Skip to content

Credit Creation Regime Change: From Central Banks to Private Sector

English Summary

Andy Constan argues that the market is undergoing a fundamental regime change from central bank-dominated liquidity provision (2010s) to private sector credit creation (2020s+), driven by massive "promises" in AI capex, onshoring, and fiscal deficits that far exceed what commercial banks can fund through money creation alone—forcing asset sales and fundamentally altering asset price dynamics in ways most investors haven't yet grasped.

The Regime Shift: Push vs Pull

The 2010s problem: Banks were able to lend (Fed stuffed them with reserves via QE) but not willing (no demand). "Pushing on a string."

The 2020s solution: Real demand has emerged (AI capex, onshoring promises) creating pull for credit. Banks want to be price makers, not price takers—they lend when borrowers pay up for it.

The key difference: - Fed QE era: Money creation without credit demand → asset inflation, minimal real economy impact - Private credit era: Credit demand without sufficient money creation → real economy stimulus, asset sales to fund

Three Massive "Promises" That Need Funding

  1. AI Capex
  2. Hyperscalers (Oracle, Meta, Google, etc.) issuing corporate bonds to build data centers
  3. Nvidia, utilities, construction all getting paid

  4. Order books are full, spending is happening NOW

  5. Onshoring / Foreign Direct Investment (FDI)

  6. South Korea, Saudi Arabia, Japan promising to build manufacturing capacity in US
  7. Tied to Trump tariff negotiations (SCOTUS decision pending)

  8. Not economically efficient—pure "insurance cost" for national security + supply chain resilience

  9. Federal Deficit

  10. Ongoing fiscal gap needs continuous funding
  11. Treasury issuance competing for same capital pool

Total promises >> what a stable economy normally needs

Money Creation vs Credit Creation: The Critical Distinction

Money Creation (Fed or Commercial Banks): - Creates spendable deposits out of thin air - Fed: Buys financial assets → creates reserves + deposits - Commercial banks: Make loans → creates deposits instantly (borrower can spend) + loan asset (promise to repay) - Impact: Stimulative + inflationary, debases all assets (stocks, bonds, gold, crypto, real goods)

Credit Creation (Private Sector): - Does NOT create new money - You lend me your bank deposit → I can spend, you can't → money "activated" - The circular loop: - T0: Borrower issues debt (e.g., Oracle corporate bond) - Money spent in real economy (Nvidia chips, utilities, construction wages) - T1: All spending becomes someone's savings - Savings fund the original debt issuance - No new money needed for the circle to work

The timing problem: Between T0 and T1, there's massive supply of assets that must be absorbed. This forces: - Asset sales to free up bank deposits - Rising interest rates / credit spreads as lenders demand higher yield for increased risk - Potential cascade: Sell BTC → fund redemptions → sell stocks → tighten margin → more selling

Why Banks Can't Save Us This Time

US banking system capacity: - ~10% capital-to-assets ratio (20 trillion in assets) - At most $2 trillion of credit creation capacity available - Problem: Promises (AI + onshoring + deficit) far exceed $2T - At 8.5% capital ratio = severe stress

Implication: These promises WILL be funded via credit creation (asset sales), NOT money creation (bank loans).

Asset Price Implications: The New Bifurcation

❌ Bearish for Bonds: - Massive corporate bond issuance (AI capex) - Government bond sales to fund FDI (foreigners selling treasuries to invest in US factories) - Credit spreads widening (Oracle bonds out 80bps from tights) - Real economy spending (not financial assets) → rates rise

✅ Bullish for Equities (conditional): - Real economy spending stimulates GDP - 10% earnings growth forecast (consensus) - Even with 2-point multiple contraction, stocks still up 8% - Growth-positive environment

❌ Bearish for Speculative Assets (Gold, Crypto): - No longer benefit from "asset inflation" (money creation going into financial assets) - Money flowing into real economy, not speculation - Crypto = "tip of the spear for liquidity conditions" (plur_daddy quote echoed here)

The paradigm shift: - 2010s: Central bank money creation → asset inflation → all assets up - 2020s: Private sector credit creation → real economy investment → stock/bond bifurcation

