SpaceX IPO

[Psi]

GEO would have too much latency though, for agentic AI you need something that can query Earth-bound datacentres and do something then iterate until a solution converges.  Usually such solution is wrong, but it will be slow and wrong if you add 700ms of latency to each query.

I was assuming this was more for AI/LLM training and long running AI processing of datasets stored at the space datacenter, rather than for serving humans on earth waiting for a reply to their LLM query or crunching earth based datasets in real-time.
Training and running AI processing doesnt care about latency if the datacenter has the dataset.
But even if it was used for earth queries, 1400ms roundtrip latency to GEO isn't that bad to get a reply to your LLM query.

But that was just an assumption on my part.

If the plan is LEO i will get my popcorn now as that will be some drama to watch.

[BitBangBro]

It the plan is LEO i will get my popcorn now as that will some drama to watch.

I have a feeling that the plan is to lure the capital into investing into development of HPC space-rated systems.
They might not be all that great for orbital AI, but SpaceX sure could use them for Mission To Mars and other things, predominantly mass-surveillance, military programs/"Space Force" etc.

[Psi]

I have a feeling that the plan is to lure the capital into investing into development of HPC space-rated systems.
They might not be all that great for orbital AI, but SpaceX sure could use them for Mission To Mars and other things, predominantly mass-surveillance, military programs/"Space Force" etc.

Space based technology does tend to cause a really large amount of trickle down technology in other sectors.

[Stray Electron]

 OMG! You guys are talking like Engineers and not like politicians or stock brokers who have something to sell!

[watchmaker]

OMG! You guys are talking like Engineers and not like politicians or stock brokers who have something to sell!

After 40 years of investing, I have concluded IPOs are a loser for the subscriber and a winner for the Corp CFO.  I would wait until it settled down.  Many lose value after 4 or 5 weeks.

[tom66]

GEO would have too much latency though, for agentic AI you need something that can query Earth-bound datacentres and do something then iterate until a solution converges.  Usually such solution is wrong, but it will be slow and wrong if you add 700ms of latency to each query.

I was assuming this was more for AI/LLM training and long running AI processing of datasets stored at the space datacenter, rather than for serving humans on earth waiting for a reply to their LLM query or crunching earth based datasets in real-time.
Training and running AI processing doesnt care about latency if the datacenter has the dataset.
But even if it was used for earth queries, 1400ms roundtrip latency to GEO isn't that bad to get a reply to your LLM query.

But that was just an assumption on my part.

If the plan is LEO i will get my popcorn now as that will be some drama to watch.

Inference (responding to a query) is 100x more resource intensive than training in terms of AI power and datacenter size.  So it wouldn't make sense to only put training data centers in space, if this was the plan.  It would have to be inferencing, but that just doesn't make sense because of latency.  You could put them in LEO but yeah, 5 year lifespan and GPUs burning up on re-entry sounds pretty bloody wasteful to me.

[Psi]

After 40 years of investing, I have concluded IPOs are a loser for the subscriber and a winner for the Corp CFO.  I would wait until it settled down.  Many lose value after 4 or 5 weeks.

I think we're all on the same page with that. I don't think anyone is suggesting to buy in to the IPO.
IPOs are notorious for
- Crashing and burning from their overly inflated starting price before beginning a slow recovery.
- Shooting up from a buying frenzy and then collapsing hard before beginning a slow recovery.

Getting into an IPO at the start is probably the highest risk you can take.

OMG! You guys are talking like Engineers and not like politicians or stock brokers who have something to sell!

Were in an engineering forum
This isn't an investment thread, or it wasn't meant to be. I probably should have been more clear with the thread title.

[tom66]

OMG! You guys are talking like Engineers and not like politicians or stock brokers who have something to sell!

After 40 years of investing, I have concluded IPOs are a loser for the subscriber and a winner for the Corp CFO.  I would wait until it settled down.  Many lose value after 4 or 5 weeks.

