WTF Just Happened To Global Markets

Jeff Snider · Atlas Financial

Neoclouds are more fragile than hyperscalers because they lack ad, retail and software businesses to absorb losses.

Snider notes CoreWeave fell about 13%, Nebius about 16% and another name about 17% in the selloff. He argues neoclouds depend far more on continuous AI construction than Microsoft, Amazon or Google, which have advertising, retail and software businesses that can subsidise years of uncertain returns. If projects are delayed, neocloud revenue is delayed. If projects are cancelled, expected revenue disappears while financing obligations remain. The weakness predates today's move.

“And these companies are far more dependent on continuous AI construction than say Microsoft, Amazon, and Google. They don't have gigantic advertising, retail, and software businesses that are capable of subsidizing years of uncertain returns.”at 4:39

Jeff Snider · Atlas Financial

Political pushback on data centers will raise financing costs before any project is actually stopped.

After the midterm results, Democrats swept Congress and won key governor races, and Bernie Sanders called for a national data center moratorium. Snider says a moratorium is not law and may never be, but markets had assumed data centers could be built as fast as financing, equipment and power arrived, treating local resistance as a minor inconvenience. Data centers strain grids, water and land, and voters are noticing. Snider argues lenders will reprice risk once they judge future construction less predictable, so one proposed moratorium can have a much larger financial effect than its probability suggests.

“And if lenders conclude political risk has made future construction less predictable, the cost of financing rises before a single project is officially stopped.”at 5:06

Jeff Snider · Atlas Financial

Hyperscalers, though very profitable, are being forced to conserve cash and find new financing for AI spending.

Snider cites Amazon spending over $77B on property and equipment in one quarter, producing a free cash outflow of about $42.4B, and setting up supplier financing that could allow it to borrow up to $80B against hardware. He notes Alphabet has over $1 trillion in reported future payments tied to data centers, compute and energy, and Microsoft offered voluntary buyouts to about 60,000 employees while continuing its AI build-out. He argues these are not speculative startups, yet even they must conserve cash, use new financing structures and rank AI above other expenses.

“These companies, these specific ones I just went over, are not speculative startups. They are among the most profitable corporations in history. Yet even they are being forced to conserve cash, find new financing structures, and prioritize AI investments over all other kinds of expenses.”at 10:44

Jeff Snider · Atlas Financial

Lenders balk at financing AI chips without supplier guarantees, showing collateral alone is insufficient.

Snider cites a reported $63B financing package for Broadcom chips and networking equipment intended for Anthropic. Banks nearly backed away, lenders demanded higher rates and stronger loss protection, and Broadcom had to provide residual value support because banks would not rely entirely on the chips as collateral. Snider reads this as evidence that uncertainty about AI hardware values is rising and that financing is becoming harder, even for the most valuable chips.

“Banks nearly backed away from a reported $63,000,000,000 financing package for Broadcom chips and networking equipment that was intended for anthropic. Lenders demanded higher interest rates and stronger protection against those losses.”at 11:30

Jeff Snider · Atlas Financial

The AI capex boom has been one of the few legitimate economic tailwinds for the US and the world.

Snider argues that AI capital expenditure has been one of the few real sources of economic support in the US and globally, pointing to how much global trade this year has been boosted by the technology sector. He ties this to his bond market view: if the AI boom turns into a bust, one of the economy's few major investment engines disappears. That makes a bust more damaging to growth than to tech stocks alone.

“The AI capital expenditure boom has been one of the few, one of the few legitimate economic tailwinds for The United States and really around the rest of the world.”at 17:40

Jeff Snider · Atlas Financial

The Treasury rally reflects expected energy-driven demand destruction and weak fundamentals, not just an AI bust.

Snider says the two-year yield fell to about 4.48%, nearly half a point from its high, and the ten-year to around 4.8%, about 50 basis points below its peak. He says the rally is real, not a blip, and started before the election. The earlier yield rise came from September seasonal dealer constraints and Fed repricing, which have now faded, so the market can price fundamentals. Bonds are pricing energy-driven demand destruction, weak real-economy fundamentals, tighter credit and the possibility that an AI bust removes an investment engine. The result is a flight to safety with lower growth and inflation expectations.

