News
SpaceX repairing heat shield, reinstalling Raptors on first orbital-class Starship
SpaceX has begun reinstalling three of the six Raptor engines that will power the first orbital-class Starship and repairing the heat shield that will hopefully protect it on its first trip to space.
Known as Starship 20 or S20, the 50m (~165 ft) tall steel rocket prototype has been stationed at one of SpaceX’s two suborbital testing pads since August 13th. No testing has been done, though, and a small army of SpaceX technicians and engineers have instead spent the last three or so weeks effectively turning a collection of steel tanks, tubes, and parts into a functional rocket. While it’s unclear why SpaceX chose to do that outfitting work at an unsheltered launch pad, new activity suggests that it may be almost complete.
Exactly one month ago, SpaceX stacked Starship S20 on top of Super Heavy Booster 4 (B4) on August 6th, briefly creating the largest rocket in history and completing a fit test that was admittedly just as much a photo op. Ship 20 was rapidly destacked and returned to SpaceX’s Starbase factory, where all six of its Raptor engines were removed. About a week later, Ship 20 returned to the pad and has remained installed on Suborbital Pad B ever since.
At the time, the implication was that SpaceX had removed Ship 20’s engines to allow the prototype to complete cryogenic proof testing with hydraulic thrust simulators. However, despite having carefully modified Pad B over several weeks for that exact purpose, those modifications were rapidly removed before Ship 20’s second rollout. Precluding a proof test with thrust simulation, the next logical conclusion was that SpaceX would still perform a cryogenic proof test before reinstalling Ship 20’s Raptors and moving on to a static fire campaign.
Now, even that appears to have been p1recluded. Instead, as if Ship 20 were the second or third or fourth in a series of prototypes, SpaceX rolled three center Raptors to Pad B on September 5th and began installing the engines on Starship on the 6th. It’s hard to say anything with confidence given how chaotically Starship S20’s to-be-determined qualification testing has changed in the last several weeks but, with plenty of uncertainty, Raptor installation implies that the vehicle will perform its first ambient pressure and cryogenic proof tests with engines installed.
It remains to be seen if Ship 20’s three vacuum-optimized Raptor engines will also be installed over the next few days (seemingly the logical assumption) or if SpaceX will instead complete proof tests and center Raptor static fire testing before finally moving into new territory. SpaceX has never static fired more than three Raptors at once and certainly never tested multiple Raptor Vacuum (RVac) engines in close proximity – let alone all six simultaneously.
Meanwhile, much of the focus of the last few weeks appears to have been on finishing Ship 20 plumbing and avionics wire runs, though it’s hard to say exactly what has been done. What is extremely visible and easy to follow, though, is the process of finishing the first orbital-class Starship heat shield and repairing a few hundred tiles broken during its pathfinder installation. SpaceX has installed 500-1000+ tiles on flown Starship prototypes like SN15 but the company has never come close to the ~15,000 needed to cover the entire windward side of the world’s largest rocket upper stage.
SpaceX has undertaken that process for the first time over the last six or so weeks and unsurprisingly seen a number of successes and failures. At some point along the way, a significant fraction of the ceramic, dinner-plate-sized tiles SpaceX technicians installed chipped, broke, shattered, or ran into other fitment issues. Over the last month or so, a great deal of progress has been made fixing those problem tiles and SpaceX has also more or less completed tile installation on the angular ‘aerocovers’ that protect Starship’s flap mechanisms – requiring dozens of custom tiles with complex shapes and curves.
As of September 6th, Starship S20’s heat shield appears to be around 95% complete and the installation of Raptor engines implies that the rocket’s plumbing, avionics, and tankage are also nearly finished. In other words, after many weeks of work, SpaceX’s first orbital-class Starship prototype could be ready to kick off cryoproof and static fire testing just a week or so (and maybe less) from now. Stay tuned for updates!
Elon Musk
SpaceX just combined Grok, Cursor, and X into one mega plan
SpaceXSI merges Grok, Cursor, and X Premium into one plan, raising questions for Tesla owners.
SpaceX Super Intelligence has rolled out The X Subscription, a single plan that combines Grok, Grok Bot, Cursor, and X Premium.
There are five levels, including a free plan that includes daily Grok access and a starter allowance for Cursor’s coding agents. Plus is $8 per month, Premium is $30, Super is $100, and Ultra is $200. Usage scales as you move up, with Premium offering four times the Plus allowance, Super 15 times, and Ultra 50 times. Annual billing brings the prices down a bit further, with Premium at $300 per year.
We’ve unified plans across Grok, Cursor, and X Premium.
Introducing The X Subscription. https://t.co/SvkrVvBa0p
— SpaceX Super Intelligence (@SpaceXSI) October 12, 2026
The $30 Premium tier is where SpaceXSI says it packs the benefits of SuperGrok, Cursor Pro, and X Premium+, which previously added up to $90 per month. Grok Bot, the agent that handles tasks like ordering food or managing a calendar, is included from Premium upward but not on Plus. Paid plans share one weekly usage pool across Grok and Cursor, and each plan includes at least twice the coding usage of the comparable Cursor plan. Third party models in Cursor remain available on Premium and above, but they are billed separately at API list prices.
That last detail matters because of what has happened around Cursor this year. SpaceX exercised its option to buy the company for $60 billion in June, and the deal closed in August. Weeks later, OpenAI said it will cut Cursor’s access to its models on November 12. Bundling Cursor with Grok usage gives SpaceXSI a reason to steer developers toward its own models.
