NVIDIA Stock to $60?
Table of contents
NVIDIA stock has been under pressure lately. Restrictions and tariffs could stand to hurt their international business, and fears over the success of China’s hardware-light LLM, DeepSeek, have investors worried. Could NVDA stock crash to $60 per share? In this video, we’ll talk about what might cause NVIDIA stock to crash and some important points for investors to watch.
I saw this clickbait headline on a Forbes piece recently – NVIDIA to $60. Could NVIDIA stock fall below $60 by the end of this year? Absolutely. They could announce a two-for-one split tomorrow. It’s not a good question to ask at all.
Here’s a better question to ask – What could cause NVIDIA shares to decline significantly in value? Today, we’re going to look at three scenarios that could cause a significant drop in NVIDIA’s share price. First, you could have consumers pull back on spending. Could existing hardware be used more efficiently? Absolutely. That’s what we’re seeing with DeepSeek. That means that a build it and they will continue to buy it may not hold true in the future.
That massive surge in AI funding for startups could result in a lot of fallout. Second factor – Margins normalizing from their currently elevated levels. Finally, the most visible factor – that would be geopolitical constraints decreasing demand for NVIDIA’s products. We’re going to talk about what steps the US may be willing to take to exert their dominance in AI.
Valuing NVIDIA
So when it comes to valuing NVIDIA stock, we typically look at price-to-earnings ratios and you can see they’re elevated at the time. You can calculate these yourselves. You should be doing that so you understand the mechanics of these. You can compare their current P/E ratio to the S&P 500 and see that it’s overvalued.
But then you can also look at expectations of future growth. That’s why tech stocks enjoy high valuations. When we look at the NVIDIA earnings trend over time, you can see here that earnings on a quarterly basis have been increasing sharply over time and, consequently, the share price has been doing the same.
Now there’s two ways to increase earnings. You can either increase your margins or increase your revenues. When it comes to their margins, these are artificially inflated due to high demand.
Risks for NVDA Stock
The challenge is going to be maintaining these juicy margins, not growing them. So when we look at their key customers, we see a lot of customer concentration risk. So somewhere around 34% or around 44 billion of their total revenues come from three customers and that’s about 39% of their data center revenues.
These are guessed to be the hyperscalers that are building out their data centers and any pullback in spending from these three large clients, or any major customer that they have, could be very impactful on NVIDIA’s earnings growth. Now when it comes to AI startups, I don’t think this is as visible to retail investors but, on our side, we’re always paying attention to startups getting funding so we can keep our fingers on the pulse of technology trends. And the amount of funding going into AI startups is truly astronomical. Not only just from an anecdotal perspective but also from the actual numbers.
So AI startups globally raised around hundred billion dollars in venture capital last year, so that’s said to be anywhere from a 60 to 80% increase from the prior year. And when you look at global VC funding, in general, in 2023, it was somewhere around $460 billion and it dropped about 25% in 2024. So even with that decline in VC funding, you saw a lot of money going to AI startups.
We know that nine out of 10 startups are going to fail. I think that number goes much higher when we start to throw common sense out the window in favor of hype. AI startups likely contribute to around 10 to 20% of total AI compute demand with the rest driven by hyperscalers and more established AI firms like OpenAI, which would still be considered a startup but not at the scale that they’re growing.
Now when capital goes away, these firms are going to be forced to use resources a lot more efficiently and also they’re going to either have to start paying their own bills or they’re going to go out of business.
How to Know if GPU Demand is Subsiding?
So as we’ve always seen in the past, high-profile startup failures really are the canary in the coal mine for when a particular technology boom is starting to subside. Now we haven’t seen many of those yet.
And I think when it comes to revenue growth for NVIDIA, you’re not going to immediately see a decline as you have that heavy hyperscaler spending which isn’t going to turn around very quickly. It will likely persist. Where you might start to see some failures are around the core type business model. So they’re currently handling the demand overflow for AI compute. And when the hyperscalers come online, they’re going to start to run into some problems. We covered that in our recent piece on the core IPO.
NVIDIA’s Extra High Margins
Now the first sign of demand erosion is going to be when you see the price premium that NVIDIA has enjoyed start to evaporate. That’s going to be reflected in their margins and I love this chart here.
