Bond Market About To Flip: A Major Shakeup Is Coming | Jim Welsh

VanEck Semiconductor ETF

Across the shows: 1 speaker in other episodes, past 30 days

• Jim Welsh · Macro Tides LLCBearish

  SMH completed an ABC rally and should fall below 503, possibly toward 460.

  SMH bottomed near 503, rallied to about 607, pulled back to about 540 and topped at 639. Welsh reads that as a completed ABC corrective rally, so SMH is heading lower. The RSI near 30 made a bounce due, and today's bounce is that bounce, but he still expects SMH to break below 503, possibly to about 460. He calls semis an over-owned sector that is unwinding even though earnings have been excellent, and money has rotated into it because its earnings look safe.

  “So when we got there, my take was we've just completed an ABC rally off of five zero three. We're going lower. I still believe we'll see SMH below the five zero three level.”at 7:13

• Jim Welsh · Macro Tides LLCBearish

  Semiconductors are highly cyclical, and the AI-driven chip ETF faces downside risk before a rally.

  Welsh calls semiconductors one of the most cyclical sectors, noting that the SMH semiconductor ETF dropped more than 65% in the 2001-02 downturn and during the financial crisis. He says the market is now built on AI and semiconductors, so the sector is vulnerable to a sharp unwind. He also says that if SMH falls below 503 it would set up a great rally, so he sees downside first, with a buying opportunity at that level.

  “The thing I point out in semiconductors, it's one of the most cyclical sectors there are. In two thousand one and two and in the financial crisis, SMH dropped by more than 65%.”at 43:26

NVIDIA

Across the shows: 17 speakers in other episodes, 14 bullish and 6 bearish calls, past 30 days

• Jim Welsh · Macro Tides LLCBearish

  Nvidia's 8% S&P weight means an AI unwind would hit the broader economy through wealth effects.

  Welsh points out that Nvidia is about 8% of the S&P 500 and that so much wealth is tied up in AI stocks. When this unwinds, he expects an impact on the economy through the wealth channel. He frames this as a risk that will play out at some point, not a timed call.

  “NVIDIA is 8% of the S and P. So my point earlier in terms of at some point in time when this unwinds, it's gonna have an impact on the economy because so much wealth is tied up in this stuff.”at 43:48

US stock market outlook

• Jim Welsh · Macro Tides LLC

  Mega-cap bounce can't carry the broad market; more downside than upside, with a secular bear coming.

  Breadth is weak: about 600 more S&P stocks were down than up, the equal-weight S&P was negative, and the advance-decline line is weak. The S&P has gone sideways since June because money rotates between the Mag Seven and semis and the broad market, with not enough money for everything. The Mag Seven bounce can lift the S&P to roughly 7,750–7,800, but a weak broad market means more downside than upside. Historically every major S&P decline since 1928–29 came with a negative breadth divergence, and a secular bear market is right around the corner. It needs a reason to sell, and the consumer-spending channel would make a sustained bear worse.

  “But as long as we see the broad market acting so weak, David, I think one has to... Well, my conclusion is there's more downside risk than upside potential.”at 8:06

• Jim Welsh · Macro Tides LLC

  The US stock market is vulnerable to a sizable pullback, with more downside than upside ahead.

  Welsh says the market is vulnerable to a decent-sized pullback. He sees more downside risk than upside potential, and he points to a day when more stocks fell than rose as evidence that breadth is weak. He says a sharp move in long-term yields could support the broad market later, but for now he expects a pullback.

  “So, you know, my take is we... We're vulnerable to a a decent sized pullback here.”at 44:38

US Treasuries & yields

• Jim Welsh · Macro Tides LLC

  Treasury yields are nearing a top; expects yields to fall after a possible final bounce to a higher high.

  Welsh says the 10-year and 30-year are nearing a high. The Economist cover 'bonds are going to crash' came after Treasuries had fallen 47% since March 2020, so the trend is late and contrarian indicators point to a turn. He expects the 10-year to take out its 5.36% high while the 30-year fails to exceed 5.73%. That intermarket divergence would be the most bullish setup for bonds. Last week both fell over 30 basis points, a dip toward 5.15 and then 5.05 happened, and the down-up-down pattern looks corrective. He notes the broad market has been responding to yields more than to anything else, including Fed policy. On the global selloff, he argues few investors have a reason to sell or buy, so modest flows can move prices a lot.

