The September Effect has a reputation for making investors nervous. September has historically been the weakest month of the year for U.S. stocks across a data set stretching back 98 years. For investors watching companies in the artificial intelligence (AI) space this year, September could be a useful test of whether a stock's investment thesis rests on price momentum or on business progress.
That distinction matters for Micron Technology (MU +0.22%) and Nvidia (NVDA +1.80%). Both companies are seeing conditions shift as the AI build-out progresses into a new stage, but the more interesting story is what is happening beneath the headlines about graphics processing units (GPUs) and central processing units (CPUs).
Image source: Getty Images.
Micron is becoming more than an AI memory supplier
Micron is one of the clearest examples of how AI infrastructure is changing.
The company produces HBM4, a type of high bandwidth memory (HBM) that Nvidia is embedding in its brand-new Vera Rubin platform, but that is only part of the opportunity. Micron is also shipping SOCAMM2 low-power memory and its PCIe Gen6 9650 data center solid-state drive in high volumes. Those products address all the different memory-related needs of an AI system: accelerator memory, CPU-attached memory, and storage.
All this jargon matters because AI factories are not just collections of GPUs. As large language models are asked to handle longer contexts and more inference workloads, moving and storing data becomes a larger part of the problem. Micron says its HBM4 can deliver more than 2.8 terabytes per second of bandwidth, while its SOCAMM2 can provide up to 2 terabytes of memory per CPU platform.
To me, that is a more useful bullish signal than another quarter of strong earnings. Micron's hardware now occupies key positions in the architectures that determine how AI systems are built.
NASDAQ: MU
Key Data Points
Nvidia is building the factory, not just the chip
Nvidia's own latest product moves the point in the same direction. The Vera Rubin platform combines GPUs, CPUs, networking, storage, and software into an AI factory. The company says Vera Rubin can deliver 10 times the agent throughput of its previous Grace Blackwell platform at scale.
The company's shift toward hardware designed to support agentic AI is important. AI agents make repeated calls to models, databases, and software tools, so their operation demands far more than just GPUs. They also need networking, storage, CPUs, software, and copious amounts of memory to keep everything running. Nvidia's Vera CPUs and Vera Rubin platforms show how the company is building an architecture designed to scale across different AI workloads and environments.
NASDAQ: NVDA
Key Data Points
What should investors do in September?
The 98-year September pattern is worth knowing about, but I would not use it as a reason to sell shares of either company. Septembers come and go, but these tickers are here to stay. A calendar effect can warn investors that volatility may rise. It cannot tell them whether a company's competitive position is improving or weakening. For Micron and Nvidia, the better question is whether the AI infrastructure cycle is producing new products, customer commitments, and deployments. I think it is.
Both companies are already familiar names, which could help them continue to attract attention from those looking to invest in the AI boom. And with both stocks already included in major indexes and offering solid exposure to AI, they're well positioned to remain on investors' radar.
There are risks. Nvidia faces the challenge of sustaining demand as AI customers spend at a huge scale, while competition from other makers of GPUs and custom silicon remains a threat. Micron faces the added risk of memory cycles. Production-related decisions can alter industry economics even when long-term demand is strong. Both of these stocks could also fall for reasons unrelated to their operating businesses.
That is why September may be useful as a review period. If these stocks fall, investors can ask whether the decline reflects a weaker AI thesis or a market pattern.
