Antropics S1 warning that advanced AI could pose catastrophic risks to humanity mmmmmm
So you want my money to develop something you say could kill us all, so you can make sure it doesn’t kill us all.
I thought this was the investment pitch for the biggest IPO in history, not a fucking protection racket.
You can’t make this shit up.
⭐️ Meta seeks $95bn for AI.
⭐️ CCC junk yields hit 18%.
⭐️ OpenAI cuts val $150bn; Anthropic S-1 today.
⭐️ Musk secures GPU debt; sees neocloud wipeout.
⭐️ China AI bubble pops: $480bn gone.
⭐️ Brazil FX & stocks plunge.
⭐️ US borrows $183bn this week.
AI has made building hobby software ridiculously easy, and it’s convinced people that writing code is the same as building a company.
Everyone has the same tools now, so code alone is no longer a moat.
What still determines success is solving a problem people will pay for, distribution, economics that work and raising the fuel needed.
Then there’s legal, finance, IP, HR and a shitload of other boring stuff that still needs doing to build and run a company. All the stuff that eats into a founder’s time, pulling them away from customers and product. Ignore it and you’ll fail.
That was always the hard part. It still is.
Startups with AI are going after your customers right now. And you don't even know they exist.
AI doing their market research, finding prospects, qualifying leads, managing outreach, tracking the pipeline and running day-to-day operations, learning and improving as it goes.
Leaving them free to build a better product, talk to customers and close deals.
No departments. No layers of management. A fraction of your costs.
And you're the business they're coming after.
$50. That’s a successful break-in now, using AI agents. One bloke, under 100 characters of instructions, 1,200 targets, 100 hits.
Same week AI agents turned up inside accounts at JPMorgan, Morgan Stanley, BofA and Wells doing things nobody asked for.
We handed software the keys and made it cheap to try every lock. Black October.
open.substack.com/pub/derekw…
I reckon we're massively underestimating what happens when a founder can access the research, analysis, technical expertise and execution capabilities of an entire team for a fraction of the cost and in a fraction of the time.
Productivity on steroids.
It's going to fundamentally change the economics of building a business.
And I don't think the startup industry really gets it.
always a grea tleading indicator. Credit markets know
🦔The cost of insuring Meta’s debt against default hit a record today, and its bonds fell, after Meta said it has now agreed to about $825 billion in future leases and contracts, mostly for AI computing. That figure was close to $700 billion in July. Meta’s free cash flow also turned negative last quarter, at minus $2.1 billion, compared with plus $10.6 billion a year earlier. The stock dropped almost 6%.
My Take
Meta’s default insurance costs about 119 basis points a year, well below Oracle near 290 or CoreWeave near 1,200 last week, so lenders still see Meta as a safe borrower. I’m a bit rattled by the speed though. Meta piled on roughly $125 billion of new commitments in three months, and its ad business, one of the best cash machines in corporate America, no longer covers the spending by itself. If Meta has to borrow to keep up, I can’t think of an AI spender that won’t.
Bond buyers are already looking at a wall of AI debt, with SpaceX shopping a $40 billion deal and Broadcom lining up $60 billion of chip financing for Anthropic. Every new deal has to pay a little more to get noticed. We'll probably see Meta sell a large bond before year end at a higher rate than it would have paid in the spring. It raised the low end of its spending forecast for this year and wouldn’t give one for next year, so free cash flow probably stays negative for a while.
Hedgie🤗
Tomorrow, Friday 23 October: @Anthropic could publicly release its S-1, according to Yahoo Finance, citing three people familiar with the plans.
Where will the IPO price?
60%above $1.5t
40%below $1.5t
15 votes • 2 daysTHE AI FUNDING PROBLEM IS SPREADING. AND FRANCE IS GETTING HAMMERED.
⭐️ Nasdaq -2.2%. VIX above 19.
⭐️ China has entered a bear market.
⭐️ Meta's credit risk hits a record as AI spending keeps climbing.
⭐️ CoreWeave -7%. SoftBank -6.9%. AI financing is getting really ugly.
⭐️ France–Germany bond spreads blow out to euro crisis levels.
⭐️ UK debt passes £3tn as inflation and borrowing costs rise.
⭐️ Brent near $112 as the Iran war escalates again.
⭐️ Tesla looks like a sack of shit with no more hype or promises apart from mreging into a bigger hype story.
Yesterday I said $VIX was cheap as chips at 16.6 given what was happening underneath.
Well, there you go.
📉 Stocks: Yesterday's Bounce Is Gone
US intraday:
S&P 500: 7,652, -1.35%
Nasdaq: 26,122, -2.18%
Dow: 52,552, -0.58%
Russell 2000: -0.50% VIX: 19.02, +14.0%
The AI names are getting hammered:
$META -6.5%
$CRWV -6.9%
$IREN -8.8%
$SPCX -5.4%
$NVDA -3.5%
$LRCX -9.0%
$AMAT -5.7%
$TSLA is roughly flat after recovering from a sharp pre-market sell-off.
The interesting part is that the more capital-intensive AI businesses are taking some of the biggest hits.
🇨🇳 China: Now Officially in a Bear Market
China's CSI 300 has fallen around 20% from its June peak, entering bear-market territory.
And the technology sector has been hit considerably harder.
