5 AI Stocks Just Crashed — I’d Buy Only 1

Nvidia reports tomorrow and the AI trade is already cracking beneath the surface. Western Digital, well, that's in fact down around 44% below its high even after gaining more than 150% this year. SanDisk, that's fallen by roughly 1/3 despite a year-to-ate gain of more than 500%. And Marvel, that's down close to 30% from its high even after securing a potentially enormous partnership with Google. Micron, that's around 25% below its peak, while Wall Street expects its earnings to more than double again. And then we have Broadcom, arguably the highest quality business of the five, is almost 27% below its high. And then just take a look at Nvidia. That's fallen for seven consecutive sessions, the longest losing streak since 2022, immediately before the most important earnings report of this quarter. So, what I've done in today's episode, I've valued all five of these companies from scratch. Four fail my test at today's price. Only one receives a buy. But before we analyze them, listen to why Nvidia's report now matters far more beyond Nvidia itself. >> The fundamental reality of markets today is that you're seeing trillions of dollars investment in infrastructure for AI, data centers, cooling, etc., all around one central theme. At the moment, if you look at forward expectations for earnings growth in the S&P, over 40% of that growth is expected to be delivered by these AI companies. So the the important thing to to recognize is to some degree that's already in the price, right? That's already baked into expectations, in which case it's reflected in market realities, which means if it doesn't happen, we have a problem. >> And that's the pressure hanging over tomorrow's report. Expectations are already embedded. Nine of the 10 best performing S&P 500 stocks in just the last year, well, they sell AI hardware. And we can see that the spending boom itself is real. Chips, memory, networking storage electricity and cooling. They're all being pulled into the exact same buildout. But we're noticing the financing is becoming more expensive. The 30-year Treasury yield that's above 5% near its highest level going all the way back to 2007. And that's ultimately the bearish concern, but the broader markets not behaving as if the psych was finished. Let's take a listen to the strongest counterargument before we judge these five stocks today. >> Your chart, you're not seeing anything overly concerning, even if there was a June peak for the S&P 500. >> Mhm. Well, that's correct. And again, you know, we the June we did have a June peak on June 2nd, but we went above that. No year in history has ever peaked in June. So, history doesn't repeat itself. Often rhymes Mark Twain. uh you know tech we'll get a little more technical for a second you know just two days ago 74% of all the components in the S&P 500 were above their 200 day moving average you go back in history when that's trending higher that's usually a good sign that's the bullish counterargument participation remains broad so this is not a simple market crash argument economic activity and market breadth they both remain constructive the real question that's narrower which AI suppliers can justify their expectations if Nvidia's guidance is merely excellent rather than perfect. And before we dive into the five, just to let you know, yesterday, as we do every single week, we released our latest weekly article. This one in particular, we looked at 29 stocks near 52- week lows. And out of those four, we selected as strong buys today. You can click on the pin comment below, sign up, and read all of these straight away. Now, we're going to kick things off with WDC, Western Digital Corporation. It's the largest draw down. It closed yesterday's market $435. Although in the pre-market we can see it up around 3%, $447. And you may look at this as a bargain. It's just had a 44% decline. But a stock becomes cheaper only when the price falls faster than its intrinsic value. And the operating recovery here is strong. Consensus expects fiscal 27 to essentially have earnings around $20, rising to just shy of 32 by 2028. And this takes the forward price to earnings ratio from roughly 22 based on 27 numbers to all the way below 10 based on 2029 and 2030. That is obviously if the later forecasts are achieved where growth for the company looks very strong A+ rating from Seek Alpha. Revenue expected on a forward basis to grow more than 40%. We can see forward earnings here. They're projected to grow around 86% in just the next 12 months. and long-term expectations for EPS very bullish 70%. However, what I would say is that several trailing profitability figures here they're distorted by the sand separation and related accounting. So, I'm not going to annualize a 73% net income margin. And looking at valuation as a whole, the picture here is a bit mixed. Earnings ratio, you can argue they do look reasonable. We can see in fact 22 times. That's lower than the 5-year average by around 38% and not too dissimilar from the sector median. But if you look at other measures here, whether in fact it's sales, book value, cash flow ratios, it does remain expensive to the relative sector and in fact expensive as we can note triple digits more than their own history. And my cash flow model for the company well it begins on 2026 based on analyst expectations $3.5 billion. Then it assumes as we can see for the growth rate