Paul | Bitcoin, Crypto & Equities | Macro | Charts & On-Chain | NFA | 🇨🇦
Ontario, Canada
Joined November 2025
- Tweets44.7K
- Following689
- Followers3.7K
- Likes236K
Not Another Quant retweeted
$SOFI is making more profit from their revenue but they raised so much new equity that returns on that equity haven’t improved as much. Their CFO said they still have a lot of extra capital to grow with. To reach their ROTCE target of 20% to 30% they need to put it to work and generate much more revenue from the capital they have.
ROTCE went up to 9.9% in Q3 which is the highest it has been and most of that came from margins.
But after selling about $3.3B of new stock each $1 of equity brings in 56 cents of revenue a year compared to 68 cents before.
To hit their target, their margins have to keep going up and that 56 cents has to get a lot closer to $1.
They plan to do that by putting the extra capital into more loans. Making more loans for partners also helps because they bring in fees without using much of their capital.
But keeping more loans also means more risk from rates. This quarter their loans lost over $800M in value mostly from higher rates and their hedges made back about $734M of it.
Not Another Quant retweeted
$SOFI Q3 earnings are out and about 90% of their revenue comes from lending and banking like it did in Q2.
Lending and financial services both grew 13% from Q2 so the split stayed about the same.
The biggest change in their fee revenue is origination fees which went from $198.5M to $270.9M. 74% of their fee revenue now comes from loans compared to 72% in Q2.
Out of every $100 (adjusted net revenue $1,371.1M):
$59 lending ($807.4M)
$45 financial services and tech platform ($525.1M + $87.4M = $612.5M)
minus $4 corporate items (−$48.7M)
That $45 is:
$21 net interest income from deposits and credit cards ($281.3M)
$11 loans made or referred for other lenders and investors ($151.2M)
$13 card fees, brokerage, tech and other ($180.0M)
$59 + $21 + $11 = $91 so about $90 of every $100 ($1,239.8M) is lending or banking.
Lending share: 51% a year ago ($481.4M of $949.6M), 59% in Q2, 59% now.
Looking at the $SOFI chart we need it to clear that Sept $18 level and the 200 day EMA. Then clear $20 and hold to stay away from goblin town we just pumped out of.
More SoFi analysis to come! Two recent posts linked in the replies.
Post on ROTCE:
xitter.cf/notanotherquant/status…
$SOFI is making more profit from their revenue but they raised so much new equity that returns on that equity haven’t improved as much. Their CFO said they still have a lot of extra capital to grow with. To reach their ROTCE target of 20% to 30% they need to put it to work and generate much more revenue from the capital they have.
ROTCE went up to 9.9% in Q3 which is the highest it has been and most of that came from margins.
But after selling about $3.3B of new stock each $1 of equity brings in 56 cents of revenue a year compared to 68 cents before.
To hit their target, their margins have to keep going up and that 56 cents has to get a lot closer to $1.
They plan to do that by putting the extra capital into more loans. Making more loans for partners also helps because they bring in fees without using much of their capital.
But keeping more loans also means more risk from rates. This quarter their loans lost over $800M in value mostly from higher rates and their hedges made back about $734M of it.
SoFi’s revenue breakdown:
xitter.cf/notanotherquant/status…
$SOFI Q3 earnings are out and about 90% of their revenue comes from lending and banking like it did in Q2.
Lending and financial services both grew 13% from Q2 so the split stayed about the same.
The biggest change in their fee revenue is origination fees which went from $198.5M to $270.9M. 74% of their fee revenue now comes from loans compared to 72% in Q2.
Out of every $100 (adjusted net revenue $1,371.1M):
$59 lending ($807.4M)
$45 financial services and tech platform ($525.1M + $87.4M = $612.5M)
minus $4 corporate items (−$48.7M)
That $45 is:
$21 net interest income from deposits and credit cards ($281.3M)
$11 loans made or referred for other lenders and investors ($151.2M)
$13 card fees, brokerage, tech and other ($180.0M)
$59 + $21 + $11 = $91 so about $90 of every $100 ($1,239.8M) is lending or banking.
Lending share: 51% a year ago ($481.4M of $949.6M), 59% in Q2, 59% now.
$SOFI is making more profit from their revenue but they raised so much new equity that returns on that equity haven’t improved as much. Their CFO said they still have a lot of extra capital to grow with. To reach their ROTCE target of 20% to 30% they need to put it to work and generate much more revenue from the capital they have.