Today's Asian closes:
Hang Seng -3.03%
Nikkei -2.65%
Kospi -3.52%
Shanghai Composite -1.48%
China's weak domestic demand, disappointing corporate earnings and diminishing confidence in stimulus are weighing heavily on equities.
Meanwhile, DeepSeek has reportedly raised $15bn, backed by investors including Tencent and CATL.
So even with China's equity market in a bear market, the AI investment race continues.
Which raises the same question facing US technology investors.
How much capital gets consumed before the returns justify it?
🤖 AI: Meta Spends More. Tesla Spends Less.
$META
Revenue rose 26% to $64.6bn.
But free cash flow was negative $2.12bn in the quarter.
Annual capital spending is now guided at $135bn–$145bn, with the lower end increased by $5bn.
Meta's five-year credit default swap spread has jumped to a record 114.5bp.
The company is also introducing a new class of non-voting shares, partly to give itself greater financing flexibility.
There was a substantial $9.9bn legal charge in the results, so the reported operating profit decline isn't simply an AI problem.
But the cash being consumed is real.
$TSLA
Tesla's results were hardly inspiring.
Revenue: $28.05bn, essentially flat. Operating profit: $718m, -56%.
Net profit: $787m, -43%.
Operating margin: 2.6%.
Free cash flow: +$1.53bn.
Tesla cut its 2026 capital spending forecast from more than $25bn to around $20bn, raising questions about the pace of its AI and robotics investment.
And Musk is exploring a potential merger with SpaceX.
Because apparently Tesla's existing business isn't complicated enough and he doesn't have enough control or anywhere to hide all the failings.
So Meta is spending more than investors like, while Tesla is spending less than investors expected.
Different problems. Same underlying question.
Where are the returns?
💣 AI Credit: This Is Where It Gets Interesting
JPMorgan's high-yield AI credit basket is reportedly at record-wide spreads, 97bp beyond its previous July peak.
CoreWeave remains at the centre of the problem.
Some of its unsecured bonds have recently traded at yields of 15–16%+.
Oracle's credit spreads have widened substantially relative to other investment-grade hyperscalers.
Goldman reports that hedge funds have been aggressively selling global technology stocks at the fastest pace in a decade.
And yesterday we discussed S&P's warning about AI liabilities sitting in leases, SPVs, guarantees and financing structures.
Big Tech also has an estimated $1.09tn of future data-centre lease commitments, while extending depreciation assumptions on equipment that may become obsolete faster.
The problem is pretty obvious.
More capital spending. More borrowing. Higher funding costs. And potentially optimistic accounting assumptions.
That's not a particularly attractive combination.
🇯🇵 SoftBank Is Getting Caught in the Middle
$9984 closed -6.9% at ¥4,305, more than 50% below its June peak.
SoftBank borrowed another $11.1bn in September, including dollar bonds paying coupons as high as 9.75%, to finance its AI ambitions.
Its cumulative investment in OpenAI has reached $64.6bn.
And the cost of insuring SoftBank debt against default has climbed to its highest in almost four years.
The AI boom isn't running out of ambition. It's discovering how expensive that ambition is to finance.
🇫🇷 France: The Bond Market Is Getting Seriously Ugly
Yesterday France's 10-year yield crossed 5%, and the spread over German Bunds was around 150–157bp.
Today:
French 10Y: ~5.11% France–Germany 10Y spread: ~174bp
The widest since 2011, during the eurozone sovereign debt crisis.
And here's the problem.
US Treasuries and German Bunds are rallying as investors seek safety.
French borrowing costs are moving in the opposite direction.
With public debt around 119% of GDP, a deficit above 5%, weak growth and political paralysis, investors are demanding an increasingly large premium to lend to France.
Moody's reviews France's Aa3 credit rating tomorrow, with the outlook already negative.
That's a proper near-term risk.
🇬🇧 Britain: £3 Trillion and Counting
Yesterday's UK inflation and borrowing numbers were pretty bloody ugly.
CPI inflation: 3.5%, up from 3.1%.
Producer input inflation: 8.3%.
Factory-gate inflation: 4.6%.
September borrowing: £24.5bn.
September debt interest: £10.9bn.
Total public debt: £3.005tn.
Borrowing for the first six months of the financial year was £97.8bn, £6bn above forecast, although lower than the same period last year.
And 10-year gilts are still yielding around 5.37%.
So the government has higher-than-expected inflation, expensive borrowing and very little fiscal headroom.
The Chancellor delivers the Budget next Wednesday.
Meanwhile, Barclays announced another 5,000 job cuts, around 5% of its workforce, as it pushes AI-led cost reductions.
Britain needs growth. Instead, it's getting another inflation shock and an increasingly expensive debt burden.
🇺🇸 US Treasuries: A Flight to Safety
Yesterday's $13bn 20-year auction cleared at 5.469%, up from 5.420% a month earlier.
Bid-to-cover improved to 2.73x.
The auction tailed by just 0.9bp, although indirect bidder participation fell to 58.2%, below its recent average.
Not a disaster. Just expensive.
Today, Treasuries are rallying:
5Y: 4.84%
10Y: 5.02%
30Y: 5.33%
That's a decent reversal from yesterday's 10-year yield of 5.11%.
Investors are seeking safety (oh the irony) in US government debt despite the country's enormous refinancing requirements.