On the top, gross margin; the bottom, operating margin that shows just how dramatically their margins have spiked because they enjoy pricing power. And when we look at why these margins have enjoyed such strong growth in recent years, very quickly, there are a number of drivers for that. So, of course, as I said, high demand limited supply leads to premium prices. That’s unique to NVIDIA’s data center segment. The rest of their segments aren’t enjoying that sort of high-margin increase. You can contrast data center to the gaming segment which is actually seeing their operating margins decline over time.
You also have the fact that NVIDIA commands 90% share of data center GPUs and that lets them maintain pricing power until they start to see competition. Now one thing that’s keeping competitors at bay is CUDA, the software component of their stack. But you also have AMD, you have hyperscalers looking to develop chips in house and, of course, China we’re going to talk about that. Now what that revenue growth allows NVIDIA to do is enjoy something called operating leverage where they can spread their fixed costs over a larger revenue base.
And, of course, you have high-margin software that NVIDIA pretty much has a 100% gross margin for. And there are China-specific chips which are said to have or at least have historically enjoyed a very high margin. Reducing demand and increased competition certainly will threaten their margins because they can no longer charge as much and we may see that starting to happen.
NVIDIA Export Restrictions
So this was a press release that came from NVIDIA rather recently. It would have been several days ago by the time this video goes out. And it talks about how NVIDIA now needs to have a particular license in order to export these H20 integrated circuits that they’ve been selling to China.
So these were custom-made chips for the Chinese and, as a result, NVIDIA’s response here is to recognize about $5.5 billion of charges associated with these products next quarter. So that short-term write-off impact really is just noise. It tells us a couple things, though. First of all, that if they were able to simply take those chips and sell them to somebody else it wouldn’t require such a large write-off. So what’s happening is that other customers aren’t necessarily wanting these throttled-down chips that NVIDIA has been building for China.
Now, of course, NVIDIA would have already been thinking through these concerns and it will be interesting to see how they respond to this. But what we’re more interested in trying to understand is what does the broader scale revenue impact look like for NVIDIA? So when we look at the various types of chips being produced by NVIDIA, there’s two in particular that have been custom-made for the Chinese.
The first is the H800 that now, in October of 2023, that was actually banned for export to China. And then you have the H20 and now that’s not being banned but it’s being rather restricted with these new licensing requirements. And what’s interesting is this piece that came out by this group called IFP and it’s labeled the H20 problem inference supercomputers and US export control gaps and it specifically points to, it says, our current mistake continuing to export H20 chips. And Reuters talks about how, in just the first three months of this year, Chinese companies including ByteDance, Alibaba, and Tencent placed at least $16 billion in orders for the NVIDIA H20 server chips.
And that would more than double China’s entire existing stock of chips. Some say this is on the back of the success that the Chinese have seen with DeepSeek. We’ll talk about that. Now when we look at the geographical revenue breakdown for China, trying to figure out how much they’re selling to China refers to the customer billing location.
So, for example, Singapore represented 18% of fiscal year 2025 total revenues when it comes to their booking. However, shipments to Singapore, so actual shipments of chips, were less than 2% of fiscal year 2025 revenue. And what’s interesting is when you look at this piece here by The Diplomat, it says, “Is China’s DeepSeek using smuggled AI chips from Singapore?”
And it says the US government is probing whether the Hangzhou, I was just in Hangzhou, based AI firm has built its disruptive model on restricted NVIDIA chip.
The DeepSeek “Threat”
So that brings us to talking about DeepSeek and China’s DeepSeek out of Hangzhou launched their latest AI models which are said to be equal to or better than industry-leading models in the United States. So the amount of training required for these for this model is said to be around $6 million worth of computing power from NVIDIA H800 chips, is what they said they use.
So before they were banned, they were able to get their hands on several thousand of those chips. The way that they’ve been able to more efficiently use this rather antiquated hardware and accomplish the same results is said to be memory-saving techniques custom-chip communication schemes that they’re using. And what this allows them to do is achieve higher performance with fewer and less advanced chips. So that means, essentially, that all these hyperscalers and everybody throwing money at NVIDIA don’t need to be spending that much money to accomplish what they want.
