AI infrastructure, data centers, neoclouds, software, fintech.
Miami, FL
Joined May 2025
- Tweets3.3K
- Following780
- Followers17.5K
- Likes7.8K
Pinned Tweet
Can Anthropic Meet Its "$518 Billion of Commitments"? rittenhouseresearch.substack…
It's almost as if the "$40MM per MW!" short-term deals Nebius has announced with these "neolabs" might not be renewed into perpetuity.. lol. Positioning yourself as the "default neocloud to serve the neolabs" is all well and good until "neolab" funding dries up.
“Jane Street depends on raising money” to pay their cloud bill to CoreWeave might be dumbest CoreWeave take I’ve seen all week. And that’s saying an awful lot.
(Also the entirety of the tweet below is false and misleading, lol).
Nebius’s two anchors are cash-rich Microsoft and Meta.
CoreWeave’s growth customers (OpenAI, Anthropic, Jane Street) are private and depend on raising money themselves, so they’re less able to prepay. CoreWeave’s 10-Q says it is adding customers that “may have increased risk of insolvency,” and it carries no credit insurance.
$NBIS $CRWV
🦗 🦗 🦗
George @gnoble79, you say that "pricing on compute and tokens is collapsing at a breathtaking rate".. What data do you have that shows compute prices collapsing? Every chart I see on GPU hourly rental rates is up and to the right.
I agree that token prices are falling, but can you explain how OpenAI and Anthropic's revenue continues to grow at pretty healthy rates if token prices are collapsing?
You also claim that Anthropic is selling dollar bills for 20 cents. What data or numbers are you looking at which suggest that Anthropic has negative gross margins and is spending more on compute than revenue? For reference, I saw in the WSJ this week that for the first 6 months of 2026, Anthropic recorded $16.3 billion in revenue with $9 billion spent on computing capacity and infrastructure. Bloomberg, amongst others, also reported that in Q3-26, Anthropic's revenue was $16.8 billion, with 85% gross margins, "compute margins" of 40%+, and positive free cash flow. I assume gross margins only deduct for inference-serving compute costs, while "compute margin" is further burdened for model training costs. In any event, it would seem that Anthropic's revenue is (significantly) greater than its total compute costs, so I would love to know what you're looking at when you state that Anthropic is selling dollar bills for 20 cents.
Gerstner says CoreWeave (1.5GW active as of Q2) is "subscale" compared to SpaceX (1.7GW active as of Q3).. and that Elon has a lower cost of capital so he can scale faster (sure).
But if CoreWeave is subscale, and access to cheap debt paramount.. Why did he just invest in "Volta", which has a whopping 0MW active today, and just priced secured paper at 11%+ ..?
Why do folks like Steve Eisman and Danny Moses feel the need to just blatantly lie when talking about certain AI-related topics..?
"OpenAI.. Anthropic. These guys are burning cash like crazy".. as Eisman nods along.
It's common knowledge as reported basically everywhere that Anthropic is cash flow positive. Who knows what their financials look like in the future but as things stand today Anthropic is certainly not "burning cash like crazy." Eisman literally is a full time blogger/podcaster who follows this stuff closely. Why gaslight people like this?
NVIDIA's consensus revenue estimate for 2027 is roughly $700 billion. Vera Rubin runs ~$40B per GW. SpaceX is "exclusive to NVIDIA", per Elon.
So when Elon claims SpaceX is going to build 20GW of capacity next year, he's suggesting SpaceX will spend $800 billion on NVIDIA chips.. or greater than 100% of NVIDIA's total revenue, lol.
$SPCX $NVDA
The WSJ out this afternoon claiming to have seen yet another leaked Anthropic S-1.. and that they're breaking news ("many of these disclosures have not yet been reported!") by stating that Anthropic has now disclosed future infrastructure costs north of $600 billion (potentially rising to >$700 billion!).. Hmm..
Can Anthropic Meet Its "$518 Billion of Commitments"? rittenhouseresearch.substack…
Probably not a bad time to bump the project-level returns analysis I ran earlier this year to estimate the unlevered returns on a standard CoreWeave $CRWV Blackwell-era contract..
TLDR: 20% unlevered IRRs with assumptions including:
- 5 yr. contract term
- $27MM/MW in capex
- $12MM/MW in rev. (2.25 yr. rev. payback)
- Utility ($0.075 c/kWh, 85% utilization, 1.4 PUE), co-location ($1.65MM/MW w/ 3% annual escalator), and cust. support expenses in aggregate avg. ~24% of rev. (resulting in ~76% contract-level EBITDA margins)
- Zero prepayment
Published some thoughts on substack (no paywall) around CoreWeave's unit economics, margins, use of debt, and what I think the real risks to the business are
open.substack.com/pub/ritten…
Rittenhouse Research retweeted
oh ok
POD UP! 🚨
Ben Horowitz (@bhorowitz) fills in for @friedberg!
