The Real Risk Is No Longer AI Valuation
It Is the Rising Cost of Capital
September 4, 2026
---
Executive Summary
The market conversation is still dominated by AI valuations, Fed cuts, and equity performance.
But the more important development is happening underneath:
Global capital is becoming more expensive at the same time that governments, corporations, and the AI industry all need more of it.
This creates a different type of risk from 2008.
The potential problem is not necessarily that AI demand collapses.
It is that:
AI demand remains strong while the financial system supporting that demand becomes increasingly expensive to fund.
At the same time:
- US fiscal deficits continue to generate enormous Treasury supply.
- Japanese long-term yields have moved toward historically important levels.
- Japan is actively trying to support the yen.
- Oil is approaching $100.
- Inflation remains a constraint on central-bank easing.
- AI infrastructure requires enormous amounts of capital.
- Governments are competing with private-sector borrowers for that capital.
The emerging risk is therefore not simply an AI bubble.
It is a potential global capital-cost repricing.
---
1. The Bond Market Is Becoming the Main Variable
The most important change in the current environment is that long-term government bond yields are no longer behaving like a passive backdrop for equities.
They are becoming a source of risk themselves.
The recent global bond selloff has involved:
- US Treasuries
- Japanese Government Bonds
- UK Gilts
- European sovereign bonds
The common factor is increasingly clear:
Markets are demanding more compensation for holding long-duration government debt.
That compensation comes in the form of higher yields.
And once long-term yields rise, the consequences spread across the entire financial system.
Higher yields mean:
Government borrowing costs ↑
Corporate borrowing costs ↑
AI financing costs ↑
Equity discount rates ↑
Asset valuations ↓
The critical point is that this transmission mechanism works even if corporate earnings remain strong.
---
2. Japan May Be the Most Important Market to Watch
Japan is no longer simply a USD/JPY story.
It is increasingly becoming a global duration story.
The Japanese 10-year government bond yield has moved toward and above 3%, a level that would have seemed extraordinary under the previous Japanese monetary regime.
At the same time, Japan faces:
- rising government spending
- increasing debt-service costs
- higher required yields
- pressure to support the yen
- potential monetary tightening
This creates a difficult policy triangle.
Japan wants:
A stronger yen
but also:
low government borrowing costs
while simultaneously maintaining:
fiscal stimulus
Those objectives are increasingly difficult to achieve simultaneously.
---
3. Japan Has Already Demonstrated Its Willingness to Intervene
Japan has deployed enormous resources to support the yen.
Recent data indicate intervention on a historically large scale, with roughly $96 billion spent over the relevant period.
Yet USD/JPY has remained extremely high.
This creates a much more important question than:
"Will Japan intervene again?"
The question is:
"How much capital can Japan deploy before the market begins to trade against the intervention itself?"
Intervention can change the exchange rate temporarily.
But it does not automatically solve:
- Japan's fiscal position
- the interest-rate differential
- Japanese inflation
- JGB supply
- the underlying demand for dollars
If intervention repeatedly fails to produce a durable change in USD/JPY, the market may eventually interpret intervention as evidence of policy weakness rather than policy strength.
That is when intervention becomes progressively more expensive.
---
4. The JGB → Yen → Treasury Connection
This is one of the most important cross-asset relationships.
Japan has historically been one of the world's largest pools of capital.
If Japanese domestic bonds become increasingly attractive while currency risk remains significant, Japanese institutions have less incentive to buy foreign bonds.
The potential sequence is:
JGB yields ↑
↓
Japanese domestic bonds become more attractive
↓
Foreign bond demand ↓
↓
US Treasury demand potentially ↓
↓
US term premium ↑
↓
US long-term yields ↑
This does not require Japan to dump US Treasuries aggressively.
Even a marginal reduction in demand can matter when the Treasury market must absorb enormous amounts of new supply.
This is why Japan's bond market deserves attention far beyond Japan.
---
5. Oil Makes the Problem Worse
Oil near $95 changes the equation.
Higher oil prices can simultaneously create:
Higher inflation + weaker growth
That is the classic stagflation problem.
