Investors are simultaneously chasing AI-driven equity growth and piling into high-quality fixed income, according to FTSE Russell's September 2026 Asset Allocation Insights report. This barbell dynamic has significant implications for how advisors build and stress-test client portfolios right now.
Sovereign bond yields have reached post-global financial crisis highs across the US, UK, and Europe, driven primarily by a rise in real yields rather than inflation expectations. That distinction points to a structural reset in the price of capital, not a cyclical inflation scare that central banks can simply wait out.
AI capex is rewriting the investment-grade credit market
Hyperscaler AI capital expenditure is now estimated to exceed $1 trillion in 2027, with Nvidia guiding to 70% revenue growth for fiscal year 2028. That level of spending has direct downstream effects on the bond market that advisors holding diversified fixed income allocations may not have fully priced in.
The five largest US hyperscalers (Amazon, Alphabet, Meta, Microsoft, and Oracle) have seen their share of USD investment-grade bond issuance climb to 12.9% year-to-date in 2026, up from just 1.7% in 2024. Their share within the industrial sector of the IG market has surged to 23.6%. At the same time, utilities are issuing more debt to fund the power infrastructure that AI data centers require, adding further supply pressure.
Total USD IG issuance is running 27% higher year-over-year in 2026, compared with just 11.7% growth for high yield. If that pace holds, the composition of standard IG benchmarks and the index-tracking portfolios built around them will shift gradually, but materially.
As asset managers pivot to outsourcing and AI, the debt-fueled AI buildout is raising scrutiny around balance sheet quality at some of the biggest names in technology, with free-cash-flow margins narrowing even as revenue growth remains strong. Free-cash-flow margins have turned negative for four of the five major hyperscalers as anticipated capex commitments outpace near-term cash generation.
The real yield story and what it means for duration
The 30-year Treasury yield hit a post-GFC high of 5.3% in August 2026 after the US Treasury Department announced it would double its long-dated debt buyback program; a move the market interpreted as an attempt to suppress long-end yields rather than a pure liquidity operation.
According to the OECD Global Debt Report, interest expense across OECD countries reached 3.3% of GDP in 2025, with rising yields and expanded issuance expected to add a further 2.5 percentage points to that figure in 2026. Governments borrowing more at higher yields is the structural force driving real rates upward — and that dynamic is not dissipating quickly.
FTSE Russell's fixed income research has found a strong historical correlation between starting yields and forward returns, with correlation coefficients as high as 0.95 for two-year forward returns on short-duration Treasury indexes. The implication for advisors is that elevated entry points in government bonds look historically attractive — but the path to those returns is not without risk, given how heavily the market is pricing in further rate hikes.
With the August nonfarm payroll report coming in at 162,000 jobs and headline US CPI running at 3.4%, market-implied probability of a September rate hike has risen to approximately 94%, according to FTSE Russell's analysis of data as of September 14, 2026. This has led to rate hike bets reshaping ETF demand as advisors eye short duration and alternatives.
Barbell flows and concentration risk
Fund flow data over both three-month and six-month windows reveals a barbell pattern: capital is flowing simultaneously into US and global growth equity, particularly technology and AI-adjacent names, and into high-quality fixed income. It is not a straightforward risk-on or risk-off rotation.
Within equity markets, concentration risk is at or near historic highs. As of August 2026, the US represents 61.7% of FTSE All-World market capitalization, and technology alone accounts for 44.1% of the FTSE USA index, a weighting that now exceeds the combined weight of nine other industries including financials, healthcare, energy, and real estate.
The FTSE Russell report notes that investors holding broad US equity index exposure may be taking on unintended, concentrated bets on AI and sensitivity to a single factor, and flags options such as broader regional diversification, equally weighted indexes, or RIC-capped structures as tools to address the issue. This is particularly relevant as direct indexing adoption climbs to 83% among U.S. advisors.
Credit spreads: the quiet risk
High-yield credit spreads tightened to a 10-year low of 260 basis points in August 2026, while world IG spreads sit at approximately the fifth percentile of their 10-year range.
That is a historically thin cushion. Advisors relying on credit allocations for yield pickup over Treasuries should weigh those spreads against a government bond market that is now offering historically attractive starting yields in its own right.
The FTSE Russell report describes asset class "yields" as currently favoring government bonds over equities, and equities over credit, based on each asset class's position relative to its own 10-year history. That ordering is a significant shift from the post-GFC playbook that defined most of the last decade.
For advisors, the September 2026 picture is one of genuine complexity: the best absolute yields in government bonds since before the financial crisis, equity valuations doing heavy lifting on earnings growth expectations, and a credit market priced for perfection at exactly the moment that AI-driven debt supply is accelerating.
The barbell that global fund flows reflect may be less a strategy than a hedge and advisors building portfolios around it will need an explicit view on duration, concentration, and how long the AI capex cycle can sustain the earnings revisions currently supporting equity multiples.