  “[I] am expecting a decline in treasury yields. Now it's just a question of do we get a bounce in the near term that takes the 10 to a higher high or not.”at 16:11

• Jim Welsh · Macro Tides LLC

  The 10- and 30-year Treasury yields should fall over the next two or three months.

  Welsh expects the 10-year and 30-year Treasury yields to decline over two or three months. He says that if he is right about rates, the broad stock market would respond positively. This would let investors get the best of both worlds, with stocks benefiting even as the AI and semiconductor sector stays under pressure. He gives no specific yield level.

  “If I'm right about rates that at some point in time, we're going to see the ten year and thirty year decline over two or three months, well, then you can get the best of both worlds where the broad market responds to that.”at 44:46

AI infrastructure & data centers

• Jim Welsh · Macro Tides LLC

  AI build-out spending growth will slow next year, though still up 25–30% versus 50–60% now.

  Companies are expected to spend $1.5–2 trillion on the build-out over the next two years, and AI spending is about 2% of GDP. Welsh expects the rate of change of that spending to slow next year. It will still grow, but not at 50–60%. That slowdown could weaken the economy because the economy and market are leveraged to the AI trade. It is one of several possible triggers for the market downturn he expects, alongside rising rates and an oil shock.

  “And at some point in time, the rates of change of that spending are gonna slow, and they're gonna slow next year. It's still gonna be up, I think, you know, 25, 30%. Not gonna be up 50 to 60%.”at 26:20

• Jim Welsh · Macro Tides LLC

  AI data center buildout may overshoot demand, repeating the fiber-optic bust with heavy losses.

  Welsh says that in past booms, companies kept investing before returns arrived, and eventually there were more data centers than needed. He compares this to fiber optic cable, where builders kept pushing, returns did not come right away, and some companies went out of business, with stocks falling 80 to 90%. He expects a similar shakeout in AI infrastructure at some point, since productivity gains that would justify the spending have not yet shown up.

  “At some point in time, there will... We will see where we have more data centers maybe than we need... just like with the fiber optic cable, they kept pushing and pushing and pushing, and they didn't get a return immediately.”at 43:06

US rates & the Fed

• Jim Welsh · Macro Tides LLC

  Fed won't hike in October but will hike in December, pacing moves rather than hiking every meeting.

  The Chicago Fed president's framework says demand-driven inflation requires aggressive hikes at each meeting, while supply-shock inflation argues for pacing because monetary policy works with 12–18 month lags and supply shocks fade. After five years above 2%, the Fed has to respond even to a supply shock, but methodically. That means no October hike, even though CPI and PPI came in hot and PCE is likely to be above expectations. Nine members wanted no hike even in June. The October meeting is five days before the election, which is another reason to wait, and the Fed will likely hike in December.

  “And so to me, I just have thought they're not gonna hike in October. ... But I do believe they will hike in December.”at 30:58

• Jim Welsh · Macro Tides LLC

  A productivity boost from AI could let the Fed cut rates to offset unemployment without inflation.

  Welsh argues the Fed normally hikes when growth runs above potential, which he puts at about 1.8% to 2%, because demand pushes prices up. A sustained productivity gain of about 1 percentage point would let GDP grow faster without inflation. That would give the Fed room to respond to rising unemployment by lowering rates, perhaps not aggressively. He says Morse is hinting at this, but he also says the productivity gain is not happening yet and will not arrive as quickly as people want.

  “Well, that means you can get one percent more out of GDP with without inflation being a problem. And that would allow them, I think, to respond to the rise in unemployment by lowering rates...”at 40:35

AI bubble & tech valuations

• Jim Welsh · Macro Tides LLC

  AI productivity gains look too narrow and slow to justify current AI spending within two to three years.

  Welsh says the average company is not yet seeing meaningful productivity gains from AI. AI-related profits are up about 300% since February 2023 against 15% for the rest of the market, and 76% of this year's earnings growth comes from AI stocks. Those earnings are real, which creates a self-fulfilling flow of money into the sector. But companies spending heavily on a new technology with no immediate results may raise or hold their spending, or pull back. Greenspan's 1990s technology productivity did not show up until 2003–04, so the payoff is real but delayed, and investors are starting to ask whether companies will earn a return on their investment.

  “I'm not sure the productivity gains are gonna be as broad and as widespread and as meaningful in the next two to three years to justify a lot of spending.”at 36:18

• Jim Welsh · Macro Tides LLC

  AI market is not yet a mania, since manias need legitimate fundamentals and the rally lacks them.