The transmission mechanism is straightforward:
Oil ↑
↓
Transportation costs ↑
Energy costs ↑
Production costs ↑
↓
Inflation ↑
But simultaneously:
Oil ↑
↓
Consumer purchasing power ↓
Corporate margins ↓
Economic activity ↓
This creates a particularly uncomfortable environment for central banks.
If growth deteriorates, the market wants rate cuts.
But if oil keeps inflation elevated, aggressive rate cuts become more difficult.
---
6. This Is Why "Fed Will Cut" Is Not Enough
The equity market can still rally when investors believe the Fed will eventually cut rates.
But there is an important distinction:
The Fed controls short-term policy rates.
It does not directly control the long-term price the market demands to finance the US government.
That distinction matters enormously.
The Fed can cut the policy rate while:
10Y yield remains high
30Y yield remains high
term premium remains elevated
Treasury supply remains enormous
Therefore:
Fed cuts do not automatically mean cheap capital.
This is one of the biggest differences between today's environment and the post-2008 world.
---
7. AI Is Becoming a Capital-Market Story
AI demand can remain extremely strong while the AI capital cycle becomes increasingly fragile.
This distinction is critical.
NVIDIA can have:
- real customers
- real revenue
- real cash flow
- real demand
while the broader AI infrastructure ecosystem simultaneously develops:
- enormous CapEx requirements
- enormous debt issuance
- aggressive utilization assumptions
- expensive data-center financing
- dependence on future AI revenue
Therefore:
A real technology does not prevent a financial bubble.
The technology can be real.
The demand can be real.
The investment can still become excessive.
---
8. The AI Capital Cycle
The AI ecosystem increasingly looks like this:
AI demand
↓
GPU demand
↓
Data-center construction
↓
Electricity demand
↓
Networking
↓
Real estate
↓
Infrastructure financing
↓
Debt issuance
↓
Credit markets
↓
Capital costs
The AI story therefore increasingly depends on the bond market.
This is why the question:
"Is AI real?"
is becoming less useful.
The better question is:
"How much capital is required to monetize AI, and what happens if that capital becomes significantly more expensive?"
---
9. Debt-Supported Revenue Is the More Dangerous Dynamic
One of the most important distinctions in the current cycle is between:
Debt-created profit
and
Debt-supported revenue.
If a company borrows money and invests it productively, the debt can eventually be repaid through genuine cash generation.
But if an entire ecosystem requires continuous financing to sustain revenue growth, the system becomes dependent on capital-market conditions.
That creates a dangerous feedback loop:
More financing
↓
More CapEx
↓
More capacity
↓
More revenue expectations
↓
Higher valuations
↓
More financing
This can work extremely well while capital remains cheap.
The problem begins when the financing assumption changes.
---
10. Why the Stock Market Can Keep Going Up
This is one of the most confusing aspects of the current environment.
The bond market can be under pressure while equities continue to rally.
The explanation is that they are pricing different things.
The equity market is looking at:
Future earnings
The bond market is increasingly looking at:
Future capital supply and government financing requirements
Therefore:
Stocks are still betting on the future.
Bonds are increasingly repricing the cost of reaching that future.
These two views can coexist for a surprisingly long time.
Eventually, however, the equity market has to incorporate the cost of capital.
---
11. Gold Is Changing Its Role
Gold is increasingly interesting because it is no longer simply an inflation hedge.
It can also function as a hedge against:
- sovereign debt risk
- currency debasement
- fiscal instability
- geopolitical risk
- declining confidence in fiat purchasing power
This creates an important distinction.
Gold does not necessarily need falling real yields to perform well.
If investors begin asking:
"How will governments ultimately deal with their debt?"
the demand for assets outside the sovereign-credit system can increase.
Gold therefore becomes something closer to:
A sovereign credibility hedge.
---
12. The Emerging Feedback Loop
The biggest risk is not any single market.
It is the interaction between them.
