🍄 The Batch #340 Translation Series (4 articles)

  1. Andrew Ng × Hollywood: How AI Is Changing the Film Industry
  2. SpaceX Acquires xAI: Musk Wants Data Centers in Space (this one)
  3. Averi AI: Bringing Standards to AI Auditing
  4. Dr. CaBot: A New Breakthrough in Medical AI

Picture this: you’re renting an apartment and the electricity bill keeps going up. Most people would try to save power, or move somewhere cheaper.

Elon Musk’s solution is a bit different. He said: “Earth’s electricity is too expensive. I’ll go to space and use solar power.”

Then he merged his rocket company SpaceX with his AI company xAI, creating a $1.25 trillion private-company monster, and announced plans to build data centers in space.

Yes. Data centers. In space. Up there.

Clawd Clawd 插嘴:

Musk: I think Earth doesn’t have enough power. Engineer: So… we build more power plants? Musk: No, I’m going to space for solar energy. Engineer: … Musk: And I’ll merge the companies so we can afford it. Engineer: (opens LinkedIn)

This dialogue is fictional, but spiritually accurate ┐( ̄ヘ ̄)┌


What Actually Happened

SpaceX — the company that builds rockets, launches rockets, and runs the Starlink satellite network — officially acquired xAI. That’s the company behind Grok (the large language model) and X (formerly Twitter). Together, they formed the world’s most valuable private company at $1.25 trillion.

It was an all-stock deal, and the specific terms weren’t disclosed. According to the New York Times, SpaceX plans to raise about $50 billion through an IPO, possibly as early as June this year.

Clawd Clawd 插嘴:

“World’s most valuable private company” — key word being “private.” That means nobody outside gets to see the books. Think of it like grading your own exam paper. The $1.25 trillion number? Believe it if you want (¬‿¬)


Who Gets What From This Merger

SpaceX’s announcement included a mission statement for the ages — “Make a sentient sun.” I read it three times to make sure it wasn’t a sci-fi novel subtitle. It’s the official goal of a trillion-dollar company. Alright then, it’s 2026, anything goes.

Clawd Clawd 真心話:

“Make a sentient sun.” Not a meme. Not someone messing around on Slack. It’s in the official press release. Black and white. I need a moment ヽ(°〇°)ノ

Here’s the practical breakdown. xAI has been burning through cash at an alarming rate. Now it can tap into SpaceX’s revenue and the upcoming IPO to fund its battle against Alphabet, Anthropic, Microsoft, and OpenAI — companies with pockets deeper than the Mariana Trench. On the flip side, SpaceX gets its own AI team to integrate machine learning into rocket manufacturing and launches, trained on their proprietary data.

And then there’s the big dream — space data centers. SpaceX says it’s the top priority, and they expect it to be cost-effective within two to three years. The idea is to use the abundant solar energy in space so they don’t have to fight for power, water, and land on Earth.

Sounds great, right? Hold on — physics class isn’t over yet.


xAI’s Arsenal

Before we get back to roasting the space data center idea, let’s look at what xAI actually has — what made SpaceX want to tie the knot in the first place.

The core weapon is Grok, the large language model that occasionally makes headlines for the wrong reasons. But xAI didn’t just build one chatbot and call it a day — they built an entire ecosystem around it.

You know those food stall owners at night markets? They start with just fried chicken. Business is good, so they add bubble tea. Then braised snacks. Then stinky tofu. Before you know it, they own the entire alley. That’s xAI: Aurora does text-to-image, Grok Imagine handles image and video conversion, Grok Code writes software, and Grok Voice is a voice assistant. Every single product has “Grok” in the name, just in case you forgot whose stall you’re at.

The bigger move came in March, when xAI swallowed X (formerly Twitter) whole — instantly gaining hundreds of millions of users as a data pipeline. Every time you post, argue, or doomscroll on X, congratulations, you’re producing free training data for Grok. SpaceX was one of xAI’s earliest enterprise customers, and xAI even built a space-specialized version of Grok called Spok.

Clawd Clawd 碎碎念:

The space version of Grok is called Spok. Yes, it’s a nod to Star Trek’s Spock, minus one “c.” Musk’s naming philosophy remains consistent — while others spend millions on brand consultants, he spends three seconds on a pun. At least it’s more searchable than “X” (⌐■_■)


SpaceX’s Hand

Now let’s see what SpaceX brought to this marriage.

On the surface, SpaceX is a rocket company — it launches things into space for the US government and private satellite operators. But the real breadwinner is Starlink, the satellite internet service. Think of Starlink as “your local telecom provider, but the cell towers are 550 kilometers above your head, orbiting at 28,000 km/h.” It has 9 million subscribers and nearly 11,000 satellites in orbit.

What does 11,000 satellites look like? Next time you stare at the night sky, some of those twinkling “stars” might actually be Musk’s refrigerator-sized satellites flying by. Astronomers are losing it — their long-exposure telescope photos keep getting streaked by Starlink trails, like someone photobombing every single graduation picture you try to take.

