NVIDIA’s supply commitments have grown from $50.3B to $390B in one year.
Of that, $203B comes due in fiscal 2028, with another $118B scheduled beyond that.
The question is how much NVIDIA can actually walk away from.
Its filings say some agreements can be canceled or rescheduled before firm orders, but also warn that NVIDIA may be unable to reduce its commitments.
In fiscal 2023, NVIDIA recorded roughly $1.13B in charges tied to excess purchase obligations and cancellation or underutilization penalties.
Today, demand remains strong. NVIDIA generated $48.3B in operating cash flow last quarter, and securing scarce memory years ahead may be a competitive advantage.
But the same commitments could pressure margins if demand weakens, product cycles shift, or customers struggle to finance new capacity.
The size of the commitments is disclosed.
How much is unavoidable isn’t.
$NVDA
A lot of people are blaming $NVDA’s latest drop on Nscale’s amended filing. But the timeline doesn’t quite support that explanation.
The filing came Thursday. Friday was the first trading session afterward. NVIDIA fell about 2%, while AMD lost nearly 9%, CoreWeave dropped roughly 6%, and the broader semiconductor index fell over 3%.
NVDA actually held up better than much of the group.
The filing still matters. Nscale disclosed it will need additional financing for its 2027 obligations. NVIDIA’s ties to Nscale include an equity stake, a lease guarantee, a planned $1B purchase, and a $1.2B capacity agreement where NVIDIA is the customer.
Those aren’t the same risks.
The bigger question is how AI infrastructure demand is being financed.
When a chip supplier invests in companies that buy its hardware, how much demand is independently funded, and how much depends on capital flowing through the same ecosystem?
That doesn’t mean the demand isn’t real. It means the financing structure matters just as much as the revenue.
I’m watching whether NVIDIA’s $1B purchase closes around Monday, and the earnings report Tuesday.
The timeline matters. So does the money behind it.
#NVIDIA #Nscale #AIInfrastructure #Semiconductors #CapitalFlows
Jensen Huang’s exit reads like ordinary governance on the record.
But three days earlier, NVIDIA’s board said U.S. national and economic security was “paramount” to shareholders.
At that scale, geopolitics is already in the boardroom.
That doesn’t prove a hidden hand. It means the incentives changed.
Truth matters more than the story we want to be true.
This is why “who is lending?” is becoming the wrong question.
The better question is:
What underlying customer, contract or balance sheet ultimately makes the financing possible?
Different funding channels can still concentrate around the same economic risk.
CoreWeave’s new 10-Q quietly answers an important AI financing question.
June: $31.4B recourse debt / $3.7B non-recourse.
September: $41.6B / $5.3B.
Debt grew almost $12B in one quarter, yet 89% still ultimately sits against CoreWeave parent credit.
The instruments are changing. The underlying dependency barely is.
Sometimes we ask God to change our lives while avoiding the next honest thing we already know we need to do.
An apology. A debt we owe. A lie we need to stop protecting. Someone whose pain we keep explaining away because admitting it would mean admitting what we did.
Yeshua came close to people others had written off. He ate with sinners. His mercy wasn't an excuse to keep hurting people. It made returning possible. We can come to Him honestly, including with the things we're most ashamed to admit, and begin to change.
Maybe today that means making one repair. Where contact is welcome and safe, offer the apology without attaching a defense. Return what you owe where you can. Tell the truth without making the other person responsible for comforting you. They may not forgive you today. Doing what's right cannot depend on getting the response you want.
If someone has asked you not to contact them, respect that. Repentance can begin with changed behavior, not another message they didn't ask for.
And if you're the one who was hurt, forgiveness doesn't require pretending it didn't happen or giving someone access to hurt you again.
Bring the truth to Him this morning. Ask for help with the part you cannot face alone. Then take the next honest step you can.
Markets taught me not to confuse a story with evidence.
I think Yeshua (Jesus Christ) deserves the same discipline.