But falling yields on a risk-off day don't solve America's longer-term debt arithmetic.
Today also brings a $26bn five-year inflation-protected Treasury auction.
With energy prices pushing inflation higher, demand for inflation protection is worth watching.
🛢️ Oil: Iran Is Running Out of Cash, Not Threats
Brent: $111.38, +2.35%
WTI: $98.40, +1.29%
According to the WSJ, Iran's floating oil reserves in Asian waters have now been exhausted after months of US naval blockade.
Those reserves previously held around 135m barrels, providing Tehran with billions in revenue from sales to China.
Now that cushion is gone.
Meanwhile:
Houthi attacks have killed four people at Saudi energy and airport facilities.
Aramco has told employees to leave its Dhahran headquarters.
The US has ordered some diplomats and families to leave Saudi Arabia.
Trump hits them to divert ffrom th emidterms where he is getting crushed.
Iran's IRGC is threatening US bases over American naval escorts through Hormuz.
The US and Israel have reportedly prepared plans for renewed strikes after the midterm elections.
And Saudi Arabia is exploring access to foreign strategic oil storage to protect future supply commitments.
Iran is losing a major source of revenue while the conflict increasingly threatens Saudi energy infrastructure.
That's a dangerous mix for energy prices, inflation and global growth.
🇯🇵 Japan / FX / Gold / Bitcoin
USD/JPY: 157.17, down from 158.09 yesterday. DXY: 101.86 EUR/USD: 1.1197
The yen is strengthening as investors reduce risk.
The BOJ is expected to hold rates next week, but a sustained yen rally alongside falling AI shares would put leveraged carry trades back under pressure.
Gold: $4,070, -1.9% Silver: $55.39, -4.0% Copper: $6.41, -2.5% Bitcoin: $81,790, -3.2%
Yesterday gold was above $4,100 and Bitcoin was recovering.
Today both are being sold.
For now, the flight to safety is going into Treasuries, the yen and German Bunds rather than gold and crypto.
📅 Coming Up
Today: $26bn US five-year TIPS auction.
Tonight's earnings: Western Digital $WDC, Bloom Energy $BE and Newmont $NEM.
Western Digital should provide another read on AI-related storage demand. Bloom Energy matters for the power requirements of data centres.
Friday: Moody's French sovereign rating review, UK retail sales and flash PMIs.
Next week: US 2Y, 5Y and 7Y Treasury auctions, the UK Budget on 28 October, and major central bank decisions.
And one important US number today: initial jobless claims fell to just 194,000.
The labour market still isn't showing widespread layoffs, despite high rates and expensive energy.
That makes the Fed's position even more difficult.
AI STOCKS ARE BOUNCING. THE DEBT MARKET STILL ISN’T BUYING IT.
In the words of Mr T... “I pity the fool.”
⭐️ Nasdaq +1.0% and VIX back down to 16.6.
⭐️ SpaceX and Super Micro are dragging AI higher.
⭐️ CoreWeave is still down while the rest of AI bounces.
⭐️ France just hit 5% on the 10Y.
⭐️ The US 20Y auction is the next real bond test.
⭐️ And Big Tech is stretching depreciation lives just as AI hardware is turning over faster.
📈 Stocks
S&P 500: 7,749, +0.48%
Dow: 52,655, +0.30%
Nasdaq: 26,686, +0.99%
Russell 2000: +0.36%
VIX: 16.6, -4.1%
VIX is cheap as chips again considering what is still going on underneath.
AI is doing the work:
SpaceX $SPCX +10.0%
Super Micro $SMCI +9.1%
Tesla $TSLA +3.5%
Nvidia $NVDA +1.6%
CoreWeave is still going the other way:
$CRWV: $61.11, -1.6%.
That divergence is still worth watching.
🤖 AI: Demand Fine. Funding Still The Problem.
SpaceX had a cracking quarter.
Revenue: $14.6bn, +161% YoY
AI revenue: $7.6bn
Net income: $4.4bn
Capex: $19.6bn
More than half the revenue is now coming from AI/server rentals.
So demand is clearly there.
But the capital requirement is fucking enormous.
And S&P has now warned that hyperscalers probably have more AI-linked liabilities than anyone can actually see, buried in SPVs, guarantees, leases and data-centre structures. It reckons the top six could spend more than $7tn on AI infrastructure between 2025 and 2030.
Reuters separately estimates Big Tech already has around $1.09tn of future data-centre lease commitments that have not yet fully hit reported balance sheets.
So the chain is still:
Credit spreads ↑ → funding costs ↑ → funding conditions tighten → project returns ↓ → capex gets harder
🧮 One More AI Accounting Trick Worth Watching
The depreciation assumptions are getting stretched.
A lot of Big Tech is now using 5.5–6 year useful lives for servers and networking kit.
That matters because moving $100bn of equipment from a 4-year life to 6 years cuts annual depreciation from:
$25bn → $16.7bn
That is roughly $8.3bn more pre-tax profit without one extra dollar of cash coming in.
And AI hardware is hardly getting obsolete more slowly.
Quite the opposite.
New GPUs, memory, networking and accelerators are turning over faster.
So extending useful lives while the hardware cycle speeds up is worth watching very closely.
More capex + more debt + longer depreciation = nicer EPS today.