Tesla integrates Grok Bot into its vehicles for the ultimate personal assistant
For Tesla owners, the connection is Grok Bot. Tesla brought it into its vehicles last month, and at that point access inside the car was limited to SuperGrok Heavy subscribers. Grok Bot is now enabled across every Cybercab and the rest of the Model Y Robotaxi fleet. Neither SpaceXSI nor Tesla has said how the new plans change inside the car access for owners, or whether a $30 Premium subscription now covers Grok Bot in a personal Tesla. That is the question owners will want answered first.
The company is also changing its name along the way. The rollout follows the rebrand Musk announced on October 4, when he said SpaceXAI would become SpaceXSI, and the plan page now carries that branding.
The X Subscription is available to new customers starting now, and SpaceXSI says it will reach everyone globally over the next few days. Existing SuperGrok, Cursor, and X Premium subscribers do not have to switch. The option to move will appear gradually in billing settings, with a prorated refund for the unused time on the old plan. Switching is permanent once the accounts are unified, and old plans are closed to new sign ups.
Energy
Taiwan power firm bets big on Tesla Megapacks as AI data center thirst grows
Taiwan’s Apollo Power orders 260 Tesla Megapacks, betting AI data centers will need more electricity.
A Taiwanese power company has ordered 260 Tesla Megapack 2XL units, roughly 1GWh of storage, betting that electricity rather than chips will decide how fast AI data centers come online. Apollo Power signed the contract with Tesla’s Taiwan energy arm on October 8, according to the company’s filing with the Taiwan stock exchange.
The filing says the units are meant for behind the meter storage and resilient grid projects, with delivery in batches from the end of 2027 and sites operating in 2028. Apollo and Tesla did not disclose a price, citing a confidentiality agreement. Apollo plans to pay for the equipment through a cash capital increase and bank financing, and it noted the purchase is not revenue and does not mean it holds firm orders from its partners.
Biggo reported that only about 33 units are under signed customer contracts, with roughly 150 more in negotiation. The same report put total project investment near NT$130 billion, but that figure includes installation work, with the equipment itself making up about 40 to 50 percent. In other words, a large share of this order is Apollo reserving supply ahead of demand. Apollo’s chairman told local media global large storage supply is tight, which is why the company locked in the units early, according to Yahoo Taiwan Finance.
Apollo also launched an AI data center and grid resilience alliance with Phison, Gigabyte and Shihlin Electric. The pitch rests on a pilot at Phison’s Zhunan plant, where Apollo and Tesla Energy Taiwan installed grid forming Megapack 2XLs in June. Tesla Energy Taiwan general manager Sun Hsiao-Han said the project went from demo to 1GWh in four months, and that Tesla Energy has deployed more than 50GWh worldwide.
Taiwan’s cabinet has approved amendments that would require large energy users to install their own generation and storage, which gives enterprise storage a regulatory tailwind.
The deal fits a pattern Teslarati has tracked. Tesla’s Megablock targets 1GWh commissioned in 20 business days, a pitch aimed squarely at fast moving data center builds. Megapacks already back xAI’s Colossus in Memphis and a 400MW AI campus in Brazil. Sun Media adds that Megapack 2XL was named among the first products in Nvidia’s DSX Ready qualification for battery storage in AI factories.
Deliveries are more than a year away, so the real test is how much of that 260 unit order turns into signed customer contracts.
For a century, the car has asked the same thing of its owner, which is basically to pay attention. Even the best ride-hailing app like Uber or Lyft only moved the problem into someone else’s hands. You stopped driving, but you still sat in a stranger’s car, while half listening, and wondering whether to take a phone call in the car without being rude. The Cybercab has no wheel, no pedals and no driver. With Grok @bot now live inside it, it becomes something else entirely, and I think that matters more than the fleet count that everyone is tracking.
Grok @bot is an agent, not a voice assistant. A rider signs in once, and from then on can say what they need and watch it happen. Order dinner from a favorite restaurant. Fill the grocery cart. Move money, answer email, adjust a calendar, push a work project forward. Tesla engineer Yun-Ta Tsai confirmed it is enabled across every Cybercab and the rest of the Model Y fleet, and Tesla’s own pitch is that Full Self-Driving handles the commute while Grok handles the errands.
Think about what that does to a thirty minute ride. Uber sells you a trip, while Waymo sells you the same trip in a more private setting. Tesla is selling the thirty minutes themselves, returned to you with an assistant attached. Errands that used to eat a Saturday get done between pickup and drop off. The ride stops being dead time between two places and becomes a place of its own.
Tesla integrates Grok Bot into its vehicles for the ultimate personal assistant
Best of all, your Bots follow you into any Cybercab, and presumably your own Tesla, then logs you out automatically when you step out. Anyone can bolt a chatbot onto a screen, but much harder to own the vehicle, the autonomy stack and the assistant at once. It also quietly ties SpaceX’s Super Intelligence business to Tesla’s fleet, after Tesla converted its xAI stake into SpaceX equity. The relationship between those two companies is getting harder to describe as separate.
This is also where the hard questions start, because an agent that can spend your money in a vehicle other people also ride in needs some answers. How does it handle a voice it shouldn’t trust? What gets recorded, who can see it, and for how long? What happens when a rider leaves halfway through a task?
If Tesla gets those answers right, the scoreboard changes. We have been counting Cybercabs and wait times. The more telling number may be what riders accomplish per trip. That is data almost no one else will ever have, and it is the real reason to watch Grok @bot.
What do you think? Leave me a comment.
SpaceX just combined Grok, Cursor, and X into one mega plan
Taiwan power firm bets big on Tesla Megapacks as AI data center thirst grows
Tesla Cybercab and Robotaxi’s real advantage won’t be the ride service
The Pentagon taps Elon Musk to design the battlefield of the future
Tesla Roadster’s new patent preps white-knuckle speeds, keeping it grounded