-- CoreWeave's Tanks After Debt Selloff: Did the AI Trade Just Crack? $CRWV
-- AI Deflation and the Future of Software
-- Anthropic IPO: Red Flags, Valuation, How Much to Raise?
-- Psyop Corner with Sacks
-- Bill Gates the Doomer: What’s Behind His AI Warnings?
(0:00) Ben Horowitz fills in for Friedberg!
(2:18) CoreWeave's debt selloff: did the AI bubble just pop?
(31:14) AI's deflationary impact: decompiled software
(45:44) Anthropic's investor day gets mixed reactions
(1:03:45) Psyop Corner: Anthropic warns about self-replicating models on IoT devices
(1:22:19) Understanding Bill Gates the Doomer
Chamath is so clueless, you can't even make it up. Truly basic stuff as well. Dude claims CoreWeave's "GPU Loan DDTL 1.0 spiked to 15%".. completely unaware that the 15% yields all over the headlines this week are the secondary-market implied yields on CoreWeave's unsecured notes..
So is the Nscale IPO at $35B still happening..?
They need to come up with $27.5B for build the data centers and buy the GPUs they've already committed towards.. And that's as of 6/30, before the $45B Anthropic contract was signed.
I suspect it will become pretty clear over time that there's a big difference between modeling out the returns, sources and uses, etc. on these massive GPU deals and actually executing on them.. even just raising the capital required to fulfill existing contracts looks a lot easier on paper than it is in reality.
Saw a lot of "Brad Gerstner Sold All of His CRWV" headlines going around yesterday.. If you actually watch the entire conversation, he comes off to me as pretty positive..
"I don't own any CoreWeave today.. Over the course of the last 3 months we took our exposure down not because of anything specific to CoreWeave but.. because.. AI is overshot. We moved 10% of our book two months ago into Meta.. We had to go somewhere to get that capital..
Let me say a few things about CoreWeave. Number one, the team that's running CoreWeave, one of the best in the business. 90% of the stuff that I read out there is misinformation with respect to CoreWeave.. There's probably opportunity today in CoreWeave in both the equity and the debt. I'm sure my team is back at the shop working away on that and somebody will make the pitch to me..
I think Mike Intrator who runs that business is one of the best. I think their capital stack is good. I think their partnerships like they have with NVIDIA are fantastic. They're going to be one of the new AI hyperscalers that make it to the other side... They went public, I think, at $30 bucks a share. I don't know. The stocks at $60 a share today. It's still a double off the IPO, which occurred sometime in the last two years, right? Not terrible."
$CRWV
The FT's behavior towards CoreWeave is insane to me.. Seems like they're either being paid to trash CRWV or they're just a bunch of bitter, spiteful individuals. The reporting on CRWV's IPO was weird (hit-piece after hit-piece, comparisons to WeWork, etc), and they've clearly been hoping for CRWV to fail ever since.
But the way multiple FT "journalists" have gone on to belittle and mock CoreWeave and its co-founders.. just days after a separate FT "journalist" published an interview with CoreWeave's co-founder is crazy to me. You would never see the WSJ or Bloomberg act like this, really bizarre.
Almost as if they're jealous (which might make sense given the UK's AI-infrastructure trophy business is "Nscale".. which is trying to list in the US to avoid the UK's decrepit capital markets.. How is that process coming btw..?)
Should be pretty evident CoreWeave's cost of capital is declining over any timeframe greater than a few months.. DDTLs (the ABS, project-finance style facilities backed by specific clusters and customer contracts) comprise the majority of CoreWeave's debt.
Pricing on the DDTLs has tightened a lot over the years. From S+9.62% for DDTL 1.0 issued in 2023.. to as low as S+2.25% for DDTL 4.0 done in March 2026. Note how the older, higher-priced DDTLs 1.0 and 2.0 are declining as the facilities are fully self-amortizing.. And new, lower-priced DDTLs are drawn to fund the most recent capex.
Will be interesting to see where the next DDTL prices. Not going to argue it will be tighter than the S+2.25% (DDTL 4.0) and S+4.50% (5.0) prints from earlier this year. But probably worth noting that DDTL 5.5 was done at S+5.50% (not shown below as it was completed post quarter-end) in July 2026 when AI financing markets were similarly volatile.. And DDTL 5.5 has a 5 year term yet funds capex associated with non-investment grade customers on 3 year contracts, making it a much more aggressive structure than anything prior. In any event, where DDTL 6.0 lands seems much more important than where its (illiquid) unsecured senior notes are trading.
$CRWV
Why do people act as if Anthropic invented the concept of "adjusted operating income"..? I mean Salesforce is close to three decades old, approaching $50B+ of revenue, and still reports non-GAAP EBIT. The outrage at Anthropic adding back SBC to present adjusted EBIT like every other tech co. on earth is insane, lol.
