Trading gold by following interest rate movements this year has been completely wrong.
I've held the gold I bought for almost a year. I didn't sell it when the price was over 1000, and I never expected it to bounce back yesterday.
01
On the evening of August 5, gold rose by 4.2% in a single day, marking its largest increase in 5 months. Just a week ago, everyone was saying the same thing: there was barely any upside left for gold.
From $5596 at the end of January this year to falling below $4000 at the end of June, the price has dropped by nearly 30%. The "king of safe-haven assets" has failed to perform its safe-haven function.
Retail investors are selling, ETFs are seeing redemptions, analysts are slashing their target prices, and SPDR, the world's largest gold ETF, has seen its holdings drop directly below 1000 tons, hitting the lowest level this year.
To be honest, if you think about it carefully, don't you have that feeling?
When conflicts break out, gold is supposed to rise. This is a rule written on the first page of textbooks, an absolute truth.
After the US-Iraq war broke out in February this year, the Strait of Hormuz was blocked, the route that carries 20% of the world's seaborne oil was cut off in a single day. According to common sense, gold should have skyrocketed.
Unfortunately, that didn't happen. It kept falling without looking back. Why?
Because this time, the conflict first impacted oil prices.
As soon as the Strait of Hormuz was closed, Brent crude oil surged directly from over $80 to $112 per barrel. US inflation jumped from over 3% to 4.2%, and the Federal Reserve was pushed to a corner.
At the June interest rate meeting, 9 out of 18 committee members publicly took a stand: there will be at least one interest rate hike within the year. The exact words of the new chairman Walsh were "Those who expect easing will be disappointed".
What does a rate hike mean? The opportunity cost of holding gold goes up.
Gold is a zero-yield asset. If you don't hold it, you can buy US Treasuries to earn interest; the higher the interest rate, the less valuable gold becomes. It's that simple. The safe-haven logic was intercepted by the inflation-rate hike chain.
I looked through the statistics, and the results are surprisingly accurate.
The JPMorgan Asset Management team conducted a regression analysis. From 1990 to 2021, the single variable of the US 10-year real interest rate could explain 85% of the fluctuations in gold prices.
Do you know what that means?
For more than 30 years, gold traders only needed to track one number. When the real interest rate rises, gold falls, and vice versa. If you ask any gold trader with more than 10 years of experience, they will tell you the same thing: watch the real interest rate.
After 2022, this model stopped working. It completely failed.
When the Russia-Ukraine conflict broke out, the West froze $300 billion of Russia's foreign exchange reserves. This incident had a greater impact on central banks around the world than any research report.
Central banks suddenly realized a problem: The US dollars held in other countries' systems may not be retrievable, so they started buying gold frantically.
In 2022, global central banks' net gold purchases reached 1136 tons. In 2023, the figure was 1051 tons. In 2024, it was 1045 tons. In 2025, purchases continued to exceed 1000 tons. For four consecutive years, annual purchases have exceeded 1000 tons, while the average annual figure in the previous ten years was only over 400 tons.
Central banks are buying gold out of distrust. The old map no longer works.
According to statistics from the same JPMorgan team, the explanatory power of real interest rates on gold prices plummeted from 85% to 16% after that. Previously, 85% of gold price fluctuations could be explained by interest rates, but now only 16% can. It's more than halved.
Then, something even more interesting happened.
In the first quarter of this year, central banks suddenly bought much less gold. Data from the World Gold Council shows that global central banks' net gold purchases in Q1 were only 57 tons, compared with 244 tons in the same period last year.
Note that 187 tons of this figure were reclassified as over-the-counter transactions, the statistical caliber changed, and the real demand is not that low. But the market does not care about the caliber, it only looks at the 57 tons figure on the report.
With fewer purchases, the old map suddenly started working again.
The 10-year real interest rate surged from 1.48% to over 2.3%, and everything seemed to be back on track.
Traders reopened the old model and started tracking interest rates again, as if the four years of frantic gold buying by central banks was just an episode.
On August 5, Bloomberg strategist Cameron Crise did one thing:
He input the day's data into this old model and ran the simulation. The conclusion was that gold prices "should have fallen slightly", but instead they rose by 4.2%.