Right.  SpaceX will be employing the top analysts. If they list at, say $30 a share and the price rockets up to $60, they are leaving billions of dollars in free cash on the table.  (Yes, the company will be 'worth' more, but it isn't liquid cash like direct stock sales are.) Meanwhile, if they list at $70 and the price drops to $60 after a few weeks, then they're still happy, and most investors won't get too worried about that in the shorter term.  The incentive is always to advertise above the present value of such shares and hope suckers get dragged in.  In 6 months time the dust settles and maybe it's then worth buying into.  In rare cases they get it wrong and the stock goes meteoric on day one.  But most IPOs lose money for shareholders after a few months.

[brucehoult]

most IPOs lose money for shareholders after a few months.

If you care what the price is after only a few months then you should be going to the casino not the stockmarket. Something like this should be bought for what it will be worth after 10 or 20 years, not next month.

Sure, maybe if you wait you can buy the dip for a 10% or 20% saving. Or maybe not. But that's speculation, not investment.

[tom66]

most IPOs lose money for shareholders after a few months.

If you care what the price is after only a few months then you should be going to the casino not the stockmarket. Something like this should be bought for what it will be worth after 10 or 20 years, not next month.

Sure, maybe if you wait you can buy the dip for a 10% or 20% saving. Or maybe not. But that's speculation, not investment.

Right, but statistically if you look at the trend of most IPOs, you'd be better off not buying at day 1, but more like day 30-120.  Looking at how other IPOs perform it's a totally valid strategy.

So hold off, and keep it for 5 years+.   Or buy it early and hope that it turns out to be a good one.   But don't buy it early expecting a deal, because the data is there, it's rare for IPOs to rocket.  Their values are usually a fair representation of the open market value, plus a bit of a premium.

[Stray Electron]

After 40 years of investing, I have concluded IPOs are a loser for the subscriber and a winner for the Corp CFO.  I would wait until it settled down.  Many lose value after 4 or 5 weeks.

I think we're all on the same page with that. I don't think anyone is suggesting to buy in to the IPO.
IPOs are notorious for
- Crashing and burning from their overly inflated starting price before beginning a slow recovery.
- Shooting up from a buying frenzy and then collapsing hard before beginning a slow recovery.

Getting into an IPO at the start is probably the highest risk you can take.

OMG! You guys are talking like Engineers and not like politicians or stock brokers who have something to sell!

Were in an engineering forum
This isn't an investment thread, or it wasn't meant to be. I probably should have been more clear with the thread title.

  Somebody can't recognize sarcasm.

[brucehoult]

After 40 years of investing, I have concluded IPOs are a loser for the subscriber and a winner for the Corp CFO.  I would wait until it settled down.  Many lose value after 4 or 5 weeks.

I think we're all on the same page with that. I don't think anyone is suggesting to buy in to the IPO.
IPOs are notorious for
- Crashing and burning from their overly inflated starting price before beginning a slow recovery.
- Shooting up from a buying frenzy and then collapsing hard before beginning a slow recovery.

Getting into an IPO at the start is probably the highest risk you can take.

OMG! You guys are talking like Engineers and not like politicians or stock brokers who have something to sell!

Were in an engineering forum
This isn't an investment thread, or it wasn't meant to be. I probably should have been more clear with the thread title.

  Somebody can't recognize sarcasm.

We're in an engineering forum.

[ejeffrey]

Yep, astronomers would lose their shit at 1-2km wide solar panels in LEO.  GEO makes more sense if you want to stream data to it directly and makes the size less of an issue if it's that far away.
I guess if you already have crazy amounts of solar power you might be able to give it lots of hall thrusters and move it out into a really high orbit after launch. But perhaps that power is always better used to run GPUs so the cost to put it directly into a high orbit is acceptable.

(Not suggesting it's actually going to happen, just theorizing)

All of this (not just SpaceX, other startups and big companies are claiming they will do this)  is based on -- predicted but as yet unrealized -- starship launch cost of $100-$200 / kg.  That's an LEO target, GEO will still be a multiplier larger. 