The AI Capex Problem: Job Destruction, Not Creation

Traditional investment cycle: 1. Build factory 2. Hire workers 3. Workers have jobs → consumer confidence 4. Consumers lever up (mortgages, cars, credit cards) 5. Consumption drives more investment 6. Cycle continues until rates too high

AI investment cycle: 1. Build data center 2. Replace workers (AI automates jobs) 3. Workers lose confidence 4. Consumers DON'T lever up 5. ??? (cycle breaks)

Historical precedent: Every major technology (railroads, electricity, internet, Excel) caused job displacement fears, but people always "figured it out" and transitioned to new productive work.

This time different? Andy is "hesitant to say this time is different," but acknowledges: - AI may significantly reduce jobs - Could impact desire to procreate → population concerns - Transition period could be long and painful

Key concern: If AI doesn't create offsetting jobs, consumers won't lever up for consumption. This means: - Only AI capex pulling credit (not broad economy) - In 2-3 years, need either: - Path A: AI successfully transitions workers to new jobs - Path B: Government steps in (UBI, etc.) - If neither happens → credit crunch when projects fail to pay off

The Onshoring Problem: Economic Inefficiency

Why onshoring hasn't happened: US manufacturing is too expensive. If it were economically efficient, companies would already be doing it.

What forcing it means: - Worse product for same money, OR - Same product for more money - It's pure insurance cost (national security, supply chain resilience) - Not economically productive

The FDI funding problem: - Foreign companies (Hyundai, Saudi firms) previously chose not to invest in US → revealed preference that it's a bad investment - Now promising to do so because of Trump tariff threats - Must sell US treasuries to fund → short-term treasury supply pressure - Will they actually follow through? Uncertain.

Market Outlook: The "Nirvana" vs "Likely" Scenarios

Nirvana Scenario (low probability): 1. All spending/investment gets funded at current interest rates 2. Projects pay off rapidly and successfully 3. Creates jobs (or at least doesn't destroy them) 4. Consumer confidence → consumers lever up 5. Virtuous cycle: savings reinvest → banks repaid → more lending 6. Result: Stocks rally, bonds flat to down modestly

More Likely Scenario: 1. Interest rates rise as supply hits market 2. Higher rates → projects canceled 3. No offsetting consumer confidence boost (AI destroying jobs) 4. Fed eases, but not enough to offset earnings disappointment 5. Result: Stocks struggle, bonds struggle, recession risk

Current Reality (H1 2026): - Spending is coming (order books full) - Growth will be strong initially - Watch for: When does market discount the END of spending? - Oracle bond spreads already widening (80bps) → data center stocks crushed (CoreWeave, etc.)

Foreign Capital Flows: The Hidden Mechanism

Why has US had "infinite bid" for debt? - US runs trade deficit → buys foreign goods - Foreign countries get dollars → buy US assets (treasuries, stocks) - As long as trade deficit persists, there's structural bid

What changes with FDI promises? - FDI is additional promise (not just recycling trade deficit) - Foreigners must sell existing US treasury holdings to fund factory investments - Short-term: Treasury supply pressure, rates rise - Alternative: Large dollar depreciation could incentivize (make US manufacturing cheaper)

The credibility question: Will they actually do it? History says no (they already had the option and declined).

What to Watch: Tracking the Regime Shift

  1. Corporate bond issuance & spreads
  2. Oracle bond spread movement

  3. Credit market capacity to absorb

  4. Who's funding the promises?

  5. Bank lending data
  6. Asset sales (treasuries, equities, crypto)

  7. Foreign reserve changes

  8. Spending velocity

  9. AI capex announcements vs actual deployment

  10. FDI commitments vs groundbreakings

  11. Labor market signals

  12. Job displacement from AI (leading indicator of consumption)
  13. Consumer confidence

  14. Credit card / mortgage applications

  15. When does market discount the end?

  16. Forward earnings guidance
  17. Order book deterioration
  18. Project cancellations (interest rate sensitivity)

The Bottom Line: Fighting the Last War

Most investors are stuck in 2010s playbook: - Obsessing over Fed balance sheet flows - Waiting for QE to save everything - Expecting all assets to inflate together

New playbook (if Andy is right): - Fed sidelined unless crisis - Private sector credit creation is the game - Stock/bond bifurcation (not "rising tide lifts all boats") - Real economy matters again - Traditional business cycle returns

The risk: This regime shift requires promises to be funded AND pay off. If either fails, it's not just a correction—it's a credit crunch with limited Fed ammunition.

