Check your bond sleeve. You might be hedging yourself.
The five AI hyperscalers issued $219 billion of investment-grade debt this year to fund data centers. US corporate issuance hit $1.68 trillion through August, up 27%.
Bank of America's math: at this pace, Alphabet, Amazon, Meta, Microsoft and Oracle will be 6.3% of the entire US investment-grade market by end of 2027. They were 3% at end of 2025.
Doubling in two years.
So an investor holding tech equities and bonds for safety now holds tech on both sides.
Worth being precise about what this does and doesn't mean. This is AA and A rated paper. The credit quality is among the best in the index, and issuer overlap is not the same as risk overlap.
But it matters in one specific scenario: if the stress is about the AI capex cycle itself, both sleeves reprice on the same news. That's a different failure mode than 2022, which was a rates shock hitting everything at once through the discount rate.
Meanwhile stock-bond correlation turned positive in March and reached a multi-decade high in June, so the two sleeves are already moving together for unrelated reasons.
None of this makes bonds uninvestable. A 4.8% ten-year and a 5.27% thirty-year is normal by historical standards, and there's real income here for the first time in years.
What's weaker is the assumption underneath 60/40 that the sleeves are independent. They're less independent than they've been in decades, and increasingly made of the same companies.
$NVDA $GOOGL $SPY
The five AI hyperscalers issued $219 billion of investment-grade debt this year to fund data centers. US corporate issuance hit $1.68 trillion through August, up 27%.
Bank of America's math: at this pace, Alphabet, Amazon, Meta, Microsoft and Oracle will be 6.3% of the entire US investment-grade market by end of 2027. They were 3% at end of 2025.
Doubling in two years.
So an investor holding tech equities and bonds for safety now holds tech on both sides.
Worth being precise about what this does and doesn't mean. This is AA and A rated paper. The credit quality is among the best in the index, and issuer overlap is not the same as risk overlap.
But it matters in one specific scenario: if the stress is about the AI capex cycle itself, both sleeves reprice on the same news. That's a different failure mode than 2022, which was a rates shock hitting everything at once through the discount rate.
Meanwhile stock-bond correlation turned positive in March and reached a multi-decade high in June, so the two sleeves are already moving together for unrelated reasons.
None of this makes bonds uninvestable. A 4.8% ten-year and a 5.27% thirty-year is normal by historical standards, and there's real income here for the first time in years.
What's weaker is the assumption underneath 60/40 that the sleeves are independent. They're less independent than they've been in decades, and increasingly made of the same companies.
$NVDA $GOOGL $SPY
