
🫐 Blueberry Capital 🫐
559 posts


@OnodaCapital How many Bitcoin shit cos can fit in a port?
English

@partners_road We are paid to own stocks that go up and short stocks that go down. Not much more than that.
English

@dalibali2 No one understands it man. Need ppl to pound the table on “compute is a strategic advantage”
English

@JaredKubin They are going to do $75Bil+. Still another 2x to go
English


@partners_road Yeah look you can be bearish on the stock just don’t say stuff that’s not true. It’s not cool man
English

Yeah I’m looking at my notes right now, the cogent guy did not give any supply demand run down on AI training/ inference. He actually said compute as a % of total capex will increase. He also said the competitive AI landscape is “great for infrastructure layer of businesses, such as CoreWeave”
English

@blubercap So it happened. The “fossil” was there on a panel with a Corweave guy who explained the s/d? Is that what you’re saying, douchebag? Whether it was today or yesterday, who fines a shit. You’ve shown yourself to be a joke.
English

@partners_road Dude I was in the room. It was yesterday, that guy was a fossil but nonetheless he said the demand was there for CRWV. Idk where you get your information to spew nonsense like this
English

@blubercap It did happen. The Cogent Communications guy who detailed the supply/demand dynamic was on a panel this morning with a guy from Coreweave. Nice try, though.
English

@partners_road River this isn’t gonna normalize for the next 6 years
English

@ShanuMathew93 @tylermacro10 Is anyone earning in the mid or high end? Thabks
English

1 GW AI data center economics are mostly a sensitivity table based on the assumptions you use.
Three revenue cases:
Low: ~$7B
(500k GPUs × 80% utilization × $2/GPU-hour × 8,760 hours)
Mid: ~$17B
(750k GPUs × 85% utilization × $3/GPU-hour × 8,760 hours)
High: ~$32B
(1M GPUs × 90% utilization × $4/GPU-hour × 8,760 hours)
Against ~$50B of capex, revenue payback looks like ~7 years, ~3 years, and ~1.6 years.
That framing is incomplete because the operating cost stack is large:
Power: ~$0.7–1.5B
(1M kW × 80–90% utilization × 8,760 hours × $0.08–0.15/kWh × 1.15–1.25 PUE)
Facilities operations: ~$0.2B
(1,000 MW × ~$175k/MW/year)
IT service and maintenance: ~$1.5–3.0B
(3–6% of $50B capex)
Labor, software, insurance, security, admin: ~$0.2–1.0B
(2–3% of revenue, case-dependent)
Depreciation: ~$6.5–8.0B
($15B GPUs/servers over ~4 years + $7.5B networking over ~6 years + $17.5B power/cooling over ~15 years + $10B shell over ~25 years)
That changes the payback math:
Low case: likely uneconomic. EBIT is minimal or negative after full cost burden.
Mid case: roughly ~8–10 years to EBIT payback.
High case: roughly ~3 years to EBIT payback, but only if GPU density, utilization, and rental pricing all hold near the bull-case end of the range.
The key variables have to include other factors beyond just power cost. They are GPU count per GW, realized rental price, sustained utilization, service burden, depreciation life, and refresh economics.
The main question is what happens in year five onward. A cluster still earning $3/GPU-hour has a very different return profile than one earning $0.50/GPU-hour after the next hardware cycle. We’ve seen demand and rental prices hold up, if they compress for whatever reason, the math changes very quick.
David Sacks@DavidSacks
Back-of-envelope numbers for 1 gigawatt data center: All-in Capex: ~$50 bn Enterprise revenue generated: ~$25-30 bn/year Electricity cost: $1-2 bn/year ~2 year payback. The boom is real.
English

@MetacriticCap @LongOnlyLarry Exactly. Plus they would probably get a higher multiple
English

@LongOnlyLarry @blubercap Automatic Data Processing is an industrial.
What is AGI if not Automatic Data Processing??
English

@qcapital2020 They are going to do a trillion in ebit over the next few years
English

@ScroogeCap @RealNickMugalli Scrooge how many Blackwell and Rubins does xAI have?
English

This is a catastrophic signal for neoclouds like $CRWV and $NBIS.
These companies exist because compute was scarce. When someone like SpaceX enters the rental market with 220,000 GPUs at once, they become the 800lb gorilla.
SpaceX has the advantage of vertical integration and they can underprice neoclouds into oblivion just to keep their utilization rates at 100%.
Anthropic was a crown jewel customer for specialized clouds. If they are moving their heavy lifting to SpaceX then the moat for neoclouds just evaporated.
Now regarding xAI:
Colossus 2 is supposed to bring them toward 1 million GPUs, but if they can't even utilize the first 200k for a winning internal model, the Gigafab becomes a massive, expensive monument to overcapacity for them.
If xAI had a model capable of leapfrogging GPT-5 or Claude 4, they would be using every single one of those 220,000 H100s/GB200s to train it. You don’t let a direct competitor take over your primary training ground unless:
- xAI’s research might have hit a wall where throwing more compute at Grok isn't yielding proportional returns.
- Maintaining a 300MW facility with 200k+ GPUs costs billions in power and debt service. If they can’t justify the internal training run right now, they have to rent it out to stop the bleeding.
- Musk will clean up the books for a SpaceX IPO. Renting out the hardware makes SpaceX a high-margin infra provider rather than a high-risk research lab.
$TSLA
Claude@claudeai
We’ve agreed to a partnership with @SpaceX that will substantially increase our compute capacity. This, along with our other recent compute deals, means that we’ve been able to increase our usage limits for Claude Code and the Claude API.
English










