AI Debt Met an Oil Shock: Nasdaq 100 Suffers Its Worst Week in 18 Months

By The News Desk, Previously

October 17, 2026

The Nasdaq 100 has just had its worst week in 18 months, and the story behind the fall is not a single bad earnings report. According to Binance’s TradFi analysis, published October 17, artificial intelligence debt met an oil shock — and investors, for one week at least, stopped believing that the spending would pay for itself on schedule.

Binance reported that the Nasdaq 100 fell 5.5% over the week to close at 29,187.55, falling 2.5% in the final session alone. The Philadelphia Semiconductor Index declined for an eighth consecutive session, Binance said, down roughly 16% and shedding about $2 trillion in market value in its longest losing streak since 2008. Oil, meanwhile, pushed above $110 a barrel on supply fears, adding an inflation and energy-cost shock to a market already worried about how much borrowed money is holding up the AI build-out. Every figure in this story is attributed to the outlet that published it, because in a week like this, whose number you are reading matters as much as the number itself.

The number that changed the conversation: CoreWeave’s debt

Binance’s TradFi piece singled out one company and one figure as the centre of the repricing. CoreWeave, the Nvidia-backed data centre operator, fell almost 12% on Friday, Binance reported, extending its weekly decline to 25% — its worst week since November last year — and leaving the stock more than 50% below its year-to-date high. Some of its bonds yielded above 15% at one point, Binance said.

The reason that yield matters, Binance argued, is the balance sheet behind it: CoreWeave’s total debt has grown past $51 billion. A 15% yield on a small borrowing is a curiosity. A 15% yield against $51 billion of obligations is the market repricing what it costs to finance AI infrastructure altogether. Binance’s conclusion was that the concern has shifted — from whether demand for computing power exists, to whether the returns justify the borrowing used to build it.

Binance also corrected what it called the week’s most misread move. Strategy, the company formerly known for holding Bitcoin on its balance sheet, fell 9.2% on Friday, Binance reported — but not because of anything to do with Bitcoin. MSCI had said it would adopt a new screening approach that could put the company’s index membership at risk, and index removal forces selling by passive funds regardless of what any manager thinks of the company. Strategy fell 9.2% on a day Bitcoin lost about 1%, Binance noted, precisely because the cause was index mechanics, not crypto.

Bitcoin held up better than the companies built around it

That divergence ran through the whole week. The task brief for this cycle, drawing on the same Binance analysis, noted Bitcoin fell about 1% on Friday against the Nasdaq 100’s 2.5% fall, holding around $80,000. The asset at the centre of the last speculative cycle was, this week, the calmer one.

The companies caught between the two cycles were not calm at all. Former Bitcoin miners that have reinvented themselves as AI hosting companies — Hut 8, IREN, CleanSpark and Cipher Mining among them — fell about 6% on Friday, according to the Binance-linked reporting summarised for this story, repriced on data centre funding costs rather than on the price of Bitcoin. The logic is straightforward once it is stated: a miner that has signed long leases to host AI computing, borrowed to build data centres and promised future revenue to lenders no longer trades mainly on the coin it used to mine. It trades on the cost of its debt. When that cost rises, it falls with the AI complex, however steady Bitcoin itself may be.

The bubble checklist: spending still rising, stress in the debt

If Binance supplied the week’s numbers, madrobot.blog supplied its framework. In an October 17 checklist asking whether the AI bubble is popping, madrobot.blog concluded that the cracks are in the most indebted corners of the boom, not yet in demand for AI chips — and then tested that claim sign by sign.

On spending, madrobot.blog found no retrenchment at the top. Alphabet raised its 2026 capital spending range to $195 billion to $205 billion in July, the checklist said, and Meta narrowed its range upwards to $130 billion to $145 billion. On the chip demand that spending buys, it found the same: Nvidia revenue rose 106% to $96.2 billion last quarter, madrobot.blog reported, with $108 billion forecast for the next. Big Tech is not cutting, and the chipmaker at the centre of the build-out is not stalling. Those signs, in the checklist’s own traffic-light terms, are not flashing.

