Forget Memory Stocks, Nvidia Is the Cheapest Quality Chip Name in the US

Nvidia has been lagging the market and is now cheap as…

Here’s a sentence we didn’t expect to be writing in 2026: Nvidia (NASDAQ: NVDA) has quietly become one of the cheapest large-cap semiconductor stocks in the United States. Not cheap in the “it fell over” sense. Cheap in the sense that the price you pay today has drifted a long way below what the business is actually doing. And in a market where investors have spent the last twelve months chasing every memory chip with a pulse, that’s exactly the kind of mispricing we like to go hunting for.

Let’s set the scene. Over the past year the semiconductor trade has rotated hard. Money that used to sit comfortably in the AI accelerator names, like Nvidia, has sprinted into the memory manufacturers, and nowhere is that more obvious than Micron (NASDAQ: MU). Micron has had an absolutely massive run, with the share price re-rating on the back of tight DRAM and HBM (High Bandwidth Memory) supply, surging pricing and the AI data-centre build-out soaking up every module the industry can produce. On the numbers of the moment, Micron looks like it can do no wrong.

Despite the HBW boom, Micron is still a cyclical stock at heart

The trouble is that memory has always been the most brutally cyclical corner of the chip world, and nothing about 2026 has repealed that law even though HBM has arguably added another layer of growth underneath the big memory players. Micron doesn’t own its pricing; the DRAM and NAND cycle does. When capacity catches up with demand, and in memory it always eventually does, pricing rolls over, margins compress and earnings can halve in the space of a couple of quarters. You don’t have to take our word for it. Just look at where the brokers have Micron’s earnings heading through FY29. The consensus estimates don’t extrapolate today’s boom in a straight line; they build in the down-leg of the cycle, with earnings expected to soften materially as supply normalises. The consensus EBITDA forecast for Micron for FY29 is down 10% on the prior year! That’s the market quietly admitting what memory investors always forget at the top: this is a commodity business wearing an AI costume.

Nvidia is a very different animal. It sells the compute platform, the software moat and the ecosystem that the entire AI industry is being built on top of. Its earnings are far less hostage to a single spot price, and yet, and this is the bit that has us leaning in, it’s trading like the risky, late-cycle name in the group rather than the quality compounder it actually is.

The headline multiple already looks silly

Start with the crudest measure everyone knows. Nvidia is currently trading on a forward P/E of around 18x. Eighteen. For context, the Nasdaq-100 as a whole sits well above that (23x), and even the broad S&P 500 is trading on a higher multiple than Nvidia right now (20x). Read that again. The single most important company in the AI supply chain, the business whose chips every hyperscaler is fighting to buy, is priced at a discount to the average American large-cap. Twelve months ago that would have sounded absurd. Today it’s simply what the screen says.

So, how did we get here? A combination of things: the money rotating into memory, some understandable nervousness about the pace of AI capex, and the sheer size of Nvidia’s earnings base making the growth rate optically “slow” to headline-chasing investors. The market has decided that big means done. We think that’s a mistake.

Our preferred metric is flashing BUY

At Stocks Down Under and Pitt Street Research we don’t put too much faith in a single-year P/E, because it tells you nothing about growth, and P/E is a very flawed metric to begin with (but that’s a story for another time). Our preferred valuation measure is EV/EBITDA divided by EBITDA growth, a kind of enterprise-value PEG ratio. It answers the only question that matters: how much are you paying for each unit of growth you’re buying?

On that measure Nvidia is, frankly, in bargain-bin territory. Running EV/EBITDA against forecast EBITDA growth, we get roughly 0.2x in FY27, 0.29x in FY28 and 0.38x in FY29. Anything under 1.0x tells you the market is valuing the company below the growth it’s actually forecast to deliver. Nvidia isn’t a bit under 1.0x, it’s a fraction of it. In plain English: you’re paying about twenty cents on the dollar for FY27 growth. The valuation is dramatically disconnected from the earnings trajectory the brokers are penciling in.

A remake of the AfterPay movie

Where have we seen this movie before? AfterPay in 2020. Long-time readers will remember we were banging the drum on AfterPay when it was screening on almost identically depressed growth-adjusted multiples, right before the share price went vertical and the company was ultimately swallowed by Block (ASX: XYZ). We’re not saying Nvidia gets acquired; it’s far too big for that. We’re saying the set-up is very similar. When a genuine category leader trades on a growth-adjusted multiple this low, the risk-reward tends to be violently skewed in the buyer’s favour!

A great opportunity to get into Nvidia on the cheap

None of this is a knock on Micron, it’s a fine business and it’s had a fabulous run. But memory is a cycle, the brokers know it, and the estimates through FY29 make that plain. The heat that’s been powering the memory names looks to us like it’s starting to come out of the trade.

And that’s precisely why we keep coming back to Nvidia. On an 18x P/E that undercuts the index, and a growth-adjusted EV/EBITDA multiple that looks like a typo, we think Nvidia is probably the most attractive chip stock on the board right now, with the balance sheet, the moat and the long-term runway to back it up.