Hello,

Welcome to our flagship newsletter, MarketBeat Daily Ratings.

We'll deliver the latest "Buy" and "Sell" ratings from Wall Street's top-rated analysts directly to your inbox each morning.

But first, we need you to do two quick things:

1. Hit reply, and send a simple "Yes." Just one word. This tells Google (and other emails providers) that you actually want to get our newsletter.

2. After that, this link to confirm your subscription. That will tell us that you received our welcome email and that we should start sending your daily report. 

Confirm your subscription here.

After you have completed these two steps, we would like to gift you a free copy of one of our most popular investing reports: 7 Stocks to Buy and Hold Forever. You can download the report with this link.

Thank you again for subscribing. We look forward to being an important part of your investing journey

Matthew Paulson
Founder and CEO, MarketBeat.

P.S. If you didn’t intend to subscribe, no problem—you can unsubscribe with this link.

(ARReply-161)

NVIDIA Just Named AI's Next Bottleneck—And These 3 Stocks Sit Right In It

Written by Bridget Bennett. Article Posted: 9/20/2026.

AI's loudest voices spent the week telling everyone to slow down, and the market took them at their word. AI infrastructure names sold off during a month that is traditionally unkind to stocks.

Jason Bodner, co-founder of quantitative research firm MoneyFlows, thinks investors are reacting to the wrong headline. The one that matters came out of NVIDIA's (NASDAQ: NVDA) August earnings call, when CEO Jensen Huang guided to roughly 70% revenue growth next fiscal year and then told analysts that demand is running well above that level.

Silver's record run faded but this project never slowed (Ad)

Silver is up more than 50 percent over the past year, yet interest has faded since its January record.

The metal gave back roughly half that move and most traders moved on to other trades.

One explorer kept its drill rig running through the pullback, working the site while attention was elsewhere.

See what this silver explorer has been drilling for

That gap is the whole story. AI's constraint right now is not demand; it is the physical hardware that moves data around inside a data center. NVIDIA named the two pinch points: memory and silicon photonics. Bodner's three picks sit on top of both.

Slowing the Models Does Not Slow the Buildout

Anthropic CEO Dario Amodei argued in a Sept. 12 essay that AI companies should deliberately slow the pace at which they improve model capabilities. OpenAI's Sam Altman endorsed the idea hours later, and Elon Musk backed it too. Meanwhile, a researcher who had worked at both Anthropic and OpenAI resigned days earlier with a public warning about existential risk.

Bodner reads the corporate half of that debate very differently than the tape did.

Regulation, in his view, gives incumbents licensing regimes, capital minimums and a moat that smaller competitors cannot clear. It also makes AI more acceptable for health care and government procurement. His analogy is the airline industry, which lobbied for a federal regulator and ultimately got mass adoption rather than grounded planes.

The macro backdrop gave sellers cover. The Federal Reserve raised rates on Sept. 16 rather than cutting them, oil is sitting near $100 a barrel amid U.S.-Iran escalation, and September has a long history of erasing gains.

What is not deteriorating is profit growth. Second-quarter beat rates across the S&P 500 ran well above their five-year average, and earnings growth reached its strongest level in years even after stripping out a couple of outsized outliers. That is a sentiment problem, not an earnings problem.

Micron Owns the Memory Shortage NVIDIA Keeps Naming

Micron Technology (NASDAQ: MU) is the fastest-growing producer of high-bandwidth memory (HBM), the stacked chips that sit beside a graphics processor and feed it data. Bodner's framing: A GPU is the world's fastest chef, and HBM is what keeps the ingredients on the counter instead of scattered across town.

Management has said industry demand continues to significantly exceed supply and that the tightness could persist well beyond next year. Micron's HBM output has been selling out ahead of production, and analysts have responded with steadily higher forward estimates. Bodner's point is that a stock can be up several hundred percent and still look cheap if its forward numbers are moving faster than its price.

NVIDIA is effectively underwriting the shortage. Its supply commitments more than doubled in a single quarter, primarily because of memory procurement. Fabs do not go up overnight, capacity is spoken for, and pricing power follows.

Coherent Sits Where Copper Runs Out

Copper wiring is fine at the chip level, but it becomes problematic across a data center. Resistance, heat and distance all pile up, so the AI industry is shifting from copper to optical connections. That transition depends on silicon photonics, which Huang has called the bottleneck of the coming decade.

Coherent Corp. (NYSE: COHR) supplies the lasers, transceivers and optical materials behind that transition. NVIDIA is investing $2 billion in Coherent alongside a multibillion-dollar purchase commitment and future access to its capacity. Lumentum Holdings (NASDAQ: LITE) received an identical commitment the same day, suggesting that NVIDIA is buying insurance against strain in a supply chain it expects to tighten.

The stock has taken a hard pullback along with the rest of the AI complex. Bodner's read is that the group has already absorbed most of its selling while energy, health care and gold rallied.

Tower Semiconductor Makes the Wafers Photonics Runs On

Tower Semiconductor (NASDAQ: TSEM) is the pick-and-shovel layer. It is a specialty foundry that manufactures wafers based on designs from other companies, which means investors do not have to guess which transceiver vendor will win.

Its silicon photonics business has grown from a rounding error into an important company engine, and management raised its long-term targets on the back of that growth. Tower has locked in more than a billion dollars of 2027 photonics contracts, with customer prepayments already in hand. That is demand booked before the capacity exists.

What Could Break the Setup

The doomsday case is that regulation kills AI demand. Bodner does not buy it, and adoption data is on his side.

The real risk is capital intensity. Tower is committing billions of its own money to a Japan expansion, and heavy capital expenditures can outrun actual demand if photonics adoption slows. Memory has broken investors' hearts before. These three stocks trade on the assumption that tightness persists, and the same forces that lifted the group could unwind it together.

Watch supply commitments and customer prepayments rather than safety headlines. Those are the factors actually moving these three stocks.

Readers can get Jason Bodner's MoneyFlows institutional money flow research at its discounted annual rate and track where the big money is moving next.