Market Lab Report - More on long bond yields vs stock market; Amodei's AI warning

Long bond yields vs stock market

GDP and earnings beats push **EPS** up. Rising long-bond yields push **P/E** down, because a 5% Treasury is now the “risk-free” hurdle. Future cash has to be discounted harder.

So the market only goes up if **earnings grow faster than the multiple shrinks**.

If yields rise because **the economy is strong** (more demand for capital, higher real growth, companies printing money):

- Bond prices fall (yields up).
- Stock *prices* often still rise, because EPS is rising.
- Capital Group looked at many periods since 1962 where the 10-year jumped at least 75 bp over six months. Stocks often wobbled, then worked through it. They tended to *break* when the yield rise got very large — on the order of **2–2.5 percentage points**, not 20 basis points in a week.

If yields rise because of **oil, a hot CPI print, and a Fed that may hike**:

- That is a **discount-rate shock** without a matching earnings gift that week.
- High-P/E, long-duration names get sold first.

Both things can be true in the same month: the *year* is a growth/earnings year; the *week* is an inflation/Fed week.

Stock market goes higher if:

- 2026/27 earnings stay near those mid-20% then mid-teens paths,
- the 10-year settles rather than marching toward 5.5–6%+,
- oil does not restart an inflation scare,
- the “why” of yields stays “hot GDP / AI capex,” not “Fed lost the inflation plot.”

It stalls or falls if:

- yields rise another big step and P/E keeps compressing,
- 2027 earnings growth disappoints (estimates already slow next year),
- an oil/war/PPI loop forces hikes that hit demand,
- the index is still priced as if peak AI margins last forever.

## How to hold this in your head

1. GDP and earnings are the **fuel**.
2. The long bond is the **speed limit**.
3. This year the fuel has been real. The speed limit just tightened.
4. . A 5% 10-year with 25% earnings growth can still lift stocks. A 5.5% 10-year with fading 2027 growth can stop them even if GDP looks “fine.”
5. Watch **earnings revisions** and the **real 10-year (TIPS)** — those two tell you which horse is winning.

Anthropic's warning

Anthropic told the industry to slow down. They said AI could destroy humanity. A staffer quit after a few weeks and shouted the same warning. The press got the letter before the tweets went up. That looks like a stunt, not a revelation.

They have said the last six or seven models were too dangerous. Those models did not crash the banks. They made people more productive. The “10% chance we all die” figure is made up.

Watch what they do, not what they say. Nobody at these labs is freezing hiring or cutting spend. They want rivals to slow. They do not plan to slow themselves. Anthropic is still racing toward a huge IPO.

A U.S. slowdown would help China. Competition is safer than a federal board packed by the big labs. Markets punish bad products faster than regulators do.

Own your own intelligence. Do not outsource it to the lab giving the loudest speech this week. Keep accelerating. That is how you stay in the game.

Furthermore, AI now attacks and defends.

Most new bugs get exploited the same day they go public. Patch Tuesday is dead.

New models can write working exploits and also write the fixes. For 12–24 months, the best tools sit more with defenders than attackers. That lead will fade.

Humans cannot keep up. Use AI to find holes, patch them, and watch your own agents. Old C/C++ bugs can be rewritten away. Scams and stolen logins will remain.

This is why Anthony Pompliano and other accelerationists say it is dangerous to slow down AI. Extra evaluators are fine; a unilateral U.S. freeze is not. If U.S. labs pause and China or attackers do not, defense loses the only edge it has. Act now or fall behind.

On Fed hikes

**More stock market downside if:**

- 10-year holds **above 5%** and DXY stays over 100
- Oil stays elevated and core inflation does not roll over
- TGA refill after quarter-end drains reserves

**Higher if:**
- Next inflation prints cool enough that October is taken off the table
- Dollar fades and global funding conditions ease again
- Earnings season confirms AI spend is still self-funding (cash flow, not just debt)

### Practical read
From **Fed + global liquidity alone**, the base case is **not “stocks head higher from here.”** It is:

1. **Range / grind with a downward tilt** while the market maps how many hikes Warsh actually delivers.  
2. **Leadership narrow**: quality, cash-flow AI, and anything that doesn’t need cheap credit. Rate-sensitive small caps, transports, and high-duration growth get hit first (already visible).  
3. **Better risk/reward later** if long yields break down or the Fed is forced to pause after one hike. That is not the message from yesterday.

So: **don’t treat this as a liquidity-fueled buy-the-dip regime.** Treat it as a market that can bounce on positioning but needs *easier financial conditions* (weaker dollar, lower real yields, or faster global liquidity growth) before a sustained move higher is the high-probability path. Until then, the Fed just made the cost of capital the story again.