SpaceX Discussion Thread

You make some good points about reusing existing buildings. The one that uses an Electrolux factory probably sucks :) They did some quick and dirty with the gas generators and apparently got sued or had permit troubles.

But Google seems willing to pay $920M a month to use them. Maybe Anthropic is not so picky.

I am not in a position to know or judge the details. Now that they are public, it will be interesting to see how much information comes out on the earnings calls.

Another thing I wonder about is how much effect each Starship test launch will have on the stock price. I would hate to see a "rapid unexpected disassembly" of the SPCX support level.

I bought a Model Y to do due diligence on my TSLA investment and risk my life daily going 75 mph with a neural network driving the car. I can't do that with a rocket. I guess I will just have to roll the dice.
 
One more thing, xai colossus is built on H100 (Hopper) GPU. These are three generations old, behind Blackwell and Rubin. Hopper (March 2022), Blackwell (2024), Blackwell Ultra (2025), Rubin (2026). Hopper GPUs are still useful due to supply chain issues but in compute years they are ancient. I think people are not aware of these generations when they see that what is being sold is capacity on 4+ year old technology. This is all disclosed in the S1 filing. I downloaded it and started reading some of the details and there is some concerning cautionary tales in it.

Link to S1 is here. spacexS1
 
I asked Gemini why Google would lease compute from such a shoddily built data center and it pointed out that Collossus 1 was a mismatch of different GPU hardware and that xAI had had problems maximizing their training. I think the older GPU's are being used for inference workload.

Collossus 2 has 520,000 GB200 and 30,000 GB300, so that is why SpaceX trains its models there.

The S1 is huge, but I should probably download and read the fine print. However, from my dot com IPO experience (bad omen) those documents go out of their way to scare investors so you cannot sue them.

But I suppose those preaching the "retail is the exit liquidity" gospel might have a point. I certainly get that feeling with BTC.

What does your user name refer to? I thought it would be a highway but Grok suggested it might be computer networking lingo. What did you do before you retired?

I was mainly a C, C++, C# programmer who started a small B2B company in the 1980's, that ran out of money so I did contract work, then went back to the company as an employee. in 2001 I set up our small "data center" that grew to 10 racks and took care of the cisco router and SQL databases. I got driven out as the big company that bought us was moving to AWS.
 
joesxm3 said:
I asked Gemini why Google would lease compute from such a shoddily built data center and it pointed out that Collossus 1 was a mismatch of different GPU hardware and that xAI had had problems maximizing their training. I think the older GPU's are being used for inference workload.

Collossus 2 has 520,000 GB200 and 30,000 GB300, so that is why SpaceX trains its models there.

The S1 is huge, but I should probably download and read the fine print. However, from my dot com IPO experience (bad omen) those documents go out of their way to scare investors so you cannot sue them.

But I suppose those preaching the "retail is the exit liquidity" gospel might have a point. I certainly get that feeling with BTC.

What does your user name refer to? I thought it would be a highway but Grok suggested it might be computer networking lingo. What did you do before you retired?

I was mainly a C, C++, C# programmer who started a small B2B company in the 1980's, that ran out of money so I did contract work, then went back to the company as an employee. in 2001 I set up our small "data center" that grew to 10 racks and took care of the cisco router and SQL databases. I got driven out as the big company that bought us was moving to AWS.
Click to expand...
My user name is from the Sony Playstation game Gran Turismo. Route246 is one of the most popular tracks in the game's series. It is highly accurate because the main straight runs directly in front of where their offices were before they moved and the site dimensions and artifacts are real and most are still there today.

I read more than half of the S1, the parts that are relevant to the financial viability of the stock. It is very difficult to discern how the needle to profitability gets threaded for this company. Many selling points have to perfectly converge in optimal ways in order to get to break-even. The Starlink service is profitable for now but it also burns a lot of capex due to the relatively short life of the satellites that need to be continuously refreshed as older ones decommission. It will also diminish in terms of scale compared to the projections stated in the S1 (my assessment).

In a nutshell, the high capex requirements is a sunk cost and one of the needles has to thread utilization to well over 50% (currently around 10%) in order to stop burning through so much cash. The utilization also needs to be for workloads that are paying a very high premium price at current market rates. It appears to be challenging as token generation is getting more efficient, more competitive and smarter meaning cost per token is going to degrade quickly as China and open source models become more relevant.