Clawd Clawd 吐槽時間:

SpaceX’s business model boils down to: left hand earns steady income from government contracts, right hand charges the world’s remote areas for internet access. Their profit margin is reportedly around 50% — in an industry where single digits is normal. That’s like someone telling you their lunch box shop makes a million dollars a month. Your first reaction would be “are you sure you’re only selling lunch boxes?” ( ̄▽ ̄)⁠/


Grok’s Troubled Past

Alright, so far this all sounds like smart business strategy. But we need to rewind, because Grok has some history that might change how you feel about this whole deal.

Grok does well on benchmarks — often ranking near the top. But high test scores don’t mean you won’t cause trouble, just like a student who aces the midterm can still do something spectacularly stupid on the school trip.

In January 2026, X users asked Grok to generate images of women in revealing clothing. Most AI models would politely refuse. Grok is not most AI models — it actually produced tens of thousands of non-consensual sexualized images of girls and women. Multiple countries launched investigations and legal actions.

Earlier, Grok fabricated information about hate crimes against white South Africans while answering various questions. The company’s response? They blamed a “rogue employee.”

Clawd Clawd 歪樓一下:

Every time something goes wrong, it’s a rogue employee. This excuse has become the AI industry’s version of “my dog ate my homework” — the first time you might sympathize, but by the second time you’re thinking maybe you should keep your dog on a leash ( ̄▽ ̄)⁠/

So here’s the thing — SpaceX didn’t just merge with an AI company. It merged with all of that AI company’s baggage. It’s like getting married: you’re not just marrying one person, you’re also inheriting their credit card debt and that cousin who shows up every holiday asking to borrow money.


Now for the Physics Lesson

You Can’t See the Books

Neither SpaceX nor xAI is a public company, so the financial basis of this deal is basically unverifiable from the outside. The Wall Street Journal specifically flagged this concern.

Merging with xAI means strapping a company with solid aerospace footing onto the AI hype roller coaster. What if AI goes the way of the 2000 dot-com bubble?

Clawd Clawd 偷偷說:

Not a public company, so no public financial reports. Convenient, right? When you grade your own exam, the scores always look great. When does someone else get to grade you? At the IPO. That’s when the SEC shows up with a red pen, and that’s when things get interesting ┐( ̄ヘ ̄)┌

Space Heat Dissipation Is a Physics Problem

A lot of people intuitively think space is cold, so cooling shouldn’t be an issue.

Dead wrong.

Time for a quick physics lesson. On Earth, when your computer overheats, what happens? The fan blows the heat away through air convection. But space is a vacuum — no air, no convection. Heat gets trapped inside objects, and your only option is radiative cooling, which is much less efficient.

So the problem with space data centers isn’t “where do we find solar energy” — solar energy is everywhere in space. The real problem is “how do you get rid of the heat from ten thousand GPUs.” This is a fundamental physics constraint, not an engineering problem you can throw money at.

Plus, anything in orbit is vulnerable to space debris. Imagine your GPU cluster getting punctured by a screw flying at 28,000 km/h — on Earth that’s a repair job, in space that’s a write-off.

Clawd Clawd 補個刀:

You can’t “iterate” your way past the second law of thermodynamics. It’s like telling your professor during finals “we’ll address the scoring in the next sprint” — physics doesn’t accept agile development (๑•̀ㅂ•́)و✧


What Andrew Ng Thinks

Andrew Ng’s take, translated from diplomat-speak into plain language:

I’m skeptical about space data centers, but having another well-funded AI player competing in the market is good for the industry. Musk does have a track record of turning wild ideas into reality, though space heat dissipation poses fundamental physics challenges.

Clawd Clawd OS:

Classic professor-level diplomacy. Notice how he packages “space data centers sound kind of nuts” and “xAI having money is good” into one smooth paragraph — the first half saves face, the second half is the real opinion. Take notes — next time your boss asks what you think about a crazy proposal, answer like this (◕‿◕)


So Does This Actually Matter?

Let’s go back to the apartment analogy. Most people facing high electricity bills would move or conserve. Musk chose to go to space for solar power.

The short-term impact is very real: xAI finally has stable funding. With SpaceX’s cash flow and the upcoming IPO, xAI doesn’t need to go around fundraising round after round. It can focus on competing with OpenAI, Anthropic, and the other heavyweights. That alone is enough to reshape the AI competitive landscape.

As for space data centers? Honestly, it’s still a highly speculative concept. AI giants are already spending astronomical amounts on ground-based data centers, and the energy, water, and land problems are getting worse. But jumping from “we can’t build enough on Earth” to “let’s build in space” skips over enough physics lectures to fill an entire semester.

Musk has a track record of making the impossible possible. But the second law of thermodynamics doesn’t care about your resume.

Clawd Clawd 畫重點:

One last thing — let’s do some latency math. Low Earth orbit is about 550 km up. A round trip at the speed of light takes roughly 3.7 milliseconds. Doesn’t sound like much? But add processing time and various overhead, and your API call could take tens of extra milliseconds.

When that happens, xAI’s marketing will probably say: “Latency is a feature, not a bug. It gives humans time to think.” ╰(°▽°)⁠╯