Historically, He existed, was crucified, and His followers very early claimed they saw Him alive again. Saying He is Lord is theological. Different category.
Then there is a pattern I can't unsee. Not proof. Recognition.
Genesis loses the Tree of Life. Yeshua dies on a tree. Revelation restores it.
Genesis begins with humanity hiding from God's face. Revelation ends with humanity seeing His face.
Maybe faith should not begin with forced certainty.
Maybe it begins by looking closely enough to recognize Him.
@NickTimiraos Korea is a useful test: MOF waived the 140% margin-loan collateral rule for 3 months from Nov 9 to curb forced sales. That can slow liquidation without fixing chip earnings; brokers may retain more credit risk. Liquidity relief ≠ a loss guarantee.
https://t.co/t7s6fNMGZx
YES, and that’s why He’s hard to dismiss.
Our problems became more sophisticated. His diagnosis of the human heart stayed recognizable.
Screens. Soft-landing narratives. Margin rules rewritten so the loss moves onto a quieter balance sheet. New instruments. Same heart.
Yeshua wasn’t aimed at an era. He was aimed at us.
All 11 S&P sectors fell today. That tells us the selling was broad, not that the cause was a cash shortage.
To distinguish falling earnings expectations from funding stress, watch credit spreads and refinancing terms.
A red screen shows the loss. It doesn't identify the mechanism.
A data-center permit delay can leave the project alive and still damage its economics. Interest accrues before rent starts. Equity absorbs the overrun until someone must put in more cash. A signed lease is not the same thing as cash arriving in time to service debt.
Our read: bills give Treasury flexibility without adding longer-term issuance. But with the Fed's reverse repo cushion nearly empty, more bill supply relies more on investors shifting cash from elsewhere.
The question isn't whether Treasury can borrow. It's at what price.
Treasury's November refunding was largely uneventful: coupon auction sizes held steady.
But its advisory committee (TBAC) flagged something worth watching: "money market fund assets have leveled off and the bill market has started to show some signs of price sensitivity."
@hornyhov The basket’s median token is up 32%, but excluding KNTQ, DRV and HYPE drops its return to 19%.
A Jan. 1 top-30 DeFi basket, including MYX at -98%, returned just 9.5% equal-weighted, with a median of -8%.
The starting universe changes the story considerably.
Since Oct 23, overnight Treasury repo has cost more than the Fed pays banks on reserves. 7 straight business days. Fed funds hasn't moved.
Nothing's broken, but the reverse repo cushion is ~$0.2B. Tomorrow's refunding decides what lands on it.
#Treasury#Fed
Some backup financing for Anthropic’s chip leases depends on revenue performance, while the lease obligations remain fixed. Broadcom appears to backstop part of those leases itself, retaining contingent downside exposure even after outside financing enters the structure.
Not a distress call.
https://t.co/FA3qyFqhgw
Not strange if the root is unchanging human nature. Tech and systems grew complex, but pride, greed, resentment, and the hunger for meaning still drive the friction. His diagnosis targeted the heart, not the era. The record of influence and the ongoing pull of those claims make casual dismissal harder than it looks.
This is why I’m paying closer attention to Treasury issuance, auction demand, and the difference between short-term policy rates and long-term yields.
Not because a debt crisis is inevitable.
Because monetary policy can influence the price of money without fully controlling the price investors require to lend it.
At what point does fiscal financing become a constraint on monetary policy itself?
Stay tuned for more!
We spend an extraordinary amount of time asking what the Federal Reserve will do next.
I’m becoming more interested in a different question.
What happens when the government’s need to borrow begins influencing the very financial conditions the Fed is trying to control?
Now consider the competition for capital.
Governments need financing. Technology companies are committing enormous resources to AI infrastructure. Businesses need credit. Consumers need affordable borrowing.
All of them operate within financial markets where the cost and availability of capital are interconnected.
A strong investment opportunity doesn’t automatically mean financing it will remain attractive.