Alphabet (GOOG 0.15%) (GOOGL 0.57%) just dropped a bombshell that is sending waves through the stock market, especially in the memory chip sector.
Before Alphabet’s announcement, the assumption was that memory capacity for AI computing chips would be in a supply constraint for several years, leading to soaring memory prices.
However, Alphabet’s latest breakthrough may have changed that notion, which has triggered a sell-off in Micron‘s (MU 0.49%) stock, one of the leading memory unit providers. The damage isn’t just isolated to Micron; it has also affected Sandisk and SK Hynix, other leading memory providers.
But is this reaction warranted? Or is there a real issue here?
Image source: Getty Images.
TurboQuant is reducing the amount of memory needed for generative AI
Generative AI requires a ton of memory capacity. As models are trained on more and more information, accessing some of the training data or previous conversations requires a ton of memory capacity. Google refers to this as a “needle in a haystack,” which is a perfect description of the challenge these models face. However, Google may have solved part of this issue with its TurboQuant algorithm. This algorithm reduces the amount of memory needed by generative AI models by six times, which logically triggered a collapse in memory stocks.
Since the paper was published on March 24, Micron’s stock is down nearly 20%, matching the sell-off in its peer, Sandisk. The assumption is that the customers are going to need six times less memory than before, which would clear the memory bottleneck that AI hyperscalers are currently facing. Additionally, Google made this algorithm open-sourced, so everyone, including its competitors, has access to it. With every AI company needing potentially less memory and the ability to integrate this breakthrough overnight, there is a lot of fear in memory chip stocks.
Today’s Change
(-0.49%) $-1.82
Current Price
$366.03
Key Data Points
Market Cap
$413B
Day’s Range
$340.50 – $366.90
52wk Range
$61.54 – $471.34
Volume
2M
Avg Vol
41M
Gross Margin
58.54%
Dividend Yield
0.14%
However, I think that may be a bit too broad an assumption.
The Jevons Paradox may be at work here
One item that isn’t discussed is how constrained AI models were before this breakthrough. Memory was a limiting factor, and with companies like Micron only being able to fulfill half to two-thirds of customer orders, there was no end in sight. Additionally, Micron projected the high-bandwidth memory market, the type primarily used for AI, would expand from $35 billion to $100 billion from 2025 to 2028 due to increased memory demand.
Just because AI companies can be more efficient with memory doesn’t mean prices will automatically fall. The Jevons Paradox notes that when something becomes cheaper or more efficient, demand can actually rise due to it becoming more accessible. That could happen with memory chips, as demand may rise because AI is becoming better at utilizing them.
This would keep the status quo going for Micron, making the sell-off a potential buying opportunity. The market has been fairly bearish on Micron’s stock since its latest earnings reports, and shares are down 30% since then, despite guiding for its revenue to rise to $33.5 billion, up from $23.9 billion in its latest quarter, and $13.6 billion the quarter before that.
MU Revenue (Quarterly) data by YCharts
I think investors don’t need to rush and buy Micron’s stock right now, as the TurboQuant news is still being digested. Investors should keep an eye on consumer memory prices (by looking at RAM stick prices) and see if they fall. If they do, then Micron’s stock could be facing falling product prices, which would give investors a reason to steer clear. If prices stay elevated, then that’s an indication that there is still a large memory supply constraint, and the Jevons Paradox is real and active for memory demand.
Personally, I think the TurboQuant breakthrough is huge news. Still, AI companies are going to find a way to utilize all of the freed up memory capacity differently, leading to still-elevated memory chip demand, making Micron a smart buy now.