But honestly, the whole thing about orbital data centers is just a scam, and I'm pretty sure most of the people promoting it realize that.  It's just a way to goose stock prices by generating hype.

The fundamental problem with the idea it isn't radiation, or cooling, or launch cost, or latency, anything technical.  It's that it doesn't actually do anything.  A datacenter in space is *less* useful for serving requests from terrestrial clients than one on the ground, and it costs more.  The only reason to launch AI processors into space is if you need them in space, for instance as part of an imagine sattelite or space probe, or even a communication satellite.  Just servicing requests from the ground will never make sense.

A big part of this is that people misunderstand the costs with datacenter water and power.  People say things like "water and power are the biggests costs of operating a data center"  -- This is somewhere between a lie and completely misleading.  Water is a negligible cost, and optional: it's used over air cooled chillers because it's incredibly cheap.  Electricity is a major fraction of operating costs, but mostly that's because operating data centers is cheap.

Consider a hypothetical data center unit that is similar in size to the planned Starlink V3.  That's the upcoming version of the highest volume satellite ever built, and one that's been heavily cost optimized.  It is expected to weigh 2000 kg, and it's solar arrays will probably produce around 80 kW of power.  A "data center node" would look a bit different with more power going to compute and less to communications, but a lot of it will be the same.  If you launch this into a sun synchronous terminator orbit so that it has 100% sunlight, it will produce about 3500 megawatt hours over a 5 year lifetime.  At wholesale electricity rates in the US, that's avoiding maybe $150,000 to $200,000 (source: https://www.eia.gov/electricity/monthly/update/wholesale-markets.php, and also unverified reports of what hyperscalers typically pay).  There is simply no way that you can take an equipment rack, redesign it to survive launch and zero g operation, add solar panels, a bunch of communication hardware, navigation thrusters, fuel, etc, and launch it into space for $200,000.  And if more generation capacity is needed, it's cheaper, easier, and faster to deploy solar, wind, and battery storage, and natural gas plants on earth than to launch them into space.  This is doubly true if you build for your own usage and don't have to wait for grid interconnect permissions, which is definitely lagging in some places.

Of course, not everywhere has cheap land and abundant solar and wind capacity.  IF you want to build a data center near new york city, you probably aren't going to have on-site solar power generation.  And your power will be more expensive than I quoted above.  But if you are willing to build your data center *in space* you can put it in western Texas.

Plus, when you build a data center on earth and your computer fails or becomes functionally obsolete after a few years, you pull it out of the rack, shred the hard drives, and send it for recycling.  Then you install a new server in it's place and get to reuse all your infrastructure.  With a satellite you just will watch it burn up on reentry.

[watchmaker]

most IPOs lose money for shareholders after a few months.

. But that's speculation, not investment.

I love it!  But "speculation" and "investing" need to be defined in any such conversation.  Two of those words that sound like they mean something but actually mean different things to different people.  Unlike ohms or volts.

[brucehoult]

most IPOs lose money for shareholders after a few months.

. But that's speculation, not investment.

I love it!  But "speculation" and "investing" need to be defined in any such conversation.  Two of those words that sound like they mean something but actually mean different things to different people.  Unlike ohms or volts.

It's quite simple.

INVESTMENT is based on predictions of what the COMPANY will do, creating new products, the demand for them, ability to scale production, margin between costs and sales, etc. It is based on market trends, news of plans, financial reporting etc -- fundamental analysis.

SPECULATION is based on predictions of what other STOCKHOLDERS will do, will they be happy or sad or panicked, buy or sell. It is based on psychology, fortune tellers and seers, "technical" analysis of short term movements in the stock price. "Head and shoulders", "double/triple top/bottom", "falling/rising wedge (and breakout)" etc).

[Analog Kid]

INVESTMENT is based on predictions of what the COMPANY will do, creating new products, the demand for them, ability to scale production, margin between costs and sales, etc. It is based on market trends, news of plans, financial reporting etc -- fundamental analysis.