The problem arrives later if the kit has to be impaired or replaced faster than the accounts assumed.
📈 Treasuries / 20Y Auction
Current:
5Y: 4.93%
10Y: 5.11%
30Y: 5.41%
Treasury is selling $13bn of 20Y bonds at 1pm ET.
That is today’s proper duration test.
The 20Y is around 5.48% going into it.
If that auction tails badly, the sovereign-funding problem is straight back on the table.
If it clears well, then Treasuries really are reclaiming some of their haven role.
🇫🇷 France: 5% Has Arrived
France’s 10Y OAT hit 5% today, the highest since 2002.
OAT/Bund is around 150–157bp, close to levels not seen since the euro crisis. Reuters
That is no longer some little French political wobble.
It is:
weak growth
big deficits
protests
budget problems
political uncertainty
and a proper sovereign risk premium
The CAC is actually up today, which makes the bond move even more interesting.
Equities are shrugging.
Bond investors are not.
🇯🇵 Japan: Still Part Of The Debasement Trade
Japan still has the same ugly mix:
USD/JPY: ~158
2Y JGB: ~1.9%
5Y JGB: ~2.3%
10Y JGB: ~2.9%
30Y JGB: ~3.5%
BOJ still wants to tighten.
Government still wants fiscal support.
Debt keeps growing.
Yields stay historically high.
And the yen is still weak.
If Japanese investors can earn materially more at home, that matters for who keeps buying US Treasuries too.
🛢️ Oil / Energy
Brent: ~$108.8
WTI: ~$97.3
Oil is stuck between two opposing forces.
On one side:
18+ vessel attacks in two weeks around Hormuz
Saudi infrastructure risk
Iran escalation
On the other:
Russia/Ukraine talks possibly restarting
more Russian barrels
some diesel relief
That is why Brent is still above $100 without running away again.
💵 FX / Gold / Bitcoin
DXY: 101.78
USD/JPY: 158.09
EUR/USD: 1.1231
Gold: ~$4,149
Bitcoin: ~$84,665, +2.4% Pasted text Pasted text
Gold holding above $4,100 with Treasury yields still north of 5% remains interesting for the debasement trade.
📅 Coming up
Meta and Tesla report after the close.
For Meta:
capex
AI monetisation
margins
depreciation assumptions
For Tesla:
AI spend
autonomy
robotics
and anything concrete on the SpaceX tie-up
These are now much more than simple EPS beats or misses.
The market wants to know whether the returns are arriving quickly enough to justify the capital being thrown at AI. Reuters
The Read
The surface looks fine.
Nasdaq +1%.
VIX 16.6.
SpaceX +10%.
Super Micro +9%.
But underneath:
CoreWeave still weak
AI liabilities still growing
Big Tech stretching depreciation lives
France at 5%
US long rates still above 5%
Brent still above $100
So yes, equities are bouncing.
The bit I still don’t trust is the funding and accounting underneath the build-out.
$SoftBank is getting properly fucked. Down 11% today and 52% from its June high.
Borrowing billions at up to 9.75% to stay in the AI game, with precious little cash flow coming from its biggest bets.
Its 5-year CDS is at 432bps, a four-year high. Bonds trading below par, with some now yielding over 10%.
Yet if OpenAI pulls off its proposed $1.4tn raise, SoftBank's 13% stake alone would be worth 15% MORE than its entire market cap.
And that's before Arm and the rest of the portfolio.
It does add up. This is what happens when paper valuations run miles ahead of cash flow, and you need increasingly expensive debt to keep the whole thing going.
Higher rates, inflation and a weakening dollar only make the funding problem worse.
The AI bubble is becoming a massive cash flow crisis that will flow all the way down into private markets, where they are still throwing cash at massive valuations.
I've spent over 30 years starting, building and investing in businesses, and I reckon we've got the whole startup thing a bit backwards.
We celebrate starting, raising and announcing.
Yet none of that means you've built a business.
Until someone wants what you're selling and is prepared to pay for it, you've got a hobby.
AI STOCKS ARE BOUNCING. THE DEBT MARKET STILL ISN’T BUYING IT.
In the words of Mr T... “I pity the fool.”
⭐️ Nasdaq +1.0% and VIX back down to 16.6.
⭐️ SpaceX and Super Micro are dragging AI higher.
⭐️ CoreWeave is still down while the rest of AI bounces.
⭐️ France just hit 5% on the 10Y.
⭐️ The US 20Y auction is the next real bond test.
⭐️ And Big Tech is stretching depreciation lives just as AI hardware is turning over faster.
📈 Stocks
S&P 500: 7,749, +0.48%
Dow: 52,655, +0.30%
Nasdaq: 26,686, +0.99%
Russell 2000: +0.36%
VIX: 16.6, -4.1%
VIX is cheap as chips again considering what is still going on underneath.
AI is doing the work:
SpaceX $SPCX +10.0%
Super Micro $SMCI +9.1%
Tesla $TSLA +3.5%
Nvidia $NVDA +1.6%
CoreWeave is still going the other way:
$CRWV: $61.11, -1.6%.
That divergence is still worth watching.
🤖 AI: Demand Fine. Funding Still The Problem.
SpaceX had a cracking quarter.