US fiscal deficit
↓
Treasury issuance
↓
Long-term yields ↑
↓
Capital costs ↑
↓
AI financing costs ↑
↓
AI CapEx becomes more expensive
↓
Risk premium ↑
↓
Equity valuations ↓
At the same time:
Japan fiscal pressure
↓
JGB yields ↑
↓
BOJ tightening pressure
↓
JPY ↑
↓
Japanese capital repatriation
↓
Foreign bond demand ↓
↓
US Treasury yields ↑
And then:
Oil ↑
↓
Inflation ↑
↓
Central banks have less room to cut
↓
Long-term yields remain elevated
The three systems reinforce each other.
---
13. Three Scenarios
🟢 Scenario 1 — Temporary Stabilization
The Fed becomes more dovish.
Rate-cut expectations rise.
Bond yields decline.
Equities recover.
USD weakens.
JPY strengthens.
Gold remains strong.
This is the most obvious short-term relief scenario.
But it does not necessarily solve the structural problem.
---
🟡 Scenario 2 — Higher-for-Longer Long-Term Yields
This may be the most important scenario.
Oil remains around:
$90–100
while:
- US fiscal deficits remain large
- Treasury issuance remains high
- JGB yields remain elevated
- AI infrastructure continues requiring capital
The result:
Long-term yields remain structurally higher.
The Fed may cut.
But the long end refuses to return to the old regime.
That would represent a major change in the cost of capital.
---
🟠 Scenario 3 — Global Duration / Credit Unwind
The more dangerous sequence is:
Bond selloff
↓
Borrowing costs ↑
↓
Fiscal deficit worsens
↓
More debt issuance
↓
More bond selling
↓
Credit spreads ↑
↓
AI financing deteriorates
↓
CapEx slows
↓
Equity risk premium ↑
↓
Asset prices ↓
At this point the problem changes category.
It stops being:
A bond-market correction
and becomes:
A credit-market repricing.
That is where the 2008 comparison becomes more relevant.
---
14. But This Is Not 2008 Yet
The distinction is important.
2008
The dominant problem was:
Private-sector leverage + housing + banking-system credit collapse + deflation
Today's potential problem is different:
Sovereign debt + fiscal deficits + duration + inflation + AI capital intensity + geopolitical energy risk
The system could therefore experience a very different type of crisis.
The danger is not necessarily:
"Banks suddenly stop lending."
It could instead begin with:
The price of capital keeps rising until too many economic assumptions stop working.
---
15. The Seven Numbers That Matter
Instead of focusing exclusively on the S&P 500 or NVIDIA, watch these:
| Indicator | Why it matters |
|---|---|
| US 10Y | Determines the baseline cost of capital |
| US 30Y | Measures long-duration fiscal pressure |
| JGB 10Y | Japan's domestic funding regime |
| USD/JPY | Measures pressure on Japan's monetary system |
| Oil | Determines the inflation constraint |
| Gold | Measures demand for monetary/sovereign alternatives |
| AI Credit / CapEx | Measures the financial intensity of the AI boom |
The particularly important combination is:
JGB 3%+ + US 30Y ~5%+ + Oil ~$95–100
If all three continue moving higher together, the issue becomes much larger than an ordinary equity correction.
---
16. The Bigger Thesis
The market is still asking:
"Is AI a bubble?"
That may be the wrong question.
The more important question is:
"Can the global financial system continue funding everything that the current economic structure expects to build?"
Governments need capital.
AI needs capital.
Data centers need capital.
Energy infrastructure needs capital.
Corporations need capital.
Consumers need capital.
And now governments are themselves competing aggressively for that capital.
The fundamental constraint may therefore become:
Not technology.
Not demand.
Not even liquidity.
But:
The Price of Capital.
---
Final Thought
The most dangerous environment is not necessarily one in which AI collapses.
It is one in which:
AI remains strong,
government borrowing remains enormous,
oil remains elevated,
long-term yields remain high,
Japan begins competing for domestic capital,
and
the Fed cannot simply suppress long-term yields indefinitely.
In that environment, the world can continue to grow while the financial architecture supporting that growth becomes progressively more fragile.
That is the scenario worth watching.
Because the next major repricing may not begin with:
"AI earnings are disappointing."
It may begin with:
"There is simply too much capital demand at the same time."
And once the market starts pricing that reality, AI, equities, bonds, currencies, commodities, and sovereign debt stop being separate trades.
They become one system.