繁體中文總結

Andy Constan 主張市場正在經歷根本性的結構轉變，從央行主導的流動性供應（2010s）轉向私人部門信用創造（2020s+），由 AI 資本支出、onshoring、財政赤字創造的大量「承諾」遠超商業銀行透過貨幣創造所能資助的規模——迫使資產拋售並從根本上改變資產價格動態，而大多數投資者尚未意識到這一點。

結構性轉變：推動 vs 拉動

2010s 的問題： 銀行有能力放貸（Fed 透過 QE 塞滿準備金）但不願意（沒有需求）。「推繩子」。

2020s 的解方： 真實需求已出現（AI capex、onshoring 承諾）創造信用的拉力。銀行想當價格制定者，不是價格接受者——當借款人願意支付更高利率時，他們才放貸。

關鍵差異： - Fed QE 時代： 貨幣創造但無信用需求 → 資產通膨，實體經濟影響極小 - 私人信用時代： 信用需求但貨幣創造不足 → 實體經濟刺激，必須賣資產來資助

三大需要資助的「承諾」

  1. AI Capex
  2. 超大規模業者（Oracle、Meta、Google 等）發行企業債建數據中心
  3. Nvidia、公用事業、建設公司都在收錢

  4. 訂單簿已滿，支出正在發生

  5. Onshoring / 外國直接投資（FDI）

  6. 韓國、沙烏地阿拉伯、日本承諾在美國建製造產能
  7. 與 Trump 關稅談判綁定（SCOTUS 判決待定）

  8. 經濟上不效率——純粹的「保險成本」（國安 + 供應鏈韌性）

  9. 聯邦赤字

  10. 持續的財政缺口需要不斷資金
  11. 國債發行與同一資本池競爭

總承諾 >> 穩定經濟正常所需

貨幣創造 vs 信用創造：關鍵區別

貨幣創造（Fed 或商業銀行）： - 憑空創造可支配存款 - Fed：購買金融資產 → 創造準備金 + 存款 - 商業銀行：放貸 → 立即創造存款（借款人可支出）+ 貸款資產（承諾還款） - 影響： 刺激性 + 通膨性，稀釋所有資產（股票、債券、黃金、加密、實物商品）

信用創造（私人部門）： - 不創造新貨幣 - 你借我你的銀行存款 → 我可以花，你不行 → 貨幣「活化」 - 循環迴圈： - T0：借款人發債（例如 Oracle 企業債） - 資金在實體經濟支出（Nvidia 晶片、公用事業、建設工資） - T1：所有支出變成某人的儲蓄 - 儲蓄資助原始債務發行 - 不需要新貨幣讓循環運作

時機問題： T0 到 T1 之間，大量資產供應必須被吸收。這迫使： - 賣出資產來釋放銀行存款 - 利率/信用利差上升（放貸者要求更高收益以承擔增加的風險） - 潛在連鎖反應：賣 BTC → 資助贖回 → 賣股票 → 緊縮保證金 → 更多拋售

為什麼銀行這次救不了我們

美國銀行系統容量： - ~10% 資本資產比（20 兆資產） - 最多 2 兆美元 信用創造容量 - 問題： 承諾（AI + onshoring + 赤字）遠超 $2T - 8.5% 資本比 = 嚴重壓力

含義： 這些承諾將透過信用創造（資產拋售）資助，而非貨幣創造（銀行貸款）。

資產價格含義：新的分化

❌ 債券看空： - 大量企業債發行（AI capex） - 政府債券拋售以資助 FDI（外國人賣美債投資美國工廠） - 信用利差擴大（Oracle 債券從低點擴大 80bps） - 實體經濟支出（非金融資產）→ 利率上升