My concern is the stock has meme-like symptoms, although it has not reached that level yet. There is a "greater fool" element to the stock's price. While Starlink and launch businesses are touted heavily in the S1 the main driver to valuation is the AI division which shows no viable way to drive to break-even at this point other than some optimistic statements and strong faith in the firm's business model.
 
Thanks. I appreciate fact-driven push back on my fan-boy mindset.

I agree that the huge TAM from enterprise AI is a puzzling leap of faith.

Maybe the more correct way to view things is that SpaceX is a company that will do great things but not in a way that people buying in now will be able to make money.

The early investors, including me buying BPTRX when SpaceX was valued at $400B, will make money. But all of the early massive gains are probably over.

I have sold 40% of my BPTX over the past week, locking in a 50% LTCG. I will probably continue to sell and maybe even sell 25% or 50% of my IPO shares. After that I can reassess whether to slowly DCA at lower prices in the future.

Well, enough FIRE and stocks for today. Off to drink some wine and listen to music at a local vineyard.
 
joesxm3 said:
Thanks. I appreciate fact-driven push back on my fan-boy mindset.

I agree that the huge TAM from enterprise AI is a puzzling leap of faith.

Maybe the more correct way to view things is that SpaceX is a company that will do great things but not in a way that people buying in now will be able to make money.

The early investors, including me buying BPTRX when SpaceX was valued at $400B, will make money. But all of the early massive gains are probably over.

I have sold 40% of my BPTX over the past week, locking in a 50% LTCG. I will probably continue to sell and maybe even sell 25% or 50% of my IPO shares. After that I can reassess whether to slowly DCA at lower prices in the future.

Well, enough FIRE and stocks for today. Off to drink some wine and listen to music at a local vineyard.
Click to expand...
I have a little experience with Baron funds back when Bob Brinker was touting them many decades ago. They were on his private recommended list and I had a small holding based on that. One month, it was announced that the fund was being removed from the recommended list without much explanation. There wasn't much public news available back then before the Internet but I sold at that time based on what was in his newsletter and move the funds to a 500-equity fund. I suspect something happened with the fund management or other thing and since it was part of the newsletter Bob never talked about it on his radio show.

I think the lesson to be learned is to look for objective data, not touts or fluff. When someone or something (i.e. S1) tells you there is high risk it means you should take that into consideration for your financial decisions. Let the masses following meme-stock mania do their thing. When it has nothing to do with fundamentals and valuation, believe them. It is your money, you should always be skeptical.
 
I am watching the Peter Diamandis podcast and they just mentioned a scary risk factor that is listed in the 38-page risk factor section of the SpaceX S1.

This is the Kessler Effect, which is a mathematical theory that with a critical mass in a given orbit there will be a collision that makes debris that causes more collisions like a nuclear reaction, making the orbital area unusable for decades.

Given the heavy reliance on Starlink and orbital data center satellites, this sounds scary.
 
joesxm3 said:
I am watching the Peter Diamandis podcast and they just mentioned a scary risk factor that is listed in the 38-page risk factor section of the SpaceX S1.

This is the Kessler Effect, which is a mathematical theory that with a critical mass in a given orbit there will be a collision that makes debris that causes more collisions like a nuclear reaction, making the orbital area unusable for decades.

Given the heavy reliance on Starlink and orbital data center satellites, this sounds scary.
Click to expand...
Starlink uses some form of collision avoidance and routinely burns to adjust their low earth orbit based on best available data so they are proactively attempting to lessen the chance of collision. The band that Starlink operates in cleanses itself within years, less than a decade I read. The band will probably soon become more crowded but Starlink currently dominates there. This cash cow is operating at a lower band with competent countermeasures to reduce risk. The risk is higher up bands like where GPS resides, from what I read.
 
joesxm3 said:
I am watching the Peter Diamandis podcast and they just mentioned a scary risk factor that is listed in the 38-page risk factor section of the SpaceX S1.

This is the Kessler Effect, which is a mathematical theory that with a critical mass in a given orbit there will be a collision that makes debris that causes more collisions like a nuclear reaction, making the orbital area unusable for decades.