  Welsh argues that manias need something legitimate to attract money and rationalize overspending. He acknowledges that AI is a real technology, but he says the market has not reached the mania stage yet. This is because productivity gains that would justify the spending have not shown up in the data, so he does not see a bubble at this point, even though he does see risks ahead.

  “In order for manias to develop, there actually has to be something legitimate going on... But it... You know, are we there yet? I don't think so.”at 44:16

AI & jobs

• Jim Welsh · Macro Tides LLC

  AI will cause job dislocations before its productivity benefits arrive over the next decade.

  Welsh expects AI's productivity payoff to be helpful over the next decade. In the meantime, job dislocations will happen, and displaced workers will need time to be retrained. This timing mismatch between spending, job losses and productivity is what he sees as a hurdle for the stock market.

  “But in the meantime, we are gonna see job dislocations. It's gonna take time for those people to be retrained and so forth.”at 36:18

• Jim Welsh · Macro Tides LLC

  Companies will adopt AI slowly, unwilling to fully commit until they see where it lands.

  Welsh argues that companies want to lift the existing workforce's productivity without cutting many jobs, but they will not jump in with both feet while they do not know the outcome. Adoption costs are not insignificant, and with the economy growing around 2% and company revenues around 3%, these resistance points mean the productivity payoff will take time to work through.

  “I just think realistically, companies are not gonna jump in both feet when they don't know what... Where they're gonna land.”at 38:48

Market sentiment, positioning & volatility

• Jim Welsh · Macro Tides LLC

  Today's market lacks the portfolio-insurance selling that drove 1987; circuit breakers guard against a repeat.

  The 1987 crash was driven mainly by portfolio insurance. Investors sold S&P futures to hedge each 3–5% decline, which pushed futures about 10% below cash. Arbitrageurs bought futures and sold cash, and this created a loop of selling. Circuit breakers introduced afterward are designed to stop that kind of cascade. Welsh says the economy then was growing about 5%, and it is roughly the same now, but the 1987 mechanism is not present today.

  “So that is a huge difference, and you noted the circuit breakers that then evolved from the crash in '87 are also there in place to prevent that kind of thing.”at 23:40

US growth & recession risk

• Jim Welsh · Macro Tides LLC

  Economy still in decent shape: middle class spending, though the bottom 30% is squeezed.

  Nominal growth is about 5% or better, similar to 1987. Middle-class households are still traveling, spending and going to concerts, while the bottom 30% of wage earners are squeezed. The biggest market declines come in recessions. Event-driven drops, like the roughly 20% fall in 1998 from the LTCM blowup, tend to be shorter. Welsh therefore sees the economy as solid for now and the risk as a later bear market.

  “the key point to me is I still think the economy is in decent shape. The bottom 30% of wage earners are getting, really, really squeezed, but, you know, the middle class is still doing pretty good.”at 24:32

US consumer

• Jim Welsh · Macro Tides LLC

  US consumer spending is concentrated: the top 10% drive almost half, so a market slump would hit spending.

  The bottom 30% of wage earners are squeezed, but the middle class is still traveling, spending and going to concerts. The top 10% account for almost 50% of consumer spending. Just as the market is leveraged to the AI trade, the economy now leans on that top decile. Those households are spending because markets are at highs. If a bear market lasts more than a few months and the S&P falls 30–40%, they will cut back, which would deepen any slowdown.

  “and the top 10% represent almost 50% of consumer spending. So just like the market is leveraged towards the AI trade, the economy now is leveraged to that top 10%.”at 24:54

Oil & fuels

• Jim Welsh · Macro Tides LLC

  Rising oil prices could deliver a knockout punch to the market, a historical bear-market trigger.

  Welsh lists possible triggers for a downturn: rates moving higher, an oil shock, and a slowdown in AI spending. Oil prices keep rising. Historically, oil shocks have delivered a knockout punch to the market, which makes them a key risk to watch alongside rates.

  “Obviously, an oil shock. Oil prices keep going up. Historically, that has delivered kind of a knockout punch.”at 26:02

Enterprise AI adoption & AI software

• Jim Welsh · Macro Tides LLC

  Business AI adoption will lag spending; firms resist replacing staff until tools prove out, taking years.

  Welsh says companies resist replacing people until they are convinced the technology works and delivers what it promises. That learning curve will take a few years. Adoption will not accelerate as fast as investors expect given how much money is being spent. This gap between spending and adoption is the mismatch he sees in the AI trade.

  “That's a learning curve that's gonna take, I think, a few years. So the adoption rate, I don't think is gonna accelerate as fast as people would like it to accelerate based on how much money is being spent.”at 38:19