PodBrowser
Forward Guidance

The Fed Is A Sideshow While The AI Trade Implodes | Weekly Roundup

Saturday, 31 October 2026 · 4 min read · Listen to the episode ↗

The Fed held rates steady, but the discussion argues that weakening growth and persistent energy-driven inflation matter less to markets than cracks in the AI trade. Microsoft and Google both fell short of capital-spending estimates, while widening credit spreads and leverage threaten to amplify a pullback. The speakers debate whether AI infrastructure can generate durable returns and how long the unwind of semiconductor and data-center overinvestment could take.

Markets had broadly priced in a Federal Reserve pause, and the Fed held rates steady, but one speaker argued that the decision was a sideshow beside the risks from AI, energy supply and slowing growth. The Fed upgraded its description of activity from growing at a “solid pace” to a “strong pace”; Hamak and Cook dissented, with Cook’s dissent reportedly her first. One speaker called the meeting unusually meaningless, while another said Jerome Powell’s press conference was his best so far. The speakers differed on the outlook: one expected further market declines to weaken growth and potentially remove a December move from consideration, but also expected the November inflation reading to mark a fourth consecutive month of accelerating inflation.

The speakers described a policy bind: two consecutive jobs reports were very negative, the prior report was revised down, and GDP growth was flatlining even as energy-driven inflation persisted. One speaker thought the Fed was probably done hiking, while markets still priced roughly three European hikes. The Bank of Canada had hiked, possibly to support its currency, and other non-US central banks might also raise rates as the dollar rallied. Europe was seen as especially exposed: natural gas there was said to cost 11 times the US price, threatening manufacturing, industrial production, reshoring and defense expansion. One speaker thought the dollar could eventually peak as oil weakens growth and confidence in Fed promises fades, but said the timing could be weeks or months.

The AI investment trade showed signs of cracking as Microsoft reported $38 billion in capital spending against $45 billion expected, and Google reported $51 billion against $55 billion. One speaker also described Google search revenue as plummeting, largely because of AI and changing consumer habits, and viewed slowing forward spending estimates as a more useful warning than lagging GDP data. AI infrastructure had made a major contribution to recent growth: computer equipment investment had its largest contribution since 1959, while equipment investment and goods imports rose together for three quarters. But real rates were said to be at or above GFC-era levels, and credit spreads were widening, particularly for AI and hyperscaler borrowers. Hyperscalers were relying more on credit markets as cash flow fell, cutting buybacks and using cash flow, equity issuance and debt to fund expansion.

Leverage had amplified the AI boom and was already helping drive its unwind, one speaker argued. SOXL 3x ETFs had roughly a third of a trillion dollars in notional exposure, and a hedge fund was described as $45 billion and four times levered. Credit spreads can widen abruptly; neoclouds may struggle to issue bonds and could be left relying on equity issuance, while crowded, index-heavy ownership makes the market reflexive and unstable. The speaker viewed the trade as more appealing to short or trade on rallies than as a straightforward long-term investment. They distinguished AI’s usefulness from investors’ ability to earn returns on the buildout, comparing it with the late-1990s telecom boom, when infrastructure builders failed but later users benefited. Anthropic’s S1 offered more visibility into frontier-model economics, but open-weight models and commoditized compute could pressure pricing and margins. The speaker expected a sorting-out period to identify sustainable profits.

Semiconductor and memory shares were near levels last seen during a leverage washout, but the concern had shifted toward fundamentals: hyperscalers were missing or falling short of capital-spending estimates, unlike during the earlier memory selloff. Intel fell 28% on the day of recording, and one speaker said semiconductors were not yet a buy, though some could become attractive in 2027. That speaker also warned that cyclical overinvestment in semiconductors, capital spending and neoclouds could take as long to unwind as it took to build. Political resistance to data centers and AI was described as bipartisan, with some governors shifting from support toward pauses or moratoriums. Gold also fell despite lower bond yields and more dovish policy rhetoric, trading below 3,900 and below its summer lows. Liquidity selling and energy-short countries selling reserves were offered as possible pressures, though data still showed China buying; one speaker considered gold fundamentally attractive but warned its near-term decline could be painful.

The speakers connected economic strain to inflation’s erosion of purchasing power, fewer family trips and the need to work more, arguing that rising prices can fuel unrest and push other concerns aside. They cited protests in Paris, widening perceptions that lower- and middle-income people are being squeezed, and a shift in voters’ priorities toward economic and foreign-affairs problems. Oil had risen steadily for two to three months despite falling futures positioning and low volume; one speaker predicted escalation or an attack, while acknowledging uncertainty about timing. In crypto, a $6 billion exchange hack hit sentiment: Circle and Tether reportedly froze funds that could enable recovery of almost $3 billion, and Binance said it would cover the rest. One speaker expected Binance and crypto-native users to be fine but said the breach could deter newcomers. Bitcoin’s downturn was described as “time capitulation,” with repeated upside breakouts sold; the speaker saw a bottoming process but did not expect market psychology to improve before next year, possibly Q1 or Q3.

This summary was generated from the episode transcript and can contain mistakes.