“And bonds are not merely pricing an AI bust. They're pricing energy driven demand destruction, weak economic fundamentals, and the possibility...”at 2:09

Jeff Snider · Atlas Financial

A weakening economy limits how far the Fed can keep hiking, so the rate-hike path is compressing.

Snider says the bond selloff was driven mainly by the Fed, which markets feared would respond aggressively to an energy-driven rise in headline inflation. He argues the economy is now constraining that response. The worse the outlook, the less credible further hikes become. SOFR futures compressed as traders priced less room for hikes before something breaks. Options-implied odds of a December quarter-point hike fell from about 83% to about 70%, and the odds of no change rose to nearly 30%. Snider says a hike remains possible but the Fed's margin is shrinking.

“The worse the economic outlook becomes, the less credible an extended series of rate increases also becomes.”at 8:32

Jeff Snider · Atlas Financial

The energy shock destroys demand rather than sustaining inflation, as consumers ration spending.

Snider argues the energy shock is widely described as an inflation problem, but expensive energy cannot produce an endless inflationary spiral without rapid wage growth, expanding credit or a broad rise in income. Instead it destroys volume. Households spending $100 more on fuel or power have $100 less for restaurants, retailers and other discretionary items, so one category's higher price becomes another's lost revenue. Businesses facing higher transport, electricity and production costs cannot pass them on, which squeezes margins. Weakening fuel consumption while energy stays expensive shows demand being rationed.

“That's why gasoline demand and other measures of discretionary activity are such a big part of it. If fuel consumption weakens, for example, while energy remains expensive, the economy is revealing that demand is being rationed.”at 14:11

Jeff Snider · Atlas Financial

The selloff is broadening beyond AI, signalling a weaker economic foundation than markets believed.

Snider notes Caterpillar fell 5%, JPMorgan 4%, Boeing 4% and Nike 3%, none of them direct AI infrastructure plays. Caterpillar reflects construction investment, JPMorgan the credit cycle, Boeing industrial demand and financing, and Nike consumers. When these fall alongside technology, investors are no longer only questioning Nvidia's valuation but whether the economic base under the whole market is weaker than believed. He presents this as a signal the broader market is repricing growth.

“they aren't direct AI infrastructure trades at all. Their declines demonstrate that investors are moving beyond a narrow reassessment of semiconductor valuations.”at 14:49

Jeff Snider · Atlas Financial

China's tech bubble may already have burst, and Beijing's support has not restored a durable recovery.

Snider says China's STAR 50 index of major Chinese tech and semiconductor companies has fallen roughly 40% from its mid-year peak, about $500B in market value, and the CSI 300 dropped about 8.5% in October, its worst month in over a decade. Beijing responded with central bank swap operations of over 100B yuan, cheaper buyback financing and encouragement for tech buybacks, but these have not produced a durable recovery. He argues liquidity and buybacks can slow declines but cannot create profitable end demand for AI services. He treats China's bust as a warning for US investors.

“China's technology bubble may have already burst. The star 50 index of major Chinese technology and semiconductor companies has fallen roughly 40% from its mid year peak...”at 15:44

Jeff Snider · Atlas Financial

AI valuations required too many assumptions to hold at once, so the boom is vulnerable to a bust.

Snider argues AI valuations depended on many conditions holding at once: data centers approved, power available, chips retaining value, cheap financing, exponential and continuous AI demand, customers accepting profitable prices, and a broadly resilient economy. He says markets now see a narrower path to that outcome. He adds that the boom will eventually break, though the timing is unknown, and cites China's AI stock collapse as a precedent. The view is that valuations were never realistic.

“Chinese AI stocks had already collapsed. Too many assumptions had to remain true simultaneously.”at 18:28