around 10% 8% discount rate 3% here in fact as a terminal growth rate and it gives an intrinsic price $335. Although you'll notice that is below today's value we're talking around 25%. And even the more aggressive rate here at 15% growth. Well, we can see $472 is not that much higher. We're only talking about 6% in the market price today. And Wall Street, they do actually believe the complete opposite. They see a lot of upside. Their average price target sits at $665, implying 53% upside. Although, I would say the range here very wide. $420 on the lower end, over $1,000 on the upper end. I'd actually say the spread's a bit of a warning, not reassurance. So with Western Digital, a company where we can see on the reverse DCF currently baked in expectations at today's price sits at 14.2% and using the middle rate today, we get no margin of safety, a 33% premium. So my conclusion to wrap this up for WDC, I'd reconsider this below $350 with a genuine margin of safety close to 300. The largest crash therefore in today's episode is our first failure. Next up, we have Marvel Technology, ticker MRVL. The share price is risen around 170% year to date. Yet, it remains close to 30% below its all-time high at $330. We do notice, interestingly, a strong buy from Wall Street, although a very weak buy from Seek Alpha, a hold from Quan, and this summary of the Google partnership from Reuters. Well, overall, it's extraordinary. Marvel will help develop Google's custom AI chips, and Google can potentially become one of Marvel's largest shareholders. Now, the actual filing here matters more than the overall headline. Google received warrants up to 58.97 million Marvel shares at an excis price of $26.58. And most warrants vest in $240 tranches with one tranch earned for every $500 million of qualifying custom product revenue. Multiply those terms and full performance vesting implies as much as $120 billion of qualifying purchase through 2033. But full vesting would also increase today's share count by roughly 6 12% after issuance. The deal validates demand but existing owners they pay with dilution. And the underlining business while it's already accelerating data center revenues climb from roughly $400 million per quarter to more than 1.5 billion. And consensus well they expect total revenues rise from around $8.7 billion to more than 23 billion by fiscal 29. But the problem here is the price. Marvel trades around 56 times Ford earnings and roughly 37 times the following year's estimates. And you'll also notice the forward multiple is also far above the 5-year average. So investors are already paying for a large portion of the partnership before the actual revenue arrives. And if you want to take a look at this from Simply Safe Dividends, the blue tunnel which highlights intrinsic fair price. Well, we can actually see the stock price today sits above the upper end of the fair value indicating a severely overvaluation signal. Look at the last 5 years, 10 years. Marvel well actually has historically traded at a premium for quite some time, but never this large in terms of the disconnect from the stock price and the fair value. Now, my base model assumes again following analyst expectations, $3.5 billion of 27 free cash flow, then 15% annual growth. And it values Marvel based on these assumptions at $217. The reverse DCF that requires in fact around 16.2% 2% annual cash flow growth for a decade and Wall Street's average target of $267 while it offers only mid- teens upside while individual targets it ranges from 126 on the lower end to 400 at the upper end. So Marvel today looks based on these assumptions to be trading at a premium of around 10%. My verdict for the company it'd be avoid above $220. I become interested below $190. Google. It makes Marvel more credible, but definitely not cheap. And then we move on to Sand. This, which is the stock most likely to tempt value investors. It's gained more than 520% year to date. It's fallen roughly one/3 from its all-time highs at $2,355. And it trades at apparently a very tiny multiple. We also notice Quant give it the highest rating, 4.99 out of five. Wall Street sitting on the border of four and a half to flip it into a strong buy rating with Seek Alpha coming in by but weaker than both at four out of five and consent fiscal 2027 earnings of $214 per share rising to $265 based on 2028 numbers and at the pre-market price near $1,544. This is just under 7 times 27 earnings and less than six times on 2028. Now, the reason for this is data center demand. Quarterly data center revenue increased from under $200 million to nearly 3 billion in roughly 16 months. And management estimates that flash demand can move from a historically cyclical $60 billion market to in fact more than 300 billion in 2026 and approach 500 billion by 2027 with SanDisk also expanding its technology roadmap to deliver roughly 27% annual productivity improvement. If both claims hold, the economics they can change dramatically. And when we take a look at the company from a valuation basis, well, both forward earnings and enterprise value multiples, they look extremely cheap. Yet the current sales multiple on both on a trailing 12 month and forward-looking basis, they both actually look fairly expensive in comparison to the sector. It just shows you how much margin expansions already embedded in the earnings today. And then we take a look at growth A+. In fact, the percentages