Title:   The AI Bubble, Real Yields, and Why Gold Sold Off
Show:    David Woo Unbound (David Woo, with John)
Guest:   David Woo (founder, David Woo Unbound; ex-BofA Head of Global Rates, FX & EM Strategy)
Date:    2026-JUN-04
URL:     https://www.youtube.com/watch?v=CvPC1VVZea8
Length:  ~11 min
Note:    Auto-transcript, timestamps mm:ss. Saved for personal study. A macro-strategy monologue: the 3-month
         bond sell-off is driven by real yields (oil + the AI rally), not inflation; oil is heading higher,
         but the AI trade is very crowded — so a curve steepener and long-dated gold calls make sense again.

================================================================

(00:00) Bonds and gold have sold off strongly since the year on war. But is it oil or token maxing? Why is steepener starting to make more sense again? What would it take for gold to shine again? >> [music] >> Before we go further, this is exactly the kind of question that we go deeper on inside David Woo Unbound. It is one thing to say that real yields are rising, oil is moving higher, and the AI trade may be getting crowded.
(00:40) The harder question is how to apply that view to a portfolio. This is what John and I focus on inside premium. Members get the full macro framework, John Speed the Market strategy, daily trade briefs, and the macro stock pickers, not to mention the Unbound community. You can start with a 30-day free trial. Premium is $100 per year after the trial.
(01:05) The link is in the description. Now back to the bond sell-off. Bonds and their proxies have so far been the biggest casualties of the US Iran conflict. Since the war began at the end of February, bond yields have risen, while gold and the Japanese yen have fallen. The rising yields has been most pronounced at the front end of the yield curve, but long bonds have not been spared either.
(01:29) The 30-year Treasury yield recently broke above the psychologically important 5% level. What is striking about the bond sell-off over the past 3 months is that it's been driven mainly by higher real yields, rather than higher inflation expectations. Of the 70 basis point increase in the 5-year Treasury yield since February 27th, roughly 60 basis points came from higher real yields, while inflation break-evens accounted for only 10 basis points.
(01:57) In the case of the 30-year Treasury yield, almost entire increase since February 27th has been driven by higher real yields. These developments raise important questions about the nature of the bond sell-off. If the rise in oil prices were the primary driver, inflation break-evens should have increased much more sharply relative to real yields.
(02:19) In theory, an exogenous oil price shock should push inflation expectations higher while simultaneously weakening growth expectations. So, what's going on? The release of Claude models at the start of April reignited the AI trade. The Nasdaq 100 has risen by more than 20% since then. I discussed the potential political and regulatory backlash to Claude models in a video 2 weeks ago.
(02:53) But, his breakthrough capabilities have only strengthened the view that AI will unleash a surge in productivity growth that will raise the economy's potential growth rate. This may be one reason why long-term real yields have repriced higher. Another factor is the positive wealth effect associated with the AI rally, which has convinced many in the bond market that the Fed can focus more on the upside inflation risk posed by higher oil price than the downside risk to growth.
(03:25) Of course, there's also the expectation of increased bond issuance to finance the AI-related capital expenditures. I suspect that neither higher oil price nor the AI rally alone would have been sufficient to drive real yields up as much as we've seen. It is the combination of the two that's created the conditions for the massive repricing in real rates.
(03:50) What this means is that for real yields to rise further, it will likely require both higher oil prices and a continued rally in the AI trade. Further sell-off of bonds needs two legs to stand on. What are the chances of that happening? Let's consider the outlook for oil before turning to the AI trade. US oil inventory data show another decline in crude stocks this week.
(04:19) Gasoline inventories picked up slightly, but remain at extremely low levels for this time of the year. A senior ExxonMobil executive said last week that if inventories remain at deeply depressed levels, oil prices could climb to 150 to 160 dollars per barrel over the next few weeks. Meanwhile, the Strait of Hormuz remains closed.
(04:43) Yesterday, June 3rd, not a single tanker passed through the strait. The uptick last week was largely due to some tankers shifting to a route closer to Oman that had been cleared of mines by the US during Operation Freedom. But after Iran fired on several of these vessels, it reportedly deployed additional mines.