Revenue: $14.6bn, +161% YoY
AI revenue: $7.6bn
Net income: $4.4bn
Capex: $19.6bn
More than half the revenue is now coming from AI/server rentals.
So demand is clearly there.
But the capital requirement is fucking enormous.
And S&P has now warned that hyperscalers probably have more AI-linked liabilities than anyone can actually see, buried in SPVs, guarantees, leases and data-centre structures. It reckons the top six could spend more than $7tn on AI infrastructure between 2025 and 2030.
Reuters separately estimates Big Tech already has around $1.09tn of future data-centre lease commitments that have not yet fully hit reported balance sheets.
So the chain is still:
Credit spreads ↑ → funding costs ↑ → funding conditions tighten → project returns ↓ → capex gets harder
🧮 One More AI Accounting Trick Worth Watching
The depreciation assumptions are getting stretched.
A lot of Big Tech is now using 5.5–6 year useful lives for servers and networking kit.
That matters because moving $100bn of equipment from a 4-year life to 6 years cuts annual depreciation from:
$25bn → $16.7bn
That is roughly $8.3bn more pre-tax profit without one extra dollar of cash coming in.
And AI hardware is hardly getting obsolete more slowly.
Quite the opposite.
New GPUs, memory, networking and accelerators are turning over faster.
So extending useful lives while the hardware cycle speeds up is worth watching very closely.
More capex + more debt + longer depreciation = nicer EPS today.
The problem arrives later if the kit has to be impaired or replaced faster than the accounts assumed.
📈 Treasuries / 20Y Auction
Current:
5Y: 4.93%
10Y: 5.11%
30Y: 5.41%
Treasury is selling $13bn of 20Y bonds at 1pm ET.
That is today’s proper duration test.
The 20Y is around 5.48% going into it.
If that auction tails badly, the sovereign-funding problem is straight back on the table.
If it clears well, then Treasuries really are reclaiming some of their haven role.
🇫🇷 France: 5% Has Arrived
France’s 10Y OAT hit 5% today, the highest since 2002.
OAT/Bund is around 150–157bp, close to levels not seen since the euro crisis. Reuters
That is no longer some little French political wobble.
It is:
weak growth
big deficits
protests
budget problems
political uncertainty
and a proper sovereign risk premium
The CAC is actually up today, which makes the bond move even more interesting.
Equities are shrugging.
Bond investors are not.
🇯🇵 Japan: Still Part Of The Debasement Trade
Japan still has the same ugly mix:
USD/JPY: ~158
2Y JGB: ~1.9%
5Y JGB: ~2.3%
10Y JGB: ~2.9%
30Y JGB: ~3.5%
BOJ still wants to tighten.
Government still wants fiscal support.
Debt keeps growing.
Yields stay historically high.
And the yen is still weak.
If Japanese investors can earn materially more at home, that matters for who keeps buying US Treasuries too.
🛢️ Oil / Energy
Brent: ~$108.8
WTI: ~$97.3
Oil is stuck between two opposing forces.
On one side:
18+ vessel attacks in two weeks around Hormuz
Saudi infrastructure risk
Iran escalation
On the other:
Russia/Ukraine talks possibly restarting
more Russian barrels
some diesel relief
That is why Brent is still above $100 without running away again.
💵 FX / Gold / Bitcoin
DXY: 101.78
USD/JPY: 158.09
EUR/USD: 1.1231
Gold: ~$4,149
Bitcoin: ~$84,665, +2.4% Pasted text Pasted text
Gold holding above $4,100 with Treasury yields still north of 5% remains interesting for the debasement trade.
📅 Coming up
Meta and Tesla report after the close.
For Meta:
capex
AI monetisation
margins
depreciation assumptions
For Tesla:
AI spend
autonomy
robotics
and anything concrete on the SpaceX tie-up
These are now much more than simple EPS beats or misses.
The market wants to know whether the returns are arriving quickly enough to justify the capital being thrown at AI. Reuters
The Read
The surface looks fine.
Nasdaq +1%.
VIX 16.6.
SpaceX +10%.
Super Micro +9%.
But underneath:
CoreWeave still weak
AI liabilities still growing
Big Tech stretching depreciation lives
France at 5%
US long rates still above 5%
Brent still above $100
So yes, equities are bouncing.
The bit I still don’t trust is the funding and accounting underneath the build-out.
Layoffs at @carbon_robotics as firm “realigns costs due to broader economic conditions”
Carbon Robotics is cutting staff even as it pushes toward an IPO and new products. If you supply or compete in ag robotics, prepare for tighter budgets and shifting…
wire.fusion-42.com/stories/b…
AI STOCKS ARE BOUNCING, IGNORING THE DEBT MARKET
In the words of Mr T... "I pity the fool"
⭐️ Nasdaq is up more than 1%, but AI credit is still widening.
⭐️ CoreWeave bonds are still yielding around 15%.
⭐️ Treasury yields are back up after last week’s haven rally.
⭐️ France is getting properly ugly.
⭐️ Oil has eased, but the physical market around Hormuz is still a mess.
📈 Stocks
US equities are bouncing:
S&P 500: 7,698, +0.5%
Nasdaq: 26,454, +1.2%
Dow: flat
Russell 2000: +0.4%
VIX: 18.4, -8%
Tech is doing the work.