here, they're spectacular, but many compared to today's boom with a depressed base, four-digit growth rates, they're not sustainable operating assumptions. And margins and returns, they've surged as supply tightened. It creates enormous cash flow, but it also makes this company more sensitive to pricing and future capacity. Now, my model based the starting point 27 again on analyst projections around $15 billion of free cash flow and then assumes a middle rate here of 5% growth producing an intrinsic value headline of just under $2,400. That equates to a margin of safety around 36%. Now, if we were to change the required return here to 10%, well, we can see the value does drop sharply to $1,700, which actually still indicates a margin of safety, much less obviously around 9.2%. And Wall Street's average price target, well, that's above $2,100, indicating 42% implied upside. But in fact, their lower end, that's just $1,000. So, this is an enormous range for a stock which investors do describe as cheap. So, my verdict for SanDisk is a hold, not a buy around $1,500. I become interested below, 1300, much more comfortable near $1,200. SanDisk, it may deliver the highest upside, but this is still a cycle bet, not today's winner. We then move on to Micron, which is the strongest challenger for the number one position. Before looking at his valuation, listen to Micron's chief executive explain why he believes AI has permanently changed memory's role. You know today there is no AI without memory. Me AI systems need more memory. They need higher performance memory. They need lower power memory. So the value of memory has that equation has totally changed. They our customers to drive their own growth they need more compute. They need more memory. So memory really has become a key enabler. And Jim this is not only in data centers even in your phone you know to have richer experiences in AI enabled phone you need more memory content that is management central claim memory is no longer disposable commodity input now test that claim against the valuation micron will it trade around six times calendar 27 earnings versus approximately 17 as we can see for Nvidia and 19 for TSMC and the spending estimates explain that confidence. Hyperscaler outlays on HBM server DRAM and NAND, they're projected to rise by hundreds of billions of dollars. And both DDR4 and DDR5 spot price indices, they've turned sharply higher. Pricing powers moving towards memory supplies because demand is running ahead of supply. And Goldman's estimate shows DRM under supply of 5% in 2026 and nearly 6% in 2027. Small shortage can create very large price movements and NAND is also expected to remain undersupplied that strengthens the entire memory complex although it also encourages future capacity investment and Micron's HBM position has improved from around 7% of industry shipments in 2024 to an estimated 21% from 2025 onwards and their latest quarter shows the operating leverage 41.5 billion of revenue 35 billion of gross profit and 28 billion of net profit where Consens expects fiscal 27 earnings of $155 per share, more than double what we can see for the 2026 estimates. That implies a forward multiple near six. But free cash flow margin, well, it currently trails the headline profit margins. Micron must fund fabrication capacity before investors receive the cash. And my model is actually a lot more conservative starting point than what analysts are forecasting. We have $35 billion of 26 free cash flow, 45 in 2027. Then the 10% growth moving forwards. That indicates an intrinsic value of just under $1,200. Terms of a margin of safety, we arrive at 21%. But then if we were to change the discount rate to around 10%, we can see it actually drops and therefore we get no margin of safety, a premium sitting around 14%. where Wall Street's average price target sits just ab $1,500 implying around 66 p upside but again range very very wide $361 at the low end over $2,000 at the upper end. So my verdict for Micron be at hold around $940 I'd buy below $800 and become aggressive below $700. Micron is genuinely attractive but his valuation still depends on treating peak cash flow as ultimately durable. This leaves us with Broadcom is on the lowest multiple stock. It's not suffered the largest decline and its near-term valuation. While it's not obviously cheap, in fact, year to date is up only 4% over the last year, up 22% where it sits towards the lower end of the 52- week range. Strong buy from Wall Street, respectable 4.2 out of five buy from Seeking Alpha. And what Broadcom offers is the best combination of growth, profitability, diversification, and cash flow quality. Quarterly revenue reached 22.2 2 billion, up 48% year-over-year. And Semiconductor Solutions contributed $15 billion, while Infrastructure Software added $7.2 billion. VMware gives Broadcom a recurring revenue engine outside of AI hardware. And the semiconductor business is accelerating rather than merely growing. Quality growth advanced from 5% in 2023 to 79% in just the latest quarter. And forward revenue growth is estimated near 50%. Forward Ebitar 55% and forward earnings growth that's close to 60. Profitability also very strong. We can see sitting at an A+. Brocom, it also converts far more profit into cash than most AI suppliers. We can see the trailing free cash flow margin that sits at 36% and at roughly 31 times the current Ford estimates. The stock looks expensive, but fiscal 2027 earnings around $1920. Well, that reduces the multiple to around $18.4. 