It was, um, too early to fade the chip rally.
I’ve conceded that particular point more times than I’m comfortable with lately.
It’s not that I’m bitter about “missing out,” it’s that the dizzier the heights, the worse the foreboding sense of vertigo.
On April 25, I called the semi surge “almost surely overcooked.” That was — cringe — 21% ago on the SOX.
Tuesday’s outsized rally, assuming it held, would push the gauge back into overbought territory for the first time since May 14.
The figure above’s a reminder: The SOX’s 14-day RSI spent every session from April 13 through May 14 above 70. That’s 25 consecutive trading days.
Micron was the center of attention again on Tuesday, when the shares rose 20%, if you can believe it. That’s almost unfathomable for a company that was already worth $850 billion.
Tuesday’s advance, the largest since November of 2011, tacked on $175 billion in market value, pushing Micron over the $1 trillion line.
If you ask UBS, there’s considerable room to run, and by “considerable” I mean the bank reckons the stock could nearly double over the next year on top of Tuesday’s surge.
UBS’s new price target, $1,625, would make the company worth almost $2 trillion by this time next year.
I won’t pretend to know how plausible (or not) that is, but if they’re right, it’d mean Micron’s rally from the “Liberation Day” lows would be 2,400%. “Super-cycle” indeed.
Meanwhile, Huawei says it doesn’t need ASML’s lithography machines to produce bleeding-edge chips. Without going too far into the specifics, a company executive said that with help from SMIC, China’s national champion managed to narrow what industry analysts generally described as a five-year tech gap vis-à-vis TSMC to just three using a strategy Huawei’s calling “sustainable evolution.”
Who knows if the claims are true, but if they are, it’s another testament to the idea that denying the Chinese access to technology doesn’t work and may ultimately be counterproductive. They’ll figure it out eventually, whatever “it” happens to be.
You could argue that export restrictions speed that process on several fronts. (Necessity’s the mother of invention, export controls are tantamount to claiming China isn’t capable and thus may motivate them to prove otherwise and so on.)
The Huawei executive who announced the breakthrough called it “a big leap ahead.” I would’ve found another way to say it. As stated, the innovation sounds uncomfortably close to “great leap forward.” The last one of those in China ended up with as many as 50 million people starving to death.
Grim jokes aside, the US doesn’t want Beijing to be semiconductor self-sufficient. That’d be very, very bad, and it’s the best argument for loosening restrictions on what American and European companies are allowed to sell into the Chinese market.
It’s better, some argue, to keep Beijing tethered to Western technology even if that means wittingly facilitating the PLA’s AI-enabled war-fighting capabilities.
Speaking of Western companies selling technology to Chinese firms, Bloomberg reported Tuesday that Qualcomm’s all set to supply TikTok parent ByteDance with “millions” of chips for AI data centers.
As the linked article noted, “so long as the Qualcomm chips fall within legally-acceptable thresholds, [the company’s] partners wouldn’t run afoul” of US export controls.
Qualcomm, which was down as much as 30% for 2026 as of early last month, rose more than 8% intraday to a record, before trimming gains.
The stock’s nearly doubled since April 7 and is now up some 40% for the year. Micron’s YTD advance is now 188%.
Constraining chip supply sure feels like wack-a-mole where the mole always wins.
At some point, the semiconductor market has to move toward an equilibrium. Maybe it’s Moore’s law helping chips catch up to insatiable AI compute demand, maybe it’s new fabs or expansion of existing fabs, or just more efficient models, but it’s hard to imagine these companies sustaining these massive windfalls. I’d be curious to see how much has capex gone up for semis. We know the hyperscalers are spending like drunken sailors, but I’d bet the semi manufacturers are doing the same. I need people smarter than myself to analyze how much manufacturing capacity is being added, how well the industry is tracking to Moore’s Law, and what the depreciation schedule is on the capex investments.
I definitely missed this boat. On to the next one..
MU has always been an extreme boom/bust cyclical. Once or twice or threeentimes shy.
Look at the chart for yourselves.
So much of the whole datacenter trade is predicated on endless expansion of fat inefficient LLMs. Cast an eye in the way more efficient Chinese LLM models. The US mega-scalers have too much money and pride to follow.
And don’t ignore the growing backlash across the polical spectrum (even Chip Roy in Texas!) against datacenters and AI in general. People focus on immediate benefits to themselves not long-horizon promises of drug discoveries or the need to keep up with China. Perhaps John & Jane Q Public rightfully realize that the financial foundation of AI success is allowing companies to fire people and pay the lucky survivors less.