hckrnws

But while this is a "good problem to have" it would have been an even better problem to avoid in the first place, because it seemed avoidable.

Now, I'm totally armchair billionaire-CEO-ing here, but anybody with any compute has been so obviously capacity constrained for so many quarters all the while scrambling like mad and spending obscene amounts of money to acquire even more compute. With lead times of 2 - 3 years, something Dario explicitly called out on Dwarkesh, it seemed prudent to acquire first, ask questions later. Worst case, they could have rented any extra capacity out, like Elon is doing!

Outsiders are reasonably questioning this mania but Dario, as one of the biggest believers in AI and even AGI, showing hesitancy seems uncharacteristic. I wonder if this is one of those rare cases where it would have been better to drink his own Kool Aid!

Anthropic got somewhat lucky that Elon wanted to stick it to Altman, but boy, even then he drove a hard bargain.

> or is cheaper to replace than it was to buy

will also hold true for cost of mass to orbit. There's a lot riding on making that prediction come true for SpaceX, hence all the CapEx going into Starship.

Tesla seems in a world of hurt unless robots start making space centers from moon rocks, yet that is also defying gravity.

Right now, quite a few companies are discovering that the can turn inference capacity into revenue. Anthropic also can turn inference capacity into happy customers and mindshare, and they can turn lack of inference capacity into sad customers that might jump ship to OpenAI. And Anthropic wants to IPO, and they want to be as close to #1 as possible. And this whole phenomenon, industry-wide, has caused the demand for fancy chips to outstrip supply. Two years ago, DRAM was a low-margin industry, and now it’s not. If you bought a 5090 when it came out for around MSRP, you could resell it now at a healthy profit.

xAI appears to have effectively resold their datacenter at a healthy profit.

Sure, maybe xAI will try to bet that they can build another datacenter and sell/lease it at a healthy profit, but lots of players are trying to make that bet (bottlenecked by power and chip availability). And those bets could easily fail. And the players who don’t have adequate competition (SK Hynix, Micron, TSMC, etc) are going to jack up prices to try to capture more of the upside. And players like DeepSeek and Alibaba want to drive down the need for FLOPs and DRAM, because they don’t have enough and because they have a shortage of those but they don’t have a shortage of excellent AI development talent.

Oh, and China will build its datacenters on the ground, backed by more solar capacity than SpaceX can even dream of launching, and those datacenters will compete. And CXMT and Huawei will do everything in their power to ramp their own production, and SpaceX is not about to get first dibs.

On the bright side, Tesla’s AI5 finally taped out, and SpaceX will surely get some of those.

So maybe SpaceX will find $20bn of GPUs that they can resell or lease for $40bn of discounted revenue, but they could just as easily not find those GPUs or they might only get $17bn of discounted revenue and lose money on the whole affair.

Perhaps the plan all along was to build enough rockets to blot out the sun over large parts of the Earth so only the space Chads can grow crops.

You don't even have to read the document yourself.

Plenty of people are summarising it themselves and using AI.

The loss from Xai is over 7B

Re your first statement, the problem is that there isn't enough compute out there. xAI built their own data centers (and plan to built more -> in orbit). I don't think Anthropic has done that to the same extent and it seems like they will partner with multiple vendors who can provide the compute they need.

"in May 2026, we entered into Cloud Services Agreements with Anthropic, an AI research and development public benefit corporation, with respect to access to compute capacity across COLOSSUS and COLOSSUS II. Pursuant to these agreements, the customer has agreed to pay us $1.25 billion per month through May 2029, with capacity ramping in May and June 2026 at a reduced fee. The agreements may be terminated by either party upon 90 days’ notice."

I do understand it provides stability for longer term leadership which is good, ideally there would be a sunset provision which converted the supervoting shares to normal shares after a certain amount of time.

And Starship primary mission is next generation Starlink, not NASA.

> the debris field of two Starships breaking up in medium orbit could wipe out significant parts of the starlink constellation

Do you have any evidence of that?

NASA wouldn't allow operations that could potentially destroy most sats around the world in any plan it signs.

I'll keep the below for integrity sake:

Well, i'm sure SpaceX bought Xai using some kind of prefered share/debt financing, but that's not to say that XAI had the original debt financing.

We can never know what the exact details, and the exact financing is on this debt. Maybe it's tied to Elon's Tesla Shares, Maybe it's tied to a convertible, maybe it is actual "loans" from a bank. Even at $9b in debt, and you naivly assume they are paying 10% (Def not 20% as OP claimed), you are paying $900m a year, for the entirety of xAi. Including that in the calculations to rent out the entire compute is folly. Not only is 900M not directly attributed to c1, cause it's split between c1, c2 and all the training runs, but you can never verify the interest. And even then, one month of this deal pays for the whole year of interest expense.

So go ahead and lower my estimate by 10%... doesn't make a difference.

Essentially, they are using most of C2 for Grok5. That training run is coming to an End, and they will be leasing more capacity. So taht 1.25b per month will go up to around $1.5B-$2B per month as they finish grok 5.

Edit: from cofounder of Anthropic: "will be scaling up on GB200 capacity in Colossus 2 throughout June." https://x.com/nottombrown/status/2057194829986300375

So the 1.25B is for c1, and the revenue will scale into c2. No idea how much scale, but c2 is almost double the compute? So potentially $3b a month, but probably closer to $2.5B since they get a discount.

https://archive.is/aKFYZ

You can quibble about the exact numbers, but I think it’s fairly clear at this point that inference is profitable with decent margins. Like you say, unit economics are more interesting than the profitability of the company as a whole.

The game is rigged. Our pensions and 401Ks are being forced to buy this slop. They can't resist taking all of the money off the table.

It was hemorrhaging in many cities using extremely profitable cities like London and NYC to keep their global competitiveness.

Uber was able to pivot and become financially sound with two moves: - Uber Eats becoming first party delivery to restaurants (it started as a limited selection of items from some restaurants and quickly evolved into a Doordash-esque competitor) - Uber launching a 1B+ RR Ads business - margins on this are obviously incredible

Both of those combined with discipline in their ridesharing business (exiting the China market with a sale + stake, dumping their self-driving business when it became a money sink) have led to a recovery in their stock price, but it is FAR from the crazy expectations set up for VCs. I expect those in the last round didn't get a great return, but obviously folk like Benchmark exited like kings.

Uber successfully displaced most of the alternatives, slowly raised rates, and maintained operating margin while their fixed costs didn't have to scale as much. Post Travis they've, financially, nailed it.

Tesla is mentionned 87 times. And even they explicitly state that they have strategic collaboration. They supposedly biggest projects are developped together (Macrohard, Terafab - all fugazzi, but still).

Related party transactions are close to a $1 Billion in 2025.

But you are right to call out the launch infrastructure as the true bottleneck. They have 3 pads currently under development. So in 6-9 months they'll have 4 operational pads.

Also, how do they heat tiles hold up? How fast can they catch, refurbish and relaunch is what remains.

I'm confident, and will be putting my money where my mouth is (By investing in the IPO) that they will have useful orbital payloads this year.

>you could already build a normal datacenter somewhere

Can you? from where i'm sitting it's pretty illegal to build a datacenter in the united states. There has literally been hundreds of projects to build data centers around the US that have gotten canceled due to NIMBY's. Collosus itself is plagued by lawsuits.

All those things you listed aren't deal breakers, and maybe some earth based compute will be used but we simply don't have enough power, political will or the legal system to be able to push these data centers through.

So despite the fact that it is theoretically "Cheaper" to build a ground based system, and it will take 261 starship launches just for a single GW... The math is still mathin'. $33B to make $180B.