✅ 股票看多（有條件）： - 實體經濟支出刺激 GDP - 10% 盈餘成長預測（共識） - 即使 2 點本益比收縮，股票仍上漲 8% - 成長正向環境

❌ 投機資產看空（黃金、加密）： - 不再受惠「資產通膨」（貨幣創造流入金融資產） - 資金流入實體經濟，不是投機 - 加密 = 「流動性條件的矛尖」（呼應 plur_daddy 論點）

典範轉移： - 2010s： 央行貨幣創造 → 資產通膨 → 所有資產上漲 - 2020s： 私人部門信用創造 → 實體經濟投資 → 股債分化

AI Capex 問題：摧毀工作，而非創造

傳統投資週期： 1. 建工廠 2. 雇用工人 3. 工人有工作 → 消費者信心 4. 消費者槓桿化（房貸、車貸、信用卡） 5. 消費驅動更多投資 6. 週期持續直到利率太高

AI 投資週期： 1. 建數據中心 2. 取代工人（AI 自動化工作） 3. 工人失去信心 4. 消費者不槓桿化 5. ??? （週期中斷）

歷史先例： 每次重大技術（鐵路、電力、網際網路、Excel）都引發工作取代恐慌，但人們總是「想出辦法」轉型到新的生產性工作。

這次不同？ Andy「猶豫說這次不同」，但承認： - AI 可能顯著減少工作 - 可能影響生育意願 → 人口問題 - 轉型期可能漫長且痛苦

關鍵擔憂： 如果 AI 不創造抵銷性工作，消費者不會為消費槓桿化。這意味： - 只有 AI capex 拉動信用（非廣泛經濟） - 2-3 年後，需要： - 路徑 A： AI 成功將工人轉型到新工作 - 路徑 B： 政府介入（UBI 等） - 如果都沒發生 → 專案無法回報時信用緊縮

Onshoring 問題：經濟無效率

為何 onshoring 過去沒發生： 美國製造太貴。如果經濟上有效率，公司早就做了。

強制執行意味著： - 相同錢買更差產品，或 - 相同產品花更多錢 - 這是純保險成本（國安、供應鏈韌性） - 不具經濟生產力

FDI 資金問題： - 外國公司（Hyundai、沙烏地公司）過去選擇不投資美國 → 顯示偏好這是壞投資 - 現在因 Trump 關稅威脅而承諾 - 必須賣美債資助 → 短期國債供應壓力 - 他們真的會執行嗎？不確定。

市場展望：「天堂」vs「可能」情境

天堂情境（低機率）： 1. 所有支出/投資以當前利率資助 2. 專案快速且成功回報 3. 創造工作（或至少不摧毀） 4. 消費者信心 → 消費者槓桿化 5. 良性循環：儲蓄再投資 → 銀行還款 → 更多放貸 6. 結果： 股票上漲，債券持平到小跌

更可能情境： 1. 利率隨供應上升 2. 更高利率 → 專案取消 3. 無抵銷的消費者信心提升（AI 摧毀工作） 4. Fed 寬鬆，但不足以彌補盈餘失望 5. 結果： 股票掙扎，債券掙扎，衰退風險

當前現實（2026 上半年）： - 支出正在發生（訂單簿已滿） - 成長初期將強勁 - 關注： 市場何時 discount 支出結束？ - Oracle 債券利差已擴大（80bps）→ 數據中心股票崩潰（CoreWeave 等）

外國資本流動：隱藏機制

為何美國債務有「無限買盤」？ - 美國有貿易逆差 → 購買外國商品 - 外國獲得美元 → 購買美國資產（國債、股票） - 只要貿易逆差持續，就有結構性買盤

FDI 承諾改變什麼？ - FDI 是額外承諾（不只是回收貿易逆差） - 外國人必須賣出現有美債持倉來資助工廠投資 - 短期：國債供應壓力，利率上升 - 替代方案：大幅美元貶值可能激勵（使美國製造更便宜）