ROTCE went up to 9.9% in Q3 which is the highest it has been and most of that came from margins.
But after selling about $3.3B of new stock each $1 of equity brings in 56 cents of revenue a year compared to 68 cents before.
To hit their target, their margins have to keep going up and that 56 cents has to get a lot closer to $1.
They plan to do that by putting the extra capital into more loans. Making more loans for partners also helps because they bring in fees without using much of their capital.
But keeping more loans also means more risk from rates. This quarter their loans lost over $800M in value mostly from higher rates and their hedges made back about $734M of it.
$SOFI Q3 earnings are out and about 90% of their revenue comes from lending and banking like it did in Q2.
Lending and financial services both grew 13% from Q2 so the split stayed about the same.
The biggest change in their fee revenue is origination fees which went from $198.5M to $270.9M. 74% of their fee revenue now comes from loans compared to 72% in Q2.
Out of every $100 (adjusted net revenue $1,371.1M):
$59 lending ($807.4M)
$45 financial services and tech platform ($525.1M + $87.4M = $612.5M)
minus $4 corporate items (−$48.7M)
That $45 is:
$21 net interest income from deposits and credit cards ($281.3M)
$11 loans made or referred for other lenders and investors ($151.2M)
$13 card fees, brokerage, tech and other ($180.0M)
$59 + $21 + $11 = $91 so about $90 of every $100 ($1,239.8M) is lending or banking.
Lending share: 51% a year ago ($481.4M of $949.6M), 59% in Q2, 59% now.
Q2 comparison for $SOFI
xitter.cf/notanotherquant/status…
I looked at $SOFI through bank measures yesterday because that's where its money comes from with 91% of its revenue last quarter coming from lending or banking.
Out of every $100 SoFi brought in:
$59 came from lending
$21 was net interest income mostly from deposits
$12 came from loans made or referred for other lenders and investors
$13 came from card fees, brokerage, tech and other
Minus $5 of corporate items
$59 + $21 + $12 = $92 but that's rounded up. The exact figure is $91.
SoFi also measures itself this way. Its own long term target is a 20% to 30% return on tangible equity and it was 7.1% last quarter.
What makes SoFi interesting is the products outside lending. 87% of its 24.4 million products aren't loans.
On average a loan still brings in about 10 times the revenue of one of those products.
But more new products are also coming from existing members. 51% last quarter compared to 43% the quarter before.
Five years from now do you think most of SoFi's revenue still comes from lending and banking?
Not Another Quant retweeted
The dollar’s rally has stalled at its 200 week MA. The weak euro is what is mainly keeping the dollar up.
France has a lot of debt and can’t agree on a plan to cut its deficit. A lot of hedge funds had borrowed to buy French bonds for the extra yield. When the budget news got worse and oil went back above $100 it looks like they got rinsed.
When people are worrying about France’s debt they start worrying about the euro. It looks like some of the money coming out of French bonds is leaving the euro completely. The euro is ~58% of DXY so a weak euro helps hold DXY up.
Moody’s didn’t cut France’s rating on Friday and the selling has stopped for now. But the euro could stay weak until France shows it can actually cut its deficit.
The next vote on France’s budget is November 3.
Not Another Quant retweeted
With Anthropic filing their S1 today it looks even more likely they are the customer behind 1/3 of $SPCX Q3 revenue.
In the S1 Anthropic says they’ll spend up to $120.5B with SpaceX into 2029. That’s up from about $45B when SpaceX filed to go public in May and $84.5B in the draft Reuters saw in September.
They expect all their capacity to be up and running in December and then pay SpaceX about $3.1B a month. The customer SpaceX didn’t name made up about $5.0B of their Q3 revenue which is about $1.7B a month. That would make sense if Anthropic was still getting more capacity added in Q3.
There’s also the $1.11B a month deal SpaceX said would start December 1. They never said who it was with and I think it could be Anthropic since that’s when Anthropic expects to get all their capacity.
We can’t confirm Anthropic is the customer yet because the S1 doesn’t show what they paid SpaceX in Q3. But I think the deal makes them too big to be anyone else in SpaceX’s numbers.
The new $SPCX 10-Q shows one customer SpaceX doesn’t name made up 1/3 of their revenue in Q3. That’s about $5.0B and it’s 2/3 of their AI revenue.
