

These points come from the video. They are the speakers' views and their readings of news reports, posts and public data; GNOUP has not checked the figures.

• A reported order, not hardware in orbit. Herbert opens by saying Elon Musk told the Baron investment conference that SpaceX plans to start launching orbital AI computers in the third quarter of 2027, ramping in 2028 and 2029, with Nvidia co-designing the hardware. He cites a research report saying SpaceX has ordered about 6,000 space-ready racks for 2027, which he puts at roughly 432,000 GPUs, $42 to $45 billion of hardware and around a gigawatt of computing power. Herbert is clear that no test satellite has flown yet, so these are orders and plans.
• The appeal, in their view. Herbert's argument is that AI companies on the ground are held back by power, turbines, transformers and land, while in orbit the sun supplies power and space provides cooling. If it works, he says, that bottleneck stops applying to SpaceX.
• It all comes down to launches. Brian says the racks are being redesigned from tall cabinets into flat, much lighter units, about one per satellite and roughly 40 satellites per Starship. At that rate, he estimates about 200 launches for a gigawatt, which would need far higher launch rates, three launch sites and more FAA launch approvals. He notes that all the mass humanity has launched to space in 60 years adds up to roughly what a gigawatt in orbit would require.
• Starship Flight 15's stakes. Brian calls the attempt to catch both the booster and, about 22 hours later, the ship a potentially historic step toward a fully reusable rocket. Herbert mentions an estimate from another guest that it had about a 50% chance of success, and says SpaceX plans to keep flying monthly until it works.
• A bigger Raptor could cut launches. Brian points to a Musk post saying it is tempting to build a 500-ton-thrust Raptor 4 for Starship's outer ring of engines. Combined with lighter satellites, he estimates this could cut the launches needed for a gigawatt from about 200 to around 40.
• Cooling "is solved," he argues. Brian says critics of space data centers overlook that Starlink satellites already run about 100 megawatts of solar power and radiators in orbit, and that chips running hotter shed heat far faster. Herbert's view is that companies would not commit billions without believing the engineering works, while saying we will find out.
• Rubin chips and a common design. Brian says Nvidia's Rubin GPUs, paired with its Vera CPUs, should be the main chips in 2027, and that SpaceX plans to use the same lightweight rack design in orbit and in its "Minihard" and "Megapod" data centers on the ground.

Newsworthy insights

These are claims, predictions and opinions from the video. GNOUP has not verified them.

• Solar machines. Brian says SpaceX has also ordered machines from a German company to build solar panels, and describes satellites whose arrays unfold accordion-style, each delivering, he says, a bit more power than the International Space Station.
• The Arizona pitch. Describing SpaceX's proposal for a data center in Arizona, which still needs approval, Brian says the company is pitching grid power backed by large Megapack battery banks, recycled water, little or no noise and lower property taxes for the county.
• Built to move. Brian argues that most of a data center's cost is in the chips, racks and portable turbines, which go in during the last couple of months, so SpaceX could move that equipment to another state if a local government blocks it.
• How fast AI is improving. Brian argues that AI improvement sped up from late last year and that labs are holding back internal models even better than what they have released. He cautions that an "intelligence explosion" is not guaranteed, because AI moves fastest in areas where results are easy to verify and score.
• Agents and jobs. His view is that people will increasingly manage many AI agents, moving into manager or executive roles, while others will live as "retired asset owners." His practical advice is to start using AI tools and ask them to help you learn.
• Revenue forecasts. Brian says his earlier forecast of $500 billion to $1 trillion in SpaceX annual recurring revenue by the end of 2027 matched what Musk later said on SpaceX's Q3 call.

Watch the video

Every SpaceX Investor Needs To See This Plan on YouTube (official channel "Brighter with Herbert," published November 14, 2026, about 48 minutes)

Episode guide

These chapters come from the video's own description (titles lightly reworded). Each link opens the video at that point.

• 0:00 SpaceX's orbital AI order
• 2:54 Starship Flight 15 stakes
• 6:49 Launch cadence to 2028
• 14:03 Vera Rubin chip pipeline
• 16:32 Cooling physics reality check
• 17:57 Raptor 4 boosts payload
• 23:14 Solar arrays for space AI
• 25:05 The Arizona data center pitch
• 32:28 The AI acceleration curve
• 39:29 Agents and job shifts

About this summary

This recap was prepared from the video's title, description and chapter list on YouTube and its automatic transcript. Automatic transcripts can garble names, so we left out the name of the research publisher, which we could not confirm. We also left out a political aside, the host's website promotion and the Brighter Finance, options course, referral, product, Patreon, membership and merchandise links. GNOUP has no connection to any product, offer or company mentioned and does not promote any of them. For the exact words and context, watch the video. This article is general information only. It is not investment, financial, legal or tax advice, it does not recommend buying or selling any security, and it is not a substitute for advice from a licensed financial professional.