可信度問題： 他們真的會做嗎？歷史說不會（他們早有選項但拒絕）。

需要關注的指標：追蹤結構轉變

  1. 企業債發行與利差
  2. Oracle 債券利差變動

  3. 信用市場吸收能力

  4. 誰在資助承諾？

  5. 銀行放貸數據
  6. 資產拋售（國債、股票、加密）

  7. 外匯儲備變化

  8. 支出速度

  9. AI capex 公告 vs 實際部署

  10. FDI 承諾 vs 破土動工

  11. 勞動市場信號

  12. AI 造成的工作取代（消費領先指標）
  13. 消費者信心

  14. 信用卡/房貸申請

  15. 市場何時 discount 終結？

  16. 前瞻盈餘指引
  17. 訂單簿惡化
  18. 專案取消（利率敏感度）

底線：在打上一場戰爭

大多數投資者困在 2010s 劇本： - 痴迷 Fed 資產負債表流動 - 等待 QE 拯救一切 - 預期所有資產一起通膨

新劇本（如果 Andy 是對的）： - Fed 被邊緣化，除非危機 - 私人部門信用創造是遊戲規則 - 股債分化（不是「漲潮抬升所有船」） - 實體經濟再次重要 - 傳統商業週期回歸

風險： 這個結構轉變要求承諾被資助並且回報。如果任一失敗，這不只是修正——而是信用緊縮，Fed 彈藥有限。

Key Quotes

"So I look at all of the spending, the budget, the AI, the onshoring as all a set of promises that are largely in excess of what you would normally need to have a stable economy."

"The question goes to who's creating the credit? How are these promises going to get funded? The Fed or commercial banks?"

"When you fill a bank with reserves and they don't need them, it doesn't change their behavior... They're already able. They're just not willing."

"The best way to convince a bank to lend money is to pay up for it. They want to have money pulled from them. They want to be a price maker, not a price taker."

"Credit creation doesn't need money creation... If you have money and you want to lend it to me, you can lend it to me. We don't need a bank to do that."

"The savings are what fund the original loan. And so that circle always works unless the things that are being spent on don't pay off."

"We are facing too much demand [for money], and thus a crowding out effect."

"The total amount of assets that the banking system owns is around 20 trillion... You have at most $2 trillion of credit creation from the US banking system available and the promises are in excess of that."

"Oracle's taking that money and actually spending it in the real economy, creating growth... At the same time, tremendous supply of assets that someone's going to have to buy."

"Unlike most investment cycles where you're building a factory to put workers in it, this is you're building a factory to take workers out of it."

"The reason why we aren't a manufacturing economy is because we're too expensive to manufacture here... It's a waste of money because that insurance that you're spending is not going to pay off very often."

"You have to ask whether they'll honor their promises. Remember, I don't think it's a very good investment and they don't either."

"The last 20 years you really hadn't had to do anything at all except own assets... Everything inflated... The world I just described where there's lots of borrowing today for a hamburger today... that's bad for bonds."

"If we get 10% earnings growth this year, stocks are going to be up. Even if you get a two-handle multiple contraction, stocks are still up."

"This is not a new thing. These circularities—it's just how the economy is supposed to kind of work. People are still stuck fighting the last war."

Personal Reflection

Why This Matters

This conversation is exceptional for three reasons:

  1. Mechanistic clarity: Andy doesn't just describe what's happening—he explains how it works. The distinction between money creation and credit creation is fundamental but widely misunderstood. The "circular loop" explanation (spending → savings → funding) is elegant and testable.

  2. Regime diagnosis: If Andy is right, we're witnessing a structural shift as profound as 2008. The 2010s were abnormal (central bank dominance, asset inflation, suppressed business cycle). The 2020s represent a return to normalcy (private sector credit, real economy investment, traditional cycles)—but most investors are still fighting the last war.

  3. Second-order implications: The AI capex paradox is profound. Traditional capex cycles create jobs → consumer confidence → consumption → more investment. AI capex destroys jobs → ???. If this breaks the consumption link, the entire credit cycle fails. This is why Andy is cautious despite near-term growth.

