Still cautious

Geo Chen · Fidenza Macro · August 05, 2026 at 14:10 · ⏱ 9 min read  | Read on Substack ↗
Summary
The author argues that the AI semiconductor trade has permanently topped, while the broader market's rotation into hyperscalers is a dangerous trap built on circular financing. Hyperscaler revenues are being artificially inflated by unprofitable AI labs (OpenAI, Anthropic) using the hyperscalers' own investment capital, creating a fragile ecosystem vulnerable to decelerating AI revenues and massive hidden debt.
  • •The AI semiconductor ETF (SMH) likely made an absolute top in June, analogous to ARKK in February 2021 or the Nasdaq in March 2000.
  • •OpenAI and Anthropic account for an estimated 70% of AI revenues at Google, Amazon, and Microsoft, creating massive customer concentration risk.
  • •AI model cost-per-performance is dropping rapidly (estimated 99.96% decline over the next year), which will outpace Goldman's projected 24x growth in token consumption by 2030.
  • •Google's recent Q2 EPS increase was heavily distorted by a $98 billion mark-to-market gain on SpaceX and Anthropic, adding over $6.50 per share to S&P index earnings.
  • •Major tech companies including Google, Microsoft, Amazon, Meta, and Oracle have accumulated an estimated $1.65 trillion in hidden debt to fund AI capex.
  • •Sentimentrader’s Optix Index for precious metals shows extreme washed-out positioning (readings below 35.9), historically a bullish signal for gold and silver.
Read time 9 min
Length 9,331 chars
Category finance
Ideas
Geo Chen Global macro trader; ex-head of FX trading, Credit Suisse
Futures positioning, open interest, and price action point towards washed out positioning and apathetic sentiment in precious metals, with gold looking like it's about to break out of a base.
GLD HIGH
Geo Chen Global macro trader; ex-head of FX trading, Credit Suisse
Author believes SMH made an absolute top in June and compares its trajectory to ARKK in 2021 and the Nasdaq in 2000, as capital rotates out of AI leaders.
Author believes SMH made an absolute top in June and compares its trajectory to ARKK in 2021 and the Nasdaq in 2000, as capital rotates out of AI leaders. Risk: If AI revenues at model labs hold up and no macro shocks occur, SMH could digest sideways and make new highs in late 2026 or early 2027.
SMH
Geo Chen Global macro trader; ex-head of FX trading, Credit Suisse
Google's earnings quality is deteriorating, with $6.50 of its recent EPS increase coming from mark-to-market gains on Anthropic and SpaceX rather than core operations, while it sinks up to $40B into A
Google's earnings quality is deteriorating, with $6.50 of its recent EPS increase coming from mark-to-market gains on Anthropic and SpaceX rather than core operations, while it sinks up to $40B into Anthropic to round-trip cloud revenue. Risk: Continued market willingness to value Other Income & Expenses (OI&E) mark-to-market gains as core earnings could sustain the stock price.
GOOGL
Geo Chen Global macro trader; ex-head of FX trading, Credit Suisse
Microsoft's Azure AI revenue is highly dependent on OpenAI, an unprofitable lab that requires tens of billions in funding to survive, exposing MSFT to severe counterparty and circular financing risks.
Microsoft's Azure AI revenue is highly dependent on OpenAI, an unprofitable lab that requires tens of billions in funding to survive, exposing MSFT to severe counterparty and circular financing risks. Risk: OpenAI could successfully monetize enterprise agents, driving legitimate token consumption that outpaces compute cost deflation.
MSFT
Geo Chen Global macro trader; ex-head of FX trading, Credit Suisse
Amazon is heavily exposed to the circular financing loop, having funneled $5 billion into Anthropic to buoy AWS revenues while taking on massive hidden debt to fund AI capex.
Amazon is heavily exposed to the circular financing loop, having funneled $5 billion into Anthropic to buoy AWS revenues while taking on massive hidden debt to fund AI capex. Risk: AWS's non-AI cloud business could accelerate enough to mask any potential deceleration in Anthropic-driven revenues.
AMZN
More from Fidenza Macro
July 2026 - A turning point in the AI supercycle
Jul 30, 11:29
AI vibe shift, broad market views
Jul 14, 06:24
Turning neutral on AI semis and infrastructure
Jun 23, 10:15
Thoughts on Kevin Warsh's first FOMC meeting
Jun 18, 08:53
Global macro is getting interesting again
Jun 17, 10:01

This newsletter, published August 05, 2026, features Geo Chen discussing GLD, SMH, GOOGL, MSFT, AMZN. 5 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Geo Chen  · Tickers: GLD, SMH, GOOGL, MSFT, AMZN