Gundlach and Zulauf on the Shift: Late-Cycle AI Capex and Liquidity Cracks
In DoubleLine Capital’s conversation with Jeffrey Gundlach and Felix Zulauf, the message is not simply bearish. It is more precise: AI capex still powers the growth story, but the liquidity backdrop is becoming more fragile through long rates, the dollar, and private credit.
Start with the conclusion
AI infrastructure remains the strongest growth engine in public markets.
U.S. fiscal pressure, long rates, and dollar confidence may not cushion markets as they did in earlier cycles.
Private credit is framed as a possible next source of credit stress.
The Signal & Flow reading is simple: AI is still the growth engine, but liquidity is already showing cracks in rates, the dollar, and credit.
The point is not to declare that AI is over or that equities must fall immediately. The better question is whether AI growth can keep validating today’s prices while liquidity conditions stay supportive enough to carry them.
1. The political layer: multipolarity is inflationary
Zulauf’s big-picture argument is that the world is moving from a U.S.-led unipolar order toward a multipolar regime. That transition tends to bring wars, sanctions, supply-chain separation, military spending, and energy-security concerns.
His phrase is straightforward: wars are inflationary and sanctions are inflationary. Globalization reduced costs by moving production to the cheapest and most efficient places. A geopolitically fragmented world often chooses redundancy and security over pure efficiency.
2. The economic layer: long rates may not fall the way they used to
Gundlach’s core macro argument is that the secular decline in long-term U.S. Treasury yields is over. The reason is fiscal math. Interest expense is much larger than it was in the previous cycle, and deficits remain heavy even without a recession.
The implication is important. In earlier cycles, recession risk usually pushed long rates lower, which supported both bonds and equity valuations. If fiscal pressure limits that decline, long-duration assets may not receive the same automatic discount-rate relief.
| Old reaction | New risk | Investor question |
|---|---|---|
| Dollar strengthens in risk-off | The dollar may weaken if U.S. fiscal confidence erodes | Does gold outperform the dollar in stress? |
| Long rates fall in recession | Deficits and Treasury supply may limit the decline | Does the 10-year yield stay firm during slowdown? |
| Treasuries cushion portfolios | Long bonds may become a volatility source | Do bonds actually hedge equity drawdowns? |
3. The technology layer: AI capex is real, but late-cycle signals are visible
Zulauf sees the AI and capex-driven upswing as late-cycle. He does not claim to know the exact top, but he suggests the market could peak sometime between the third quarter of 2026 and the first quarter of 2027, followed by a recessionary bear market rather than a shallow valuation-only correction.
This is not an anti-AI argument. AI is still the growth axis. The issue is that the growth requires more capex, power, GPUs, memory, data centers, financing, and prepayment structures.
Oracle’s FY 2026 release shows both sides of the story. Oracle highlighted strong cloud infrastructure growth and large AI contracts, but also reported negative free cash flow of about $23.7 billion for FY 2026 as it invested heavily in cloud infrastructure.
4. S&P 500 concentration: broad index, narrow bet
Gundlach points to the extreme concentration of U.S. equities. RBC Wealth Management similarly notes that by the end of 2025 the top 10 companies represented nearly 41% of the S&P 500’s total weight.
Concentration alone does not prove a bubble. Great companies can deserve large weights if earnings power keeps growing. The problem is that many investors believe they own a broad market while they are increasingly exposed to a small group of AI and platform leaders.
| Signal | Constructive read | Risk read | What to watch |
|---|---|---|---|
| Megacap AI concentration | Winners are earning their weight | The index depends on one theme | Top-10 earnings share versus index weight |
| Cap-weighted dominance | Quality leaders drive returns | Market breadth weakens | Equal-weight S&P 500, QQQ breadth, semis relative strength |
| AI capex surge | Long-duration productivity infrastructure | Free-cash-flow and overinvestment risk | Cloud revenue, RPO, FCF, depreciation, power bottlenecks |
5. Private credit: the next liquidity accident candidate
The strongest warning in the latter part of the conversation is private credit. Gundlach compares the tone to the 2005–2006 credit cycle. The claim is not that every private-credit asset is bad. The issue is opacity.
He points to volatility smoothing, inconsistent loan marks, small private rating agencies, limited redemption mechanics, and the complex links among private equity, private credit, insurance companies, and offshore reinsurance.
6. Investment map: separate company, price, and timing
Companies that can convert AI infrastructure bottlenecks into revenue, margins, and free cash flow. Look for pricing power, long-term contracts, power and data-center execution, and customer prepayment structures.
High-quality AI leaders where the narrative may already be heavily priced. Entry discipline requires earnings validation, rate stability, and pullback support.
Private-credit, leveraged finance, and AI-capex burdened balance sheets. These require observation until stress is visible in prices and disclosures.
Extreme Treasury restructuring or major policy shifts should be treated as scenarios unless confirmed by official action and market pricing.
Warning signals
- AI leaders fail to convert capex into revenue, margins, and free cash flow.
- The 10-year Treasury yield refuses to fall during economic slowdown.
- The dollar weakens in risk-off while gold strengthens.
- Private-credit NAV markdowns, redemption limits, or insurance linkages become more visible.
- Only megacap AI rises while equal-weight indexes and market breadth deteriorate.
Investor bias check
- FOMO: real AI growth does not justify chasing any price.
- Safe-asset bias: long Treasuries and the dollar may not hedge exactly as they did in prior cycles.
- Narrative bias: “AI capex is always good” and “AI capex is always a bubble” are both too simple.
- Confirmation bias: a strong macro warning does not mean every risk asset deserves the same conclusion.
Reader checklist
- Is AI capex turning into revenue, margins, and free cash flow?
- How do long rates and the dollar behave during risk-off periods?
- Is S&P 500 performance dependent on a small group of megacap AI names?
- Are private-credit markdowns, redemption limits, or insurance links becoming more visible?
- Is your portfolio overly concentrated in U.S. megacap AI growth?
- Are you separating a good company from a good entry price?
Final view
This conversation is best read as a structural warning for the 2026–2027 transition, not as a precise crash call. AI remains the most powerful growth axis. But liquidity is becoming more complicated than in the previous cycle.
If long rates, the dollar, private credit, and AI-capex cash-flow pressure all move the wrong way, markets may shift from “growth is real” to “the price moved too fast.” If AI earnings and free cash flow continue to validate while rates and credit stay stable, the cycle can extend.
The discipline is therefore clear: keep the long-term AI growth thesis alive, but avoid leverage and chase-buying. Separate company quality, valuation, and timing.
Telegram: https://t.me/signalandflow