(05:07)  Traffic through that corridor appears to have dried up once again. Interestingly, the market appears to remain hopeful that a deal will be soon reached. The betting odds of a deal by the end of June is still about 60% despite Iran having suspended indirect talks with the US. The market may believe the ceasefire reached between Lebanon and Israel will help revive negotiations.
(05:33) I think otherwise. I believe Iran's decision to suspend talks reflects growing doubts about whether Trump is negotiating in good faith. I do not have any inside information, but my gut tells me that Trump may have been dragging out these non-negotiations in part to give the US more time to collect intelligence on Iran.
(05:55)  After all, intelligence gathering is often most effective during periods of uncertainty when communications increase, personnel move around more frequently, and decision-makers are forced to reveal information about their intentions and capabilities. It is possible that the IGRC has concluded that continuing negotiations under these conditions carries more risk than benefits.
(06:22) It takes two to tango. If I'm right that Iran is walking away from the negotiation, Trump will have a tougher time to convince the oil market that a deal is just around the corner. This is why I think oil price is heading higher. I think either the continuation of the status quo or the end of the ceasefire should push oil price higher from here.
(06:52) The AI trade has always been driven by AI-related capital expenditures by the hyperscalers. Exactly what they're buying more of at any given moment may change, but the underlying driver remains the same. To me, the most important takeaway from the first quarter earning season is the fact that the combined CapEx spending of Microsoft, Amazon, Alphabet, Meta, and Oracle fell compared to the fourth quarter.
(07:20) On a year-on-year comparison, the growth rate also eased in Q1. Given the increased price of memory chips, we can safely assume the growth rate of AI-related CapEx in volume terms probably fell even more. In this respect, we should view the current shortage of memory chips as coming at the expense of the positive productivity impact of AI on the economy.
(07:45) Of course, the wealth effect is probably AI's most important transmission channel to the broader economy right now. It is probably why consumers are still spending despite the oil shock and personal saving rate at just 2 and 1/2%. Will retail investors continue to pour money into chasing the AI trade and driving it even higher from here? I don't pretend to know the answer, but this is what I know.
(08:12) The political and regulatory backlash to AI is growing. Even Trump has finally signed an executive order aimed at imposing greater oversight on the technology. An article in Political earlier this week suggested that Defense Secretary Pete Buttigieg has emerged as one of the strongest advocates within the administration for tighter controls on frontier AI.
(08:36) His reasoning is interesting. The wider the dissemination of frontier models, the easier it becomes for China to gain access to them and potentially use them against the United States. As I asked in an earlier video, if only a small number of users can access Claude 3, how would Anthropic monetize the technology? I remain of the view that the accessible capabilities of frontier models are plateauing.
(09:06) I see this as being the biggest problem facing Anthropic and ChatGPT ahead of their IPO later this year. For Anthropic, there's another problem. That is competition is closing in. Cursor, GitHub Copilot, OpenAI Codex are all quickly catching up to Claude Code. This is one reason why Microsoft reportedly has canceled most of its Claude Code licenses.
(09:32) This could mean that the massive growth rate that Anthropic saw in Q1 and Q2 may not be sustainable. The massive growth also reflected token maxing. Uber CEO said last week that the company ran through its token budget for the year in just 4 months and is now imposing a stricter limit on token usage. This suggests that enterprises may be transitioning from experimental usage to budget constrained usage.
(10:02) This could show up in the earnings and guidance for Q2. I see few positive fundamental catalysts for the AI trade in the near term. And technically, Intel, AMD, and Broadcom all seem to have lost some momentum. What this means is that the AI rally might become even more narrow. I think the balance of risk is starting to favor the downside as the AI trade is getting very crowded.
(10:34) Bond yields have been driven up by the Iran war and token maxing. While I think oil prices still headed higher, the AI trade looks very crowded. What this means is that the outlook for bonds is not quite as clear-cut as in the last 2 months. While I think it's still too early to start buying bonds outright, a steepener is beginning to make sense again.
(10:59) I also like owning long-dated gold calls. The yellow metal could regain traction if the air bubble were to pop. After all, if the AI trade begins to unwind, the wealth effect that has helped keep real yields elevated could quickly reverse. In that scenario, investors will likely rotate back towards traditional safe havens.
(11:22) Thank you for listening. >> Mhm.