The warning signs it did flag were in the financing layer underneath. Madrobot.blog reported that CoreWeave’s and Oracle’s bond yields hit records in October — the same stress Binance measured in CoreWeave’s 15% yields and $51 billion debt pile, seen from the bond market’s side. It reported that the International Monetary Fund counts $1.4 trillion of revenue promises tied to circular deals, in which AI firms fund and supply each other, and that none has unwound yet. The same appetite for debt-funded dealmaking in technology was visible this week in the Proofpoint–Varonis cybersecurity deal, and the week’s stress was not confined to New York — Previously also covered Indonesia’s worst stock-market day in the same risk-off stretch. It noted that revenue is growing fast at the big labs, but so are their losses and computing bills. And it recorded that the IMF, the Bank of England and the Federal Reserve have all flagged AI valuations, that traders are betting on more US rate rises, and that inflation data this week — September consumer prices up 3.7% year on year in the reporting around the sell-off — made all of that borrowing more expensive to carry.

Put the two accounts together and a coherent picture emerges, without Previously needing to invent one. Demand, as madrobot.blog measured it, is real and still growing. Spending, as both outlets measured it, is enormous and still rising. The stress, as both outlets located it, is in the debt used to bridge the gap between spending now and revenue later — concentrated in the most leveraged builders, and repriced in a single week when an oil shock and hot inflation raised the cost of carrying it.

Concentration: five companies, 93% of the gains

One further figure explains why a repricing in AI financing becomes a repricing of the whole market so quickly. Five companies — Microsoft, Nvidia, Meta, Apple and Alphabet — produced 93% of the S&P 500’s gains since July, according to the market analysis summarised for this story. When the index’s gains are that concentrated, there is nowhere inside the index to hide from the trade that is falling. A bad week for AI financing is, mechanically, a bad week for the benchmark most retirement accounts track.

That concentration cuts both ways, and the week’s other market reporting showed it. The Dow Jones Industrial Average, far less weighted to the AI builders, actually gained over the week in several of the accounts published alongside Binance’s, while the S&P 500 fell about 2% and the Nasdaq measures fell further. This was a rotation and a repricing of one crowded trade, not a uniform collapse of every stock — a distinction Binance’s own analysis made in explaining Strategy’s fall, and one worth keeping when the headlines say the market fell.

“All of this is going to pass” — Nvidia CEO Jensen Huang, via Odaily and Traders Union

The most senior voice arguing against panic this week was the chief executive of the company at the centre of the trade. Odaily reported that Nvidia CEO Jensen Huang, speaking at an event in Berlin, said the current turmoil in the AI market will eventually pass, that artificial intelligence will have a profound impact on all industries and the global economy, and that the industry is still in its early stages. In Odaily’s account, Huang said short-term changes in market news will not alter the long-term fundamentals of the AI industry, that large-scale hardware investment will ultimately pay off, and that investors should start from first principles and focus on long-term value rather than short-term fluctuations.

Traders Union, reporting the same Berlin remarks from a Bloomberg interview, quoted Huang directly: “All of this is going to pass, and the reason is very simple.” It reported that Huang emphasised the technology is still in its early stages and will continue spreading across all sectors, and that future revenues from AI adoption will, in his view, justify current investments in computing capacity. Readers should weigh that for what it is: the case made by the executive whose company sells the hardware the spending buys. It may prove right. It is not a neutral forecast, and neither Binance nor madrobot.blog treated the week’s debt stress as settled by it.

What happens next

The tests ahead are the ones the two main sources identified. For madrobot.blog’s checklist, they are whether bond stress at the weaker builders spreads or settles, whether the circular revenue promises begin to unwind, and whether AI revenue keeps pace with the losses and computing bills it is already generating. For Binance’s account, they are whether the semiconductor losing streak ends, whether CoreWeave’s financing costs come down from record levels, and whether oil stays above the levels that turned a financing worry into a market-wide shock.

What this week established, on the reporting cited here, is narrower than a bubble verdict in either direction. The Nasdaq 100 fell 5.5% in its worst week in 18 months, according to Binance. The chip index fell for eight straight sessions. The most indebted AI builder’s debt passed $51 billion while its bonds yielded above 15%. The biggest spenders, according to madrobot.blog, raised their spending plans anyway, and Nvidia’s revenue grew 106%. Both halves of that paragraph are true at once. That tension — real demand, enormous spending, and debt markets starting to charge more for the bridge between them — is the story of the week, and it will still be the story when earnings arrive to test it.