Goldman Sachs warnt: Raus aus Microsoft, Alphabet & Co.?

Finance-Focused Summary (Markets, Investing, Portfolio Construction)

Core Theme

Goldman Sachs and the video’s narrator argue that the AI buildout is becoming more expensive to finance, particularly through investment-grade (IG) corporate/project bonds. The pressure is driven by:

• Heavy debt supply from hyperscalers
• Rising credit risk and concentration concerns

The implication is that this could eventually pressure returns on capital for large tech companies.

---

Companies, Tickers, and Instruments Mentioned

Equities (Hyperscalers)

• Microsoft
• Amazon
• Meta Platforms
• Oracle
• Alphabet (implied)
• “Code” (appears to be a caption error, likely referring to Alphabet or a related big-cap)

Bonds / Projects / Credit

• Meta Platforms bonds (example maturities discussed below)
• Oracle — “Project Jupiter” (New Mexico project financing / loans)
• Emerson (appears related to a bond index/issuer concentration context)

ETFs

• iShares Dollar Treasury Bond ETF (1–3 years maturity) (captioned as “Isares Dollar Treasury Bond ETF”)

Risk-Free Benchmark Instruments

• US Treasuries (short-term yields referenced)

---

Key Numbers, Yields, and Timelines

Hyperscaler Debt Supply → Bond Market Impact

• New debt by hyperscalers
  • 2025: $108B
  • By Aug 2026: $194B
  • Full-year expectation (2026): $250B
  • Potential 2027: exceed $400B
• Project financing: “hundreds of billions” (not precisely quantified)
• Financing capacity context: referenced as about ~$2T so far (not presented as the main point)

AI Concentration / Risk Appetite Shift in IG Bonds

• Share of AI-related issuance in IG bond offerings:
  • 2024: ~1%
  • 2026: ~18%
• For maturities > 15 years: up to ~40% of new issues from the “AI complex”
• Insurer / long-term buyer behavior:
  • Q1: “queue” of ~15 insurers, each placing ≥ $50M orders for 30-year AI bonds
  • By end of Q2: fewer buyers—less than half as many wanted the deals

Corporate Bond Risk Premium / Repricing

• Average risk premium (vs government bonds) for AI corporate bonds:
  • about 74 bps → nearly doubled within a year
• Example project/corporate pricing:
  • “Debts… trading at 89–91 cents” (implies value ~89–91% of par)
  • If bought early: implied ~10% losses

---

Oracle: “Project Jupiter” (Delays, Permits, Force Majeure)

• Capacity: 2.45 GW
• Financing / loans: $18B
• Gas supply constraint: pipeline delayed due to permitting
• Delay timeline: at least 6 months, targeting Feb 1, 2027
• Air quality permit: still pending
• Force majeure: Oracle invoked force majeure
• Operational target: campus expected by 2028
  • Force majeure may allow deferral if operations don’t occur on time

---

Meta Platforms: Bond Examples (Yields, Duration, Price Risk)

Bond maturing Nov 2030

• Yield: ~4.2%
• Trading below face value
• Described as “current maturity is 5% per annum”

Ultralong bond maturing 2056

• Coupon: 6.3%
• Actual yield: ~7%
• Modified duration: ~10 years
• Duration sensitivity (illustrative):
  • If yields rise +1% (7% → 8%): bond could lose ~10% value
  • If yields rise +2%: could lose ~20% value

---

Treasuries & Alternative Returns (Duration-Aware Comparison)

• Short-term US Treasuries: “almost 5%”
• iShares Treasury ETF (1–3 years, ~2-year avg term):
  • Duration: ~1.8 years
  • Underlying yield: “almost 5%”

Currency-Hedged ETF Return Impact (EUR investors)

• Hedging cost driven by rate differential: about 1.4%
• Net return estimate for euro investors: ~3.6% (vs ~5% unhedged USD yield)

---

Portfolio Construction Recommendation (Conditional)

Defensive Allocation Suggestion

If keeping 20% in defensive cash/bonds:

• Invest 5–10% of the total portfolio in a short-term Treasury ETF (1–3 years), per the narrator.

Currency Hedging Caveat

• For euro investors, hedging reduces the estimated yield to ~3.6% rather than ~5%.

---

Frameworks Mentioned / Implied

Bond-Market Framework

• Track AI hyperscaler bond supply growth
• Monitor buyer risk appetite (e.g., insurers for long-dated AI bonds)
• Watch issuer concentration and market saturation risk (share concentrated in the AI complex)
• Expect higher yields to clear the market relative to US Treasuries

Equity Risk / Returns on Capital Check

Evaluate:

1. How quickly AI investments are scaled back
2. Free cash flow deterioration (many cited as already negative)
3. Incremental profit returns per additional $ invested (marginal profitability)

Checkpoint: if these factors “add up reasonably well,” it’s framed as “not a fundamental problem” yet.

Duration / Risk Control for Bonds

• Prefer shorter maturities to reduce duration risk
• Even “high credit quality” bonds can decline materially if rates rise
  • Example logic:
    • 2056 bond duration ~10 → higher volatility
    • 1–3 year Treasury ETF duration ~1.8 → more manageable volatility

---

Explicit Cautions / Recommendations (As Stated)

Caution on AI Corporate/Project Bonds

• Goldman Sachs is described as hardly buying AI bonds currently
• Rationale: waiting for yields/conditions to become more attractive
• Emphasis: concentration/saturation risk plus rising financing costs makes the bonds less compelling

Equities View

• Narrator says there is “no good reason to sell” shares of Amazon and Alphabet (“Code”)
• But “new purchases are not really on the cards”
• Exception: Meta, framed as having upside potential from a “post-smartphone” platform strategy

Major Risk Reminder

• Even strong companies can experience bond losses via duration if long-term yields rise.

---

Disclosures / Disclaimers

• The narrator states: “just my opinion, not investment advice.”
• Includes typical channel language (subscription/notifications), with the key disclaimer being not investment advice.

---

Presenter / Sources

• Presenter / narrator: Amin Prag
• Institution referenced: Goldman Sachs
• Companies referenced (incl. projects):
  • Oracle (Project Jupiter)
  • Meta Platforms
  • Amazon
  • Microsoft
  • Alphabet
• Additional mention: Blom Energy (referenced as a gas supply source for fuel cells)