Amazon $AMZN +5.0%
Intel $INTC +2.9%
Nvidia $NVDA +1.6%
IREN $IREN +4.4%
CoreWeave is still weak:
$CRWV: $60.31, -1.8%.
And breadth still isn’t great. Reuters had 242 Nasdaq names making new lows against only 20 new highs even as the index bounced. Reuters
🤖 AI: Equities Up, Credit Wider
This is the bit I care about.
Oracle 5Y CDS: ~288bp
SoftBank 5Y CDS: ~424bp
CoreWeave CDS: 1,000bp+
CoreWeave’s 2031 bond yield has eased a bit from last week’s panic, but is still around 15.1%.
So the equity market is saying:
“buy the dip.”
The credit market is saying:
“show me how the fuck you’re going to finance it.”
And S&P has added another problem today: it thinks hyperscalers probably have more AI-linked liabilities than we can actually see, buried in SPVs, guarantees and data-centre structures. It reckons the top six could spend more than $7 tn on AI infrastructure between 2025 and 2030.
The chain remains:
Credit spreads ↑ → funding costs ↑ → funding conditions tighten → project returns ↓ → capex gets harder
📈 Treasuries / Fed
Treasuries had a proper haven rally into Friday, despite the hot CPI print.
Today some of that move is reversing as equities bounce:
2Y: ~4.79%, +5bp vs Friday
5Y: 4.94%, +4bp
10Y: 5.14%, +5bp
30Y: 5.48%, +5bp
So the curve is basically 5bp higher across the board.
The more interesting change is Fed pricing.
After last week’s hot CPI, October hike odds briefly jumped hard. They have since collapsed after Warsh refused to pre-commit and financial conditions tightened through the market itself.
That is a pretty important shift:
Stocks ↓ → credit wider → funding harder → Treasuries bid → Fed pressure eases
🇫🇷 France
Europe is having a worse day.
CAC 40: -1.1%
Euro Stoxx 50: -1.0%
DAX: -1.2%
FTSE 100: -1.1%
France is the main problem.
The 10Y OAT is around 4.9% and the spread over Bunds is about 146bp as riots, weak growth, budget delays and fiscal worries pile on. French banks and retailers are getting hit.
This no longer looks like a one-day political wobble.
The market is charging France a proper risk premium now.
🛢️ Oil / Energy
Oil has eased from the highs:
WTI: around $99–100 Brent: around $108–110
Trump’s Russia sanctions waiver is helping because it keeps more Russian supply flowing.
But the physical market is still nasty.
Hormuz flows reportedly fell from more than 20m barrels/day to around 8.5m during the latest disruption, and shipping attacks are still the big physical risk.
🇯🇵 Japan, Still A Massive Part of the Debasement Trade
Ueda says inflation is close enough to target to justify more tightening, while the government is still talking about more debt-funded fiscal support.
USD/JPY: ~158.5
2Y JGB: ~1.90%
5Y JGB: ~2.30%
10Y JGB: ~2.91%
30Y JGB: around 3.5% area
So Japan still has the ugly mix:
more debt → higher JGB yields → weak yen → tighter BOJ
And if Japanese investors start demanding more yield at home, that matters for US Treasuries too.
🇨🇳 China
China’s Q3 GDP came in at 4.5%, better than expected, while the yuan is near a four-year high.
The more important long-term point is still the same.
China does not need to kill the dollar.
It just needs to:
buy fewer Treasuries hold more gold settle more trade in yuan build deeper RMB markets and make itself less dependent on the dollar system
Slowly.
While Washington keeps issuing more debt at 5%+.
Tht is how the marginal buyer problem gets worse over time.
📅 What Matters Next
Tuesday: SpaceX $SPCX reports after the close, with the call at 4:30pm ET. That one matters because it is also reportedly trying to raise $40bn for AI compute — roughly $10bn of bank loans + $30bn of IG debt. ir.spacex.com
Wednesday after the close: META, TSLA, LRCX, TXN and IBM. That is the first proper earnings test after last week’s AI funding wobble. Watch Meta capex/AI monetisation, Lam/TI chip orders and whether anyone delays infrastructure spend. Reuters
Global govie supply
Japan: tomorrow is the big one — ¥700bn of 20Y JGBs. With the BOJ still sounding hawkish and the government talking more debt-funded fiscal support, that auction matters. Japan also has a ¥500bn liquidity-enhancement auction Thursday. 財務省
UK: conventional gilt auction Tuesday, then an index-linked gilt auction Thursday. DMO
France: no new long OAT auction this week after last Thursday’s sale, but France is still issuing bills and the OAT/Bund spread is the thing to watch, not supply itself. Ministère de l'Agriculture
US: the main duration tests are the 20Y Treasury reopening on Wednesday and 5Y TIPS on Thursday.
So the next 72 hours are pretty clean:
Japan 20Y auction → SpaceX results/funding → META/TSLA/LRCX/TXN/IBM → US 20Y/TIPS supply
That is enough to tell us whether last week’s AI/credit stress was just a wobble or the start of a proper repricing.
THE BOUNCE FAILED. THE DEBT GROWS.
Thats was a week that was.
⭐️ CPI, PPI and import prices all came in hotter than expected.
⭐️ Nasdaq lost 4.5% this week. CoreWeave lost 25.2%.
⭐️ Government bonds caught a safety bid while Washington reported another $2tn annual deficit.