In Q2 the same customer was about $1.5B so it more than tripled in one quarter.
It looks like it could be Anthropic. Reuters reported Anthropic’s contracts with SpaceX are worth up to $84.5B into 2029 which is nearly double the original deal.
We don’t have that confirmed but if Anthropic files publicly soon then the S1 should show their SpaceX contracts.
Not Another Quant retweeted
If you're wondering why $META is red today:
$35B stock and preferred offering announced.
Up to $60B more planned. ~$95B total.
$META didn’t wait for the new Class C shares to start selling stock to pay for their AI buildout. They just announced $35B in stock offerings today and plan to sell up to $60B more once the Class C shares start trading next year.
Their filing last week said it’s getting harder to raise money because so many other AI companies need it. It looks like they want to get theirs before there is less left.
These Class C shares have no votes. I have posted a lot on this and the quote thread is linked below. This first sale can’t be Class C yet but Meta set it up so most of it can end up as Class C later. That helps protect Zuckerberg’s voting power.
$NVDA $AMD and $AVGO might buy up to $5B of it between them. Meta buys chips from all three so some of this money could go right back to them.
Not Another Quant retweeted
$META didn’t wait for the new Class C shares to start selling stock to pay for their AI buildout. They just announced $35B in stock offerings today and plan to sell up to $60B more once the Class C shares start trading next year.
Their filing last week said it’s getting harder to raise money because so many other AI companies need it. It looks like they want to get theirs before there is less left.
These Class C shares have no votes. I have posted a lot on this and the quote thread is linked below. This first sale can’t be Class C yet but Meta set it up so most of it can end up as Class C later. That helps protect Zuckerberg’s voting power.
$NVDA $AMD and $AVGO might buy up to $5B of it between them. Meta buys chips from all three so some of this money could go right back to them.
The dollar’s rally has stalled at its 200 week MA. The weak euro is what is mainly keeping the dollar up.
France has a lot of debt and can’t agree on a plan to cut its deficit. A lot of hedge funds had borrowed to buy French bonds for the extra yield. When the budget news got worse and oil went back above $100 it looks like they got rinsed.
When people are worrying about France’s debt they start worrying about the euro. It looks like some of the money coming out of French bonds is leaving the euro completely. The euro is ~58% of DXY so a weak euro helps hold DXY up.
Moody’s didn’t cut France’s rating on Friday and the selling has stopped for now. But the euro could stay weak until France shows it can actually cut its deficit.
The next vote on France’s budget is November 3.
Not Another Quant retweeted
Everyone who miss following @notanotherquant in this early phase is gonna regret it later. His posts are gems, and he always engages with his mutuals. The next bear market is gonna make him a 100k account when everyone will look for real asnwers during dark times.
Not Another Quant retweeted
Phenomenal read.
It’s just different.
I find accounts who give great value and can tell they are working hard to put out alpha.
From the start that’s all my dawg has done.
Quality matters in the neocloud space. I think lumping them all together is a mistake. Neoclouds can still get loans but it’s getting harder and how much they pay comes from how risky they look.
Lenders care a lot about how strong customers are and how long those customers are signed up for since that’s the money that is paying back the loan. How new the company is and how much debt it already has are obviously key as well.
Musk is arguing lenders have mostly stopped lending to companies that don’t bring in a lot of revenue outside AI data centers and he uses $SPCX as the example.
They have Starlink and launch revenue and last week it cost about 2.3% a year to insure their debt against default. $CRWV makes almost all their money from AI data centers and it cost about 11% a year to insure their debt.
But it doesn’t look like you need outside revenue to borrow yet. $NBIS makes almost all their money from AI cloud. They have customers like Microsoft and Meta and on Oct 17 during this selloff they got a one year $3.5B loan from Citi starting at about 6.7%.
Compare this with Volta which is a new company and got a loan at about 10.5% two days earlier. They had to pay more than first planned to get it sold.
Not Another Quant retweeted
Quality matters in the neocloud space. I think lumping them all together is a mistake. Neoclouds can still get loans but it’s getting harder and how much they pay comes from how risky they look.
Lenders care a lot about how strong customers are and how long those customers are signed up for since that’s the money that is paying back the loan. How new the company is and how much debt it already has are obviously key as well.
Musk is arguing lenders have mostly stopped lending to companies that don’t bring in a lot of revenue outside AI data centers and he uses $SPCX as the example.