4 and you can see the valuation grade overall D minus it does remain poor. That's ultimately because when we take a look in fact sales book value and near-term cash flow multiples they sit well above the sector. In some cases we're talking a tripledigit premium. So this is a quality at a price decision not deep value. Also want to point out price to earnings growth moving forward sitting 65 47% cheaper in the sector 53% lower than their own 5-year average. But I do have to say the first major warning with the company's financing is reportly discussing more than $60 billion of debt for an AI chip deal benefiting Anthropic as well as others. But before deciding whether the risk disqualifies Broadcom, listen to this explanation of why vendor financing has become one of the AI trades most important fault lines. Yeah, I I have to be clear on this point because this is among the bigger concerns for investors is this concept of circular financing or vendor financing. And so the consortium that was announced with groups like Black uh Blackstone, Black Rockck, Apollo, KKR uh it shifts the risk, it shifts it off of Nvidia's balance sheet elsewhere and that is most appropriate. But to be clear, I don't love circular financing. But let's look through this in the eyes of Nvidia. Nvidia is generating roughly a billion dollars of free cash flow every 2 days. Now, this overall clip, it does concern Nvidia, but the takeaway applies across the AI supply chain. Finance demand can accelerate revenue while transferring risk elsewhere. It makes Broadcom's proposed guarantee impossible to ignore. And again, 2026 based on analyst estimates kicking off at $48 billion of free cash flow and it assumes the middle rate growth sitting around 15%. This equates to a value of $457 and we can note a margin of safety around $22%. And if we look at the conservative growth rate at 10% it gives us $320. But if we use this middle rate 15% increase the discount rate to 10% will we actually notice an overvaluation a premium factor sitting around 17%. Where Wall Street's price target comes to $526 implying 47% upside although we can see a low target of $216. It shows the downside if AI growth disappoints. So Broadcom it advances to the final comparison but it doesn't receive an automatic buy. its financing exposure and discount rate sensitivity must now survive the same stress tests as the other four. So now we have five impressive growth stories and five very different valuations. Before naming the winner, I want to compare them using exactly the same four tests. The tests are cash flow durability, dependence on Nvidia's next guide, valuation under a higher discount rate, and the size of the permanent loss risk. So we can start with cash flow durability. Every supplier benefits while hyperscalers transfers billions into chips and infrastructure. The harder question is what happens after the initial buildout. Electricity demand well it suggests this investment cycle could last for years but durable industry demand does not guarantee that every supply preserves today's margins or today's market share. Next is the discount rate. When long bonds yield more than 5% an 8% required return leaves only a narrow premium for owning a volatile semiconductor company. and Western Digital. It fails this test immediately. At an 8% discount rate, my value is already around $100 below the market price. And to justify today's price for WDC, it needs something close to my aggressive case. It leaves almost no compensation if storage demand, margins, or execution disappoints. Marvel's demand visibility, well, it's stronger because Google's attached war investing to actual purchase. It creates an unusually direct connection between customer spending and shareholder value. But the trade-off with Marvel's dilution, if the relationship becomes large enough to earn every warrant tranch, existing owners surrender roughly six and a half% of the postisssuance company. Marvel therefore in fact needs very high growth simply to reach fair value. A company can win strategically while it shareholders earn only an average return from an expensive starting price. And SanDisk, it has the most explosive bullcase. If management is right, that premium flash demand is resetting the market. While today's earnings estimates they could prove conservative and the data send acceleration supports the argument revenue didn't merely improve it multiplied is why a stock up more than 500% can still show a seven times forward multiple. But Sandis valuation depends less on 5% future growth than on whether $15 billion of staling free cash flow represents a new flaw or in fact a cyclical peak. If cash flow normalizes lower, the apparent margin of safety will it disappears very quickly. The low mod was real, but so is the uncertainty surrounding the denominator and Micron receives the clearest direct benefit if Nvidia confirms that accelerated demand remains supply constrained. More GPUs require more HPM and a DRAM shortage strengthens pricing and Micron's also gaining HBM share rather than relying on industry pricing. It makes this cycle structurally better than a simple recovery in commodity DRAM. However, memory manufacturing remains capital intensive. reported profit can surge before free cash flow catches up because new fabrication capacity consumes cash