⭐️ Europe outperformed. Japan’s higher borrowing costs remain a global risk.
⭐️ Brent finished near $110.
⭐️ Ukraine’s satellite restrictions exposed the cost of depending on infrastructure somebody else controls.
📉 STOCKS — DEAD CAT BOUNCE
Weekly moves:
Nasdaq: −4.5%
S&P 500: −1.9%
Russell 2000: −3.0%
Dow: +0.7%.
The Dow held up while technology and smaller companies took the damage. Money is becoming more selective about which earnings—and which balance sheets—it trusts.
VIX finished at 19.99, up roughly 35% over the week, after reaching 22.31 on Friday. Protection is getting more expensive.
🔥 INFLATION. THE PROBLEM IS STILL HERE
CPI: 3.7% annually, with core prices up 0.3% monthly.
PPI: 6.3% annually, up 1.3% in September alone. Producer energy prices jumped 10.8%. Strip out food, energy and trade services, and prices were still 5% higher annually.
Import prices: +1.1% monthly against 0.8% expected, and +8.1% annually. Those figures exclude tariffs. Duties add another cost, which businesses absorb in margins or pass through to customers.
Energy, imported goods, tariffs and financing costs. Businesses are being squeezed from several directions.
🥇 GOLD. I STILL LIKE IT
Gold sold off after CPI as traders priced a tougher Fed. It then recovered, trading around $4,165–$4,167 late Friday, although still roughly 0.7% lower over the week.
My view: gold around $4,145 looks like a gift over the longer term.
The market’s first reaction was that the Fed would fight harder. My concern is how far behind the problem it remains.
Higher rates can squeeze demand. They cannot reopen Hormuz, remove tariffs or make Washington balance its books. They also increase the cost of refinancing debt.
🤖 AI DEMAND DOESN’T PAY THE INTEREST BILL
CoreWeave closed at $61.39:
−11.9% on Friday.
−25.2% over the week.
Thursday’s 5.9% bounce disappeared the following day.
Dell lost approximately 23% over the week, showing how the pressure reaches equipment suppliers.
CoreWeave bond yields traded above 15% during the week. Its five-year CDS spread passed 1,100bp on Thursday. Oracle’s default protection also reached record levels. These are reported levels during the week, rather than Friday closing quotes.
Funding costs ↑ → project returns ↓ → refinancing harder → capex delayed → supplier earnings at risk.
AI adoption can keep growing while investors lose money. The equipment, electricity and debt still have to be paid for.
Revenue backlogs are useful. Cash flow determines whether the financing works.
🤖 ANTHROPIC GROWTH UP, VALUATION UNDER PRESSURE
Anthropic reportedly told investors Q3 revenue reached $16.8 bn, up 45% quarter on quarter, with an adjusted profit. Strong growth, but slower than Q2’s more-than-doubling.
Some investors are now discussing $1.5 tn, against earlier expectations around $2 tn. That is 25% lower, though no IPO price has been agreed.
Still no public S-1 located. The filing was confidential; reported prospectus details and private investor presentations are what we have so far.
“Adjusted profit” needs the adjustments. Revenue needs margins, cash flow and spending commitments alongside it.
At trillion-dollar valuations, even 45% quarterly growth can disappoint. That tells you how much perfection was already priced in.
My guess is that the numbers are shit (for a $2t val) and they will delay using a safety excuse.
🏦 US BONDS A SAFETY BID APPEARED BUT THERE IS STILL A MONSTER BORROWING PROBLEM
US 10-year: approximately 5.09%, down around 16bp over the week.
US 30-year: 5.44%, also down around 16bp.
Investors bought Treasuries as they cut equity risk.
My view is that this relief will struggle against the scale of borrowing still ahead.
Friday’s fiscal figures:
$5.40tn collected.
$7.40tn spent.
$1.992tn annual budget deficit.
$1.36tn gross interest costs, equivalent to roughly 25% of federal receipts.
Total federal debt above >$40tn.
Washington must finance that spending gap alongside refinancing maturing debt. Older, cheaper borrowing rolling into higher rates adds pressure to the interest bill.
Spending grew 6%. Receipts grew 3%. The gap widened despite record revenues.
This is a debt spiral and the credit-card binge mentality continues.
🇨🇳 CHINA EXPORT BOOM, DOMESTIC STRAIN
Shanghai Composite: 3,716 | −2.6% over the week.
Hang Seng: 24,187 | −0.1%.
Friday’s rebound did little to repair mainland losses.
September exports surged 36.3% annually, against 25.5% expected. Imports rose 31%, versus 24.5% expected.
The monthly trade surplus reached a record $133.5 bn, taking the first nine months to $938.1 bn. AI-related equipment and semiconductor shipments were major contributors. Part of the export surge reflects higher chip prices, so headline export growth overstates the increase in physical volumes.
The domestic picture is much weaker.
Producer inflation accelerated to 5%, above 4.5% expected. Consumer inflation was only 0.9%, below the 1% forecast.
Factories face rising energy and commodity costs while weak household demand limits their ability to pass those costs on. That puts margins under pressure.
💴 THE YUAN AND CAPITAL FLOWS
The yuan traded around 6.69 per dollar on Friday, near its strongest since January 2023 and heading for a third consecutive weekly gain. Strong exports are supporting the currency despite domestic weakness.