They have Starlink and launch revenue and last week it cost about 2.3% a year to insure their debt against default. $CRWV makes almost all their money from AI data centers and it cost about 11% a year to insure their debt.
But it doesn’t look like you need outside revenue to borrow yet. $NBIS makes almost all their money from AI cloud. They have customers like Microsoft and Meta and on Oct 17 during this selloff they got a one year $3.5B loan from Citi starting at about 6.7%.
Compare this with Volta which is a new company and got a loan at about 10.5% two days earlier. They had to pay more than first planned to get it sold.
With Anthropic filing their S1 today it looks even more likely they are the customer behind 1/3 of $SPCX Q3 revenue.
In the S1 Anthropic says they’ll spend up to $120.5B with SpaceX into 2029. That’s up from about $45B when SpaceX filed to go public in May and $84.5B in the draft Reuters saw in September.
They expect all their capacity to be up and running in December and then pay SpaceX about $3.1B a month. The customer SpaceX didn’t name made up about $5.0B of their Q3 revenue which is about $1.7B a month. That would make sense if Anthropic was still getting more capacity added in Q3.
There’s also the $1.11B a month deal SpaceX said would start December 1. They never said who it was with and I think it could be Anthropic since that’s when Anthropic expects to get all their capacity.
We can’t confirm Anthropic is the customer yet because the S1 doesn’t show what they paid SpaceX in Q3. But I think the deal makes them too big to be anyone else in SpaceX’s numbers.
The new $SPCX 10-Q shows one customer SpaceX doesn’t name made up 1/3 of their revenue in Q3. That’s about $5.0B and it’s 2/3 of their AI revenue.
In Q2 the same customer was about $1.5B so it more than tripled in one quarter.
It looks like it could be Anthropic. Reuters reported Anthropic’s contracts with SpaceX are worth up to $84.5B into 2029 which is nearly double the original deal.
We don’t have that confirmed but if Anthropic files publicly soon then the S1 should show their SpaceX contracts.
$META didn’t wait for the new Class C shares to start selling stock to pay for their AI buildout. They just announced $35B in stock offerings today and plan to sell up to $60B more once the Class C shares start trading next year.
Their filing last week said it’s getting harder to raise money because so many other AI companies need it. It looks like they want to get theirs before there is less left.
These Class C shares have no votes. I have posted a lot on this and the quote thread is linked below. This first sale can’t be Class C yet but Meta set it up so most of it can end up as Class C later. That helps protect Zuckerberg’s voting power.
$NVDA $AMD and $AVGO might buy up to $5B of it between them. Meta buys chips from all three so some of this money could go right back to them.
My quote thread on $META xitter.cf/notanotherquant/status…
$META new non voting shares let Zuckerberg keep control while Meta can raise money and he can sell some stock.
Looking more into the filing Meta’s board members did push back on the size of the share dividend. Zuckerberg wanted 20 new shares for every share. The board offered 1 for every 9 and they ended up at 1 for 1. More new shares would mean more stock he could sell without giving up votes.
But they didn’t get any limits on his control. In 2016 Facebook’s board made him agree that in most cases he would lose control if he left the company. This time he said no to that kind of rule and they approved it anyway.
Last week a Delaware court blocked a similar deal at $NET because Cloudflare’s board members likely went along with what their founders wanted. Meta’s board also went along with Zuckerberg on control but nobody has sued yet so we don’t know how a court would rule.
Quality matters in the neocloud space. I think lumping them all together is a mistake. Neoclouds can still get loans but it’s getting harder and how much they pay comes from how risky they look.
Lenders care a lot about how strong customers are and how long those customers are signed up for since that’s the money that is paying back the loan. How new the company is and how much debt it already has are obviously key as well.
Musk is arguing lenders have mostly stopped lending to companies that don’t bring in a lot of revenue outside AI data centers and he uses $SPCX as the example.
They have Starlink and launch revenue and last week it cost about 2.3% a year to insure their debt against default. $CRWV makes almost all their money from AI data centers and it cost about 11% a year to insure their debt.
But it doesn’t look like you need outside revenue to borrow yet. $NBIS makes almost all their money from AI cloud. They have customers like Microsoft and Meta and on Oct 17 during this selloff they got a one year $3.5B loan from Citi starting at about 6.7%.
Compare this with Volta which is a new company and got a loan at about 10.5% two days earlier. They had to pay more than first planned to get it sold.