02
Who exactly was buying that pushed the price up by 4.2%? Let's first look at what happened on August 5.
Three things. First, the US ADP employment data was released.
This is the private sector employment report released before non-farm payrolls, which can be regarded as a preview of the official data. Only 44,000 new jobs were added in July, while the market had expected 70,000, and 95,000 were added in the previous month. The figure was directly cut in half.
Second, Trump said that the United States and Iran are close to reaching a 60-day temporary agreement to reopen the Strait of Hormuz, and Brent crude oil fell to $78 per barrel that day.
Third, the US dollar index fell below 100.
The three things all point in the same direction: rate hike expectations are fading. According to CME data, the market's bets on a September rate hike dropped from 80% to 54%.
Okay, the direction predicted by the old map is not wrong. Weak employment leads to lower rate hike expectations, which in turn leads to lower real interest rate expectations, so gold should rise. This logical chain is correct, but the magnitude does not match.
I checked the data.
The US dollar fell by less than 0.2% that day, and the 10-year US Treasury yield barely moved. According to the traditional interest rate sensitivity model, such changes can only push gold prices up by a little over 1%, but the actual increase was 4.2%, three times more.
Who was buying that extra three times of volume?
Goldman Sachs released a flash comment that day, pointing to China as the "most important immediate trigger" for this round of rally.
I looked through the data. On Thursday, the gold holdings on the Shanghai Futures Exchange increased by 19,000 lots in a single day, an increase of 6%. What does that mean?
In the past three years, there have only been four single-day position increases larger than that day. Bloomberg data also shows that China's gold ETFs have seen net inflows for 14 consecutive trading days, the longest streak since March.
Analysts from Huaan Fund said that since the gold price fell near $4000, institutional investors' interest has picked up significantly, and the correction of A-shares in the same period is also pushing funds to the gold sector.
That's not all. Ryan McKay of TD Securities wrote in a report that the gold holdings of macro discretionary funds have more than doubled since June.
What about the CTA trend model? It is still in a net short state, and the short positions have not been fully closed. The unwinding of short positions itself is fuel for the rally.
There is another buyer that you might not expect.
Tether. Yes, the company that issues USDT. I checked that it has quietly accumulated 154 tons of gold in its Swiss vaults, and added another 6 tons in the first quarter of this year.
It is the world's fourth largest gold buyer, exceeding the gold holdings of the vast majority of national central banks.
Due to the high information density of various research reports and the need for cross-verification, this fact has hardly been discussed together, but it perfectly illustrates a signal:
The people buying gold have changed. Wall Street's old model did not take this new player that is quietly hoarding gold in Swiss vaults into account at all.
Okay, we found the buyers. These buyers gathered together for different reasons:
Chinese asset investors are looking at the cost-effectiveness of gold prices falling below $4000, CTA short covering is a technical behavior, and Tether buying gold is part of its stablecoin ecosystem. None of them placed orders because "the real interest rate fell by 78 basis points".
Why? The problem with the old map is not that the numbers are miscalculated, it can only measure one type of temperature.
For example:
You have a fever, you take your temperature with a thermometer, it's 38.5 degrees, you do have a fever. Just looking at this number, how do you know if it's caused by a cold or a viral infection?
The thermometer can only tell you one thing: you have a fever. As for why you have a fever, when it will stop, and whether it will recur? It can't tell you any of that.
The old map is exactly this thermometer. I read a paper published by the JPMorgan Asset Management team this year. They did an experiment:
They replaced the "real interest rate" in the traditional model with "term premium", which is the market's pricing of the long-term credit risk of US Treasuries, and the explanatory power immediately came back.
In other words:
The market no longer cares whether the interest rate is 3% or 4%. It cares whether the US government can still pay its debts on time. This is a trust issue.
Speaking of which, there is one thing worth mentioning separately: the Strait of Hormuz.
In the first half of the year, this strait was a hammer that smashed the gold price. When it was blocked, oil prices soared, inflation rose, rate hike expectations heated up, and gold prices plummeted.