But is it really?
Just look at the current "AI" investment: how much of that is based on the usual voodoo attendant upon similar gold rushes?

And I'm not referring to speculators as you defined them, but actual investors, those with real skin in the game.

[McCarthy]

OMG! You guys are talking like Engineers and not like politicians or stock brokers who have something to sell!

After 40 years of investing, I have concluded IPOs are a loser for the subscriber and a winner for the Corp CFO.  I would wait until it settled down.  Many lose value after 4 or 5 weeks.

IPOs in the first few days never have been a proper investment vehicle to begin with, and have been highly speculative, meant for day traders only.

I trade them the moment they hit the exchange, on the 5 min, 1 min and 10 second chart, based on level 2 and the tape only. Buy high, sell higher. Buy the hype and greed, get out before the fear kicks in and sell to the losers. I hold them for seconds only.

[Simmed]

The environment of low Earth orbit also benefits from some shielding from the atmosphere and Earth's magnetic field compared to higher orbits, and low Earth orbit does not pass through the Van Allen radiation belts.

LEO is out of the question for this. No one will be willing to have bazzillion of MASSIVE DATACENTERS along with ultzramassive panel arrays buzzing so low, constantly fighting against atmospheric drag and reboosting.

i think it will become serious
if musk announces he will start creating robot satellites that refuel and repair other satellites
i guess this will truly kick off the next phase, if there is a next phase
robots in space, takes the human out of the equation
no stupid space suits

[mikerj]

If you invited the Chinese in and got out of their way (or a similar length tunnel in China) they could do it for $5-$15 billion.

You just have to hope they are using their most premium chewing gum and sawdust mix to build it with.

[Stray Electron]

i think it will become serious
if musk announces he will start creating robot satellites that refuel and repair other satellites
i guess this will truly kick off the next phase, if there is a next phase
robots in space, takes the human out of the equation
no stupid space suits

   No more space suits is just the beginning of the benefits of not having any humans onboard.  No more food or water to carry or human waste to dispose of,  no need to carry air supply or to filter the recycled air, even to pressurize the vehicle for that matter, and no more having to heat or cool the vehicle (within some limits). So suddenly you've reduced your vehicle weight be roughly one half and also made it smaller. And on top of all of that, your possible mission duration becomes almost unlimited. And then you get into much higher performance areas such as the higher acceleration and de-acceleration rates that become permissible. And finally they wouldn't have to be nearly concerned with flight safety. Not too many people care if you lose a rocket that's carrying satellites or a science experiment but lose a crew of seven astronauts and suddenly the whole world cares!

[UnijunctionTransistor]

You are correct, Stray Electron.

Any tragedy that involves human life will cause a concern several orders of magnitude higher than even the worst accident where only machines are involved.

[Analog Kid]

If you invited the Chinese in and got out of their way (or a similar length tunnel in China) they could do it for $5-$15 billion.

You just have to hope they are using their most premium chewing gum and sawdust mix to build it with.

I object to what you wrote not because it's racist (it kinda is) but because it's so wildly inaccurate.
The Chinese are kicking our ass technologically-speaking, and that only seems to be improving with time.

Your statement might have been believable two or three decades ago.

[Stray Electron]

  The Chinese might be kicking ass technology wise but a lot of their products still look they're made of chicken shit and paper mache. Have you bought any power or hydraulic tools from HF lately and taken them apart and looked at the construction?  I have, a coupe of years ago I found a "choke" in a surge suppressor that was copper wire very crudely wound around a common steel nail!  That's just one example of the poor design and construction that I've found in a lot of Chinese made products.

[McCarthy]

China never learned to properly innovate. To this day they steal and clone. Ask me how I know. Clones of my products show up on Amazon USA regularly, from Chinese sellers, and I have to make a claim with Amazon and US customs to get that shit removed and blocked every single time.

[jheissjr]

What is the link to your Amazon store?