Questions This Raises

  1. When does the market realize?
  2. Oracle spreads are out 80bps, data center stocks crushed
  3. But S&P keeps grinding higher on "10% earnings growth"

  4. At what point do equity markets price in the credit cycle risk?

  5. How much can rates rise before promises break?

  6. Andy mentions projects are already sensitive (Oracle issuance killed data center stocks)
  7. If 10-year yields hit 5%+, how many projects get canceled?

  8. What's the trigger for Fed intervention? (He says Fed easing "won't be enough")

  9. The AI employment question is existential:

  10. Every past technology disruption eventually created offsetting jobs
  11. What if AI is genuinely different? (Replaces cognitive work, not just physical)

  12. Does this force UBI or similar structural changes? (Andy mentions as "Path B")

  13. Onshoring credibility:

  14. Foreign companies already chose not to invest in US manufacturing
  15. Will they follow through on tariff-negotiated promises?
  16. What happens to rates if they don't? (Huge supply disappears → rates spike)

Cross-Reference with Plur_Daddy's Thesis

Remarkable convergence between Andy Constan (credit cycle mechanics) and plur_daddy (capital scarcity narrative):

Both argue: - ❌ Money is scarce, not abundant (opposite of 2010s) - 📉 Speculative assets (crypto, gold) get crushed - 📈 Near-term cash flow assets (DRAM/HBM, stocks with real earnings) outperform - ⚠️ Credit/liquidity shortage is the dominant force

Complementary insights: - Plur_daddy: "Deepest pockets tapped out" (Saudis, Softbank) → must sell to fund Altman - Andy: "FDI promises must sell treasuries to fund factories" - → Same mechanism, different promises

Key difference: - Plur_daddy: Focused on immediate market impact (why crypto feels bottomless, why DRAM rips) - Andy: Focused on 2-3 year structural question (will promises pay off? will consumption follow?)

What I'd Watch

Leading indicators of regime confirmation:

  1. Credit spreads by sector:
  2. AI-related corporate bonds (Oracle, hyperscalers)
  3. Manufacturing/industrial (onshoring beneficiaries)

  4. Consumer discretionary (consumption signal)

  5. Foreign treasury holdings:

  6. Saudi Arabia, Japan, South Korea central bank data
  7. Are they net sellers? How fast?

  8. If not selling → promises aren't being funded → onshoring fails

  9. AI employment data:

  10. White-collar job postings (cognitive work most at risk)
  11. Labor force participation rate

  12. Consumer confidence vs AI deployment correlation

  13. Bank lending surveys:

  14. Willingness to lend (has demand returned?)
  15. Credit standards (rising = regime confirmed)

  16. Loan growth vs money creation gap

  17. Equity sector rotation:

  18. Momentum/speculation struggling (regime confirmed)
  19. Value/cash flow outperforming (regime confirmed)
  20. If everything rallies together → regime thesis wrong

Falsification signals: - Fed forced to do large QE despite no crisis → Andy wrong, back to 2010s playbook - All assets inflate together (stocks+bonds+crypto+gold) → "fighting last war" crowd was right - Consumer credit explodes without job growth → something broke in the model

Investment Implications (If Thesis Correct)

Avoid: - ❌ Long-duration speculation (crypto, unprofitable tech, meme stocks) - ❌ Long-duration bonds (corporate or sovereign) - ❌ "Fed will save us" trades

Favor: - ✅ Equities with real earnings, near-term cash flows (Andy mentions 10% growth) - ✅ Short-duration credit (if spreads compensate for risk) - ✅ Positioning for credit cycle failure (when? 2-3 years out? Earlier if projects canceled)

Ultimate question: Does consumption follow investment? If AI doesn't create jobs, consumer credit doesn't expand, and the entire thesis collapses into credit crunch. This is the 2-3 year risk Andy keeps returning to.

Why this matters more than typical macro: We're not debating Fed policy at the margin. We're debating whether the fundamental structure of credit creation has shifted, and whether that shift is sustainable. If Andy is right, most portfolios are positioned for a world that no longer exists.