Given the heavy reliance on Starlink and orbital data center satellites, this sounds scary.
Click to expand...
Reminds me in a strange way to debris on Mt. Everest:
 
Just a quick update.

Apparently the Starfall pod was successfully deorbited, but not much news.

I saw some discussion saying that it is designed to fit into the Starship pez dispenser used to launch satellites. It was said that in addition yo being used to return product from orbital factories, Starfall could be used as a point to point fast delivery service for use cases like rapidly shipping emergency or military supplies. In this scenario Starship would carry the cargo, but the problem of landing Starship to unload would be avoided.

The other news is that, having lost all of its rockstar AI researchers, SpaceX seems to be changing strategy. Rather than trying to invent a clever AI model, it is rewriting the training code in C/C++ to make it 10x or more efficient. Basically, trying to change the game from an artist/inventor game to an engineering/manufacturing game where an advantage in compute capacity and cost control is more important than progress in model design.
 
joesxm3 said:
Just a quick update.

Apparently the Starfall pod was successfully deorbited, but not much news.

I saw some discussion saying that it is designed to fit into the Starship pez dispenser used to launch satellites. It was said that in addition yo being used to return product from orbital factories, Starfall could be used as a point to point fast delivery service for use cases like rapidly shipping emergency or military supplies. In this scenario Starship would carry the cargo, but the problem of landing Starship to unload would be avoided.

The other news is that, having lost all of its rockstar AI researchers, SpaceX seems to be changing strategy. Rather than trying to invent a clever AI model, it is rewriting the training code in C/C++ to make it 10x or more efficient. Basically, trying to change the game from an artist/inventor game to an engineering/manufacturing game where an advantage in compute capacity and cost control is more important than progress in model design.
Click to expand...
Hey, my late 90's C/C++ skills could once again be used to optimize things! :)
 
I notice that some people are selling SpaceX because of the aborted launch. IMO, If a person’s investment in SpaceX stock depends on one test launch he/she probably don’t understand what they have invested in.

Look up Gemini VI, launch shutdown. Now that was a hair raising event.
 
Last edited:
My spcx IPO investment depended on market principles and I found it odd that long-standing rules needed to be bent to favor institutional investors over retail investors. So I didn't buy any, and I made sure our portfolio had absolutely minimal exposure (We estimate 2%).

It's currently 125.05.

It appears the only strategy for those who bought are to dump it now, or hold and hope for the best.
 
In hindsight, perfect execution would have been for me to sell at $165 on the day my Fidelity anti-flip expired. I also had a chance to close my PUT at a 20% profit at one point.

My initial IPO plan was to get 50 to 100 shares to dip my toe, and I was surprised to get 400. But as usual, when the time came to trim I was optimistic. And also I was running out of room in my 2026 AGI.

I did sell about 40% of my BPTRX, starting on IPO day. So I didn't do everything completely wrong.

Also I had the sense to sell about 45% of my Micron on the way up. Not a lot at the nose bleed level, but I think almost all of my sales were above the current MU price. My average cost for the shares I still hold is around $85.
 
I was up quite a bit after launch but we are right at IRMAA level one for this year (affects 2 years out) and hence did not sell. I only have a taxable brokerage account.
 
RetiredHappy said:
I was up quite a bit after launch but we are right at IRMAA level one for this year (affects 2 years out) and hence did not sell. I only have a taxable brokerage account.
Click to expand...
Your post reminded me how stupid I actually was since the 400 shares are in my ROTH IRA :mad:

I sold all the BPTRX in IRA and a decent chunk in taxable, which is what pushed my AGI. But there was no reason not to sell the 400 in ROTH, except greed.
 
You must log in or register to reply here.

Similar threads

Jerry1
iShares® iBonds® High Yield and Income ETF Questions
Semi-Retyrd
S
camfused
CCRC vs Age in Place
5 6 7
PaulR888
P
S
Hello from DFW
Flyfish1
F
M
Repair or replace my car. Help with advice and different perspective.
2 3 4
Jerry1
Jerry1
M
Opinions re College Undergrad
2
wmackey
W

Latest posts

Back
Top Bottom