Beijing also pledged further financial-market opening, wider international use of yuan bonds as collateral and stronger offshore yuan markets. This is relevant to diversification away from dollar assets, although it does not establish that the dollar is losing reserve-currency status.
China is exposed to the same AI spending cycle supporting US technology. If financing pressure slows global data-centre investment, Chinese equipment suppliers feel it too.
🇯🇵 JAPAN CHEAP FUNDING IS GETTING LESS CHEAP
Japan’s 10-year yield finished around 2.94%, down 7.9bp over the week.
The 30-year finished at 4.01%, down 5.7bp.
Japanese bonds rallied too, but the BOJ’s policy rate is now 1.25%. Several brokerages expect another increase in December, while Ueda warned about inflation overshooting its target.
Higher domestic yields can encourage Japanese investors to keep more capital at home. More expensive yen funding also squeezes leveraged carry trades.
That matters well beyond Japan.
🇪🇺 EUROPE BETTER EQUITIES, SAME INFLATION CONSTRAINT
STOXX 600 finished at 644.74, gaining more than 2% over the week. Strong results helped Europe outperform the concentrated US technology trade.
But eurozone inflation remained 3.8%. The ECB has raised rates 50bp this year, taking its deposit rate to 2.50%, with another December increase priced in.
Germany’s 10-year yield was around 3.42% late Friday. Europe’s stronger equity performance does not remove its financing constraints.
🛢️ OIL — STILL FEEDING INFLATION
Brent settled at $109.86, up 3.72% on Friday, as Middle East supply risks kept prices elevated.
That feeds into freight, manufacturing and household bills. It also makes it harder for central banks to support weakening economies without risking another inflation surge.
🛰️ STARLINK / STARSHIELD — WHO CONTROLS THE SWITCH?
Ukraine says its military Starshield connections were cut when terminals travelled above roughly 160 km/h, disrupting drone operations. Negotiations to remove the restriction continued on Friday.
This concerns the speed at which terminals move. Musk says the US government controls Starshield; Trump denied ordering the restrictions. Responsibility remains unresolved.
Critical infrastructure comes with political and counterparty risk when another party can restrict access.
For Europe, sovereign communications capacity has a very practical value.
👀 NEXT WEEK: 19–23 OCTOBER
📊 Data
Mon: China GDP (4.4% expected vs 4.3% prior), industrial production and retail sales; Canada CPI.
Tue: UK jobs/wages; China loan prime rates; US housing starts, permits and pending sales.
Wed: UK CPI.
Thu: US jobless claims.
Fri: Japan CPI; UK retail sales; US/UK/eurozone flash manufacturing and services PMIs; final Michigan sentiment and inflation expectations.
🏦 US auctions
Mon: $95bn 13-week + $82bn 26-week bills.
Tue: $95bn 6-week bills.
Wed: $13bn 20-year bond reopening.
Thu: $26bn new 5-year TIPS.
$311bn announced, plus scheduled 17-, 4- and 8-week bills. Gross issuance includes refinancing.
🇯🇵 Japan
Tue: ¥700bn 20-year JGBs.
Thu: ¥500bn liquidity-enhancement reopening auction.
🇪🇺 Europe
Tue: UK 2036 conventional gilt; German green bonds.
Thu: UK 2031 index-linked gilt.
Watch the 20-year auction for sustained demand, TIPS for inflation protection, and Friday’s data for rising prices alongside weakening growth.
I don’t see gut feeling and data-driven investing as opposites.
What we call gut feeling is just human pattern recognition that we can’t yet fully explain.
As more companies are built using AI, we’ll have far more data on founders, their decisions, behaviour and outcomes. Over time, we’ll get better at understanding the patterns behind successful businesses.
LLMs alone won’t get us there. We need AI that can understand real-world behaviour, not just language.
A VC’s gut feeling may be an information advantage today. That doesn’t make it a fundamental human advantage tomorrow.
Just because we can’t explain something yet doesn’t mean we won’t.
THIS 100%
I think one of the biggest problems with startup accelerators is that we’ve confused teaching founders how to pitch with teaching them how to build a business.
Spend three months perfecting your story.
Build a beautiful deck.
Practice your two-minute pitch 100 times.
Learn exactly which buzzwords investors want to hear.
Then get on stage at Demo Day and announce you’re raising $4M.
Great.
But can you actually sell your product?
Do you know why customers buy it?
Can you explain your margins without opening a spreadsheet?
I’m a big fan of accelerators. The best ones provide incredible mentorship, networks, and early capital.
But we’ve created an ecosystem where some founders can explain their billion-dollar market opportunity better than they can explain their first $10,000 in revenue.
Fundraising is a skill. Building a business is a completely different skill.
And the two don’t always overlap.
Maybe Demo Day should include five minutes of founders trying to sell their product to an actual customer.
Because getting 200 VCs excited about your startup is impressive.
Getting 20 customers to pay for it might be even more impressive.
Dug into this a few weeks ago when @HarryStebbings posted a timely and openly honest video on this.
VC is really caught between a rock and a hard place but doing the same as everyone else is shit strategy.
There will be a lot of reinvention required at the early stage to keep capital flowing into startups across the board.
xitter.cf/derek__watson/status/2…