Now, this strait has become a hand that lifts the gold price. When the agreement is reached, oil prices fall, inflation expectations cool down, rate hike expectations also fall, and gold prices skyrocket.
The same strait, the same force, hit you in the first half of the year and supports you now. The transmission mechanism has completely reversed.
03
The question is: what does the new thermometer look like? Wall Street itself is arguing about it.
How intense is the debate? I ran through the data from the research reports, and you will see how divided they are.
Wells Fargo: $6100; Goldman Sachs: $4900; UBS: $5200; Deutsche Bank: $4600; JPMorgan Chase: $4500; Barclays says the fair value is about $4150; Citi: $4000.
From $4000 to $6100, the span exceeds 50%. These are the new maps in everyone's hands, and they are all different versions.
There is a report worth mentioning separately. It was published by SDIC Securities in mid-July. The title is very blunt: The historical top of gold has been basically confirmed.
Why? AI.
Their logic is as follows: AI is improving the productivity of the United States, and higher productivity will strengthen the credit of the US dollar.
What is the core logic for gold's rise in the past few years? It is exactly the weakening of the US dollar's credit. Once this logic reverses, the underlying pricing pillar of gold will be broken and gone.
After reading it, I only have one thought: it's too hasty.
To be honest, the productivity dividend of AI is far from being reflected in macro data now. Data centers are being built, chips are being purchased, and optical cables are being laid. All these capital expenditures themselves are creating inflation.
As for enterprise process reengineering? It has only just begun. SDIC itself also wrote in the report that we need to look at hard indicators such as non-farm business labor productivity and unit labor cost.
But none of these indicators have given a conclusion yet. It's like you haven't even taken the temperature, but you are already issuing a death certificate.
Central bank gold buying is happening right now. According to the World Gold Council's Q2 report, global central banks' net gold purchases reached 289 tons, a year-on-year increase of 62%, the highest Q2 figure in history.
Poland bought 51 tons in one quarter, bringing its reserves to 632 tons, with a target of 700 tons. The exact words of the governor of the National Bank of Poland: "We are taking advantage of the recent price correction." China bought 33 tons.
45% of the surveyed central banks said they will continue to increase their gold holdings in the next year, the highest proportion in the history of this survey.
South Korea is another example. In 2013, the Bank of Korea bought 90 tons of gold at an average price of $1629 per ounce. Right after the purchase, the gold price plummeted, and the governor was summoned to the National Assembly for questioning. In the following 13 years, South Korea never touched gold again.
Now it's back. The scale is small, only 4 to 5 tons a year, which is a by-product of its domestic copper and zinc smelting. The exact words of the head of the Bank of Korea's reserve department: This decision is not based on price judgment, but made after comprehensive consideration of domestic and foreign market conditions.
It bought at $1629, and now the price is $4300. It missed the previous rally, but it chooses to come back now.
To put it bluntly, the plans of these central banks are completely different from those of traders. Traders look at daily charts, while central banks look at a 5 to 10 year horizon.
They are buying insurance for their foreign exchange reserves. Insurance seems useless in normal times, but you will know its value when an accident happens.
Speaking of which, there is also a dose of cold water:
Citi and UBS once said the same thing: central banks can only put a floor under the price, they cannot drive a sustained rally. A real long-term bull market requires large-scale return of ETF funds.
What UBS's Castelli means is: central banks cannot allow the proportion of illiquid assets in their foreign exchange reserves to exceed a certain limit. Gold bars are not US Treasuries after all, and cannot be liquidated at any time.
This sentence is correct. Putting a floor under the price is part of the new map.
Previously, when the price fell by 20%, no one would buy at $4000, and a stampede would happen. Now there are buyers. This is a structural change.
It's like the floor of your house, which used to be wooden, but now it's replaced with concrete. The concrete won't make the house taller, but the house will not collapse.
Is anyone still using the old map? Yes. The Federal Reserve itself is using it.
At the FOMC meeting on July 29, the vote was 9 to 3 to keep interest rates unchanged. Three committee members voted for a rate hike. On one hand, the ADP data shows that employment is cooling, on the other hand, three committee members say inflation has not come down, and the new chairman Walsh is caught in the middle.
Tomorrow evening, at 8:30