well before it produces revenue. And at an 8% discount rate, Micron looks undervalued when we change that in fact to a 10% when it becomes overvalued. The single change moves the answer by more than $360 per share. Broadcom, well, it's different. AI semiconductors drive the acceleration, but infrastructure software supplies recurring revenue, high switching costs, and cash flow that does not depend on memory pricing. And this 36% free cash flow margin also creates room to invest, room to repay debt, room to repurchase shares, and absorb a slower quarter without breaking the long-term thesis. The weakness is that reported financing proposal. If Broadcom guarantees customer debt, some demand risk migrates back towards the supplier. So I treat that as a position sizing constraint. Now consider three Nvidia scenarios. Scenario one is a genuine beat and raise with stronger demand, improving supply and confidence that customer spending will continue accelerating. That outcome helps every stock, but Micron Sandis probably react most because their earnings contain the greatest operating leverage to both memory prices and volume. Broadcom that would also benefit, but its diverse cash flows may produce a smaller immediate reaction. That's acceptable because I'm selecting an investment, not predicting one trading session. Scenario two is strong results, but only inline guidance that may disappoint a market condition to expect Nvidia to beat increasingly extreme expectations. In that outcome, while valuation matters more than the narrative, Marvel's 56 times forward multiple, and WDC's optimistic cash flow assumptions, it leaves them especially exposed. and Micron and Sandies have lower headline multiples, but investors may still sell them if Nvidia implies that memory availability is improving faster than expected. Scenario three, that's weaker guidance or evidence that customers are delaying projects that would challenge the entire free cash flow transfer supporting supplier earnings. And under that scenario, the stocks with the largest prior gains do not automatically have the smallest downside. Expectations, balance sheet exposure, and valuation determines what happens next. So, Western Digital is eliminated first. Its base case intrinsic value is already below the market price and the aggressive scenario provides too little upside for the uncertainty. Marvel, that's eliminated next. I like the Google relationship, but I will not pay a premium multiple while also accepting performance link dilution. SanDisk, it reaches the final three. Its upside could be exceptional, but I need a lower entry price because the model begins with peak looking cash flow. Micron reaches the final two. It offers the most compelling multiple rising HBM share and powerful pricing, but the value changes true dramatically when I normalize the discount rate. And Broadcom is the other finalist. It's more expensive today, but it cash flows diversified, recurring, and less dependent on one product cycle. So, the final decision is cheap cyclicality against expensive durability. Only one gives me enough upside, quality, and resilience to buy before Nvidia reports. Here's the final ranking. In fifth place, Western Digital, avoid above $435. The draw down is large, but the valuation still assumes a nearperfect recovery. In fourth place, Marvel avoid above $220. The Google agreement is strategically valuable, but growth expectations and potential dilution, they're already reflected in the price. In third, SanDisk hold near $1,500. Its forward multiple looks extraordinarily cheap, but the valuation rests on a $15 billion starting cash flow assumption. In second, Micron hold near 940 buy below 800. This is the cheapest established memory exposure, but I still want protection from the cycle. At first place, Broadcom is the only buy today. It has the best cash flow quality, the broadest business, and the clearest path to compound beyond one memory cycle. So, tomorrow I'm watching three things from Nvidia. First, whether data center demand is accelerating fast enough to support the forecasts that are embedded across this entire group. Second, memory availability and system pricing. Higher memory prices help Micron and SanDisk, but they raise the cost of every Nvidia system. Third, financing and customer returns. Supply revenue can look excellent today while the customers free cash flow deteriorates. That imbalance, it cannot expand forever. And if Nvidia delivers stronger guidance and long-term yields remain stable, this sell-off could create selective opportunities. If yields rise in guidance merely meets expectations, expensive suppliers, well, they remain vulnerable. It's why declining share price is not enough, a 44% draw down can still be overvalued. While a smaller decline in a high quality business can offer the better risk adjusted return, and Nvidia's results will move all five, but the report will not make their balance sheets, cash flows, or valuations equal. VI trade is one theme, not one investment. So my answer before the report is Brocom as a starter buy, Micron Sandis as holds, and Marvel and Western Digital avoids at the current price. But let me know which one you'd buy before Nvidia reports, which stock I should value immediately after the results, as well as signing up to the weekly newsletter by clicking on the pin comment below. More importantly, have a great day. I'll see you all on the next