Last week, we said a daily close below $82.5K would mean the breakout had failed.
That did not happen. BTC fell to $82,563 on Monday, just $63 above the line. Since then, every daily close has been above $83.5K. September ended up 6.4%, which is rare for bitcoin. Q3 ended up about 42%.
But the bounce was weak. BTC tried to get back above $85K several times this week. Each time, sellers pushed it back down.
This tells us something. The news this week was good for BTC. Inflation came in soft. Fed speakers sounded less hawkish. The odds of an October hike fell by half. In a healthy market, news like this should push the price higher. BTC still could not hold $85K. On Wednesday, after the PCE data, it jumped to $85,650 in one hour. Two hours later, it was back at $83.4K.
When good news cannot lift the price, we take that as a warning.
We are still on alert. In the short term, we are careful. The line is still $82.5K. If BTC closes below it, the on-chain data shows the next real support is around $78K.
What Changed
Macro. Last week, we said long-term yields could keep going up even if the data got softer. The reason was that bond investors were holding too many bonds going into this hiking cycle. This week, that is what happened.
Most of the data was soft. Core PCE rose 0.2% in August. The forecast was 0.3%. Compared with a year ago, it was up 3.0%, below the 3.3% forecast. Part of this drop came from a change in how the data is calculated. That change also cut July's number from 3.3% to 3.0%.
Consumer confidence fell to 81.9, the lowest since 2014. Job openings were 7.1 million, below the forecast. New York Fed President Williams said there is no rush to raise rates again. On Polymarket, the odds of an October hike fell from 64.5% last week to about 34%.
Long-term yields did not care. The 10-year yield went from 5.17% last Friday to 5.30% on Wednesday. That is the highest since 2002. The 30-year yield reached 5.63%. The 2-year yield also went up, but much less. So the pressure is no longer coming from the Fed. It is coming from long-term bonds. Investors want to be paid more for holding them.
We do not think this will go away soon. The government is selling a lot of bonds. AI spending is also raising the return that money can earn. A slower Fed does not change either of these. Bonds have fallen a lot, but we do not see signs that the selling is over yet. The 10-year yield is now at 5.29–5.32%. If it goes above that, the next level is 5.47–5.53%.
People are spending a lot, but not from new income. After inflation, spending rose 0.6% in August, while income did not grow. The savings rate fell to 4.1%. People are using their savings to spend. Companies are still buying a lot of equipment from abroad, so tech investment has not slowed. The job market is getting weaker, but it is not breaking. Job openings fell, but private hiring was 90K, better than expected. For Friday's jobs report, the forecast is around 85K to 90K.
Oil also moved the market. Last weekend, the US said no to Iran's offer to reopen the Strait of Hormuz within seven days. Brent oil closed about 4% higher on Monday. There are also reports that Trump expects attacks on Iran to start again after the November elections. Crude oil exports from the region are almost back to pre-war levels. But fuel products like diesel are only back to about 58%. Diesel is the price to watch this winter.
AI. Last week, we said Micron's earnings were the real test. Micron passed easily.
Revenue was $54.2B. The market expected about $51.2B. Gross margin was 87%, up 2.1 points from last quarter. Adjusted earnings per share were $33.42, against about $31.6 expected. For next quarter, Micron expects revenue of $60–63B. The market expected about $56.6B. Gross margin should be about 86.25%, and management said this will be the lowest level of the whole year.
Two other points matter more than these numbers. First, most of Micron's 2027 HBM supply is already sold, at much higher prices. Second, Micron now has 26 long-term deals. They cover more than 35% of its revenue until 2030. Customers have put down $32B, and most of it is cash.
Before the report, some people worried that customers would switch to cheaper chips and buy less HBM. That does not fit what we see. Customers do not lock in supply a year early, pay cash up front and accept higher prices if they plan to buy less. Long-term deals this large also make Micron less of a boom-and-bust business. Management expects memory to get even tighter in the next two years.
Even so, the stock only rose about 1.6% after the report. Samsung and SK Hynix opened lower the next morning. Last week, we said the risk for memory is interest rates, not demand. Micron showed both sides of that this week.
The supply problem we wrote about two weeks ago is also getting worse. The most advanced chips from TSMC look short until 2028. TSMC's spending for 2027 and 2028 is now expected to be around $86B and $100B. The business is getting stronger. But with rates this high, strong and stable companies are doing better than risky, fast-moving stocks. We would not chase rebounds in the risky names.
Crypto. ETF buying has stopped. In the week before, US spot bitcoin ETFs took in about $2.4B. That was the best week of the year. This week, they took in $31M on Monday and $66M on Tuesday. On Wednesday, $149M went out. That ended nine days in a row of inflows.
Long-term holders sold about 61K BTC in the past week. New buyers took those coins and moved them off exchanges. That is why the price did not fall apart. But US buyers did not step in. All week, BTC on Coinbase was cheaper than on exchanges outside the US. Someone is taking the selling, but nobody is chasing the price.