That’s the implication now with the AI industry shifting towards inference-orientated workloads, it also requires less computational power than training. So these were the concerns that rattled the market especially NVIDIA’s stock. Now NVIDIA argues that DeepSeek’s success demonstrates the need for more GPUs especially for inference and their H20 chip was optimized for this purpose. So they’re saying, “Well, you still need our hardware, anyway.” Well, fair enough.
But there’s also some speculations around what DeepSeek has accomplished that are worth noting. So they talk about, you know, around 2,000 NVIDIA H800 chips. But there may be this temptation to brag, going back to Miles’s great leap forward, they can say, “Well, you kept us from your tools and we still accomplished what you did.”
So they could be lying about that and downplaying the real hardware that they actually used and they may be actually purchasing way more than you think. You know that 18% that’s going to Singapore, well, it ain’t going to Singapore, right? So let’s get back to talking about the impact of these new licensing restrictions on H20 sales to China. So we need to figure out how many of those chips NVIDIA was selling China. You see the breakdown here. A fair number of these are estimates but we can sort of try and validate those. So the other thing to note here is that enforcement remains challenging.
So China can certainly be consuming indirectly a lot more chips than what appears to be on the white market tin, right? There can be an entire black market and there’s actually some evidence of there being an industry that’s growing up around trying to get some of NVIDIA’s more advanced technology into China. Now when we look at this piece by The Register, it talks about how, in 2024, NVIDIA, again, selling these chips legally in 2024, was going to sell about $12 billion worth of GPUs to China, so the H20s, right?
So when you compare that $12 billion to the $130 billion they realized in fiscal 2025, that’s around 9% of revenues. Again, these are legal revenues. So when NVIDIA commented on their Chinese revenues, this was back when they were referring to Q4 of fiscal 2024 and Q1 of fiscal 2025, they reference this mid-single-digit percentage. So everything seems to be pointing to around a 9% impact. That’s if these H20s completely vanish. And one would assume that this new licensing requirement certainly restricts them but they’re not likely to completely vanish.
So the question remains, what’s the net impact on growth? What you need to remember is that even a small increase in demand can have a much larger impact on NVIDIA’s pricing leverage.
Now that licensing process is going to slow things down. It seems to be based on a volume of chips. So really to restrict large purchases. Now China can find any number of ways to get around this. There’s evidence of them engaging with Taiwan Semiconductor Manufacturing Company using shell companies. We’ll talk about that in a second.
But what’s keeping the current US administration from restricting NVIDIA in any number of ways going forward in the name of national security or whatever reasons they might conjure up? And that can impact NVIDIA’s revenues. Again, a small drop in demand can have dire consequences for their margins and for earnings which lowers the valuation which impacts the share price. And that may just be the start.
AI Chip Curbs
So you may not be aware of this framework for AI diffusion that, just days before leaving the White House, Biden introduced. This comes into effect next month. And for most countries in the world, this framework limits the AI computational power in an entire country to a fraction of that of a single top US company.
So pretty much what they’re saying is that except for the United States and its 18 allies, the rest of the world isn’t going to have access to this technology coming from NVIDIA that would allow them to build advanced generative AI models. And when you look at who’s against this policy, it’s Oracle and NVIDIA. And part of this policy surrounds where you can build data centers and the implication there is that Oracle is planning to build a lot of data centers around the globe and they’re really impacted by that, that expansion.
And for NVIDIA it’s obvious why. They’re going to be selling less products as a result of this. So you need to pay very close attention in, perhaps, this quarter and next quarter to see whether or not this starts to impact NVIDIA and the language around that. And then another interesting thing to consider here would be that when you start restricting a country’s access to technology they’re going to want to build that themselves.
And this piece here from last year talks about how NVIDIA was facing local competition for its China special GPUs, the H20s from Huawei, which they have their own 7-nanometer technology and that’s really where the tech being sold to China is limited and they’re also developing 5-nanometer.
But what’s interesting is this piece by Tom’s Hardware that talks about how TSMC is facing a billion dollar US fine for making chips for blacklisted Huawei. And what they were doing is using a whole slew of shell companies to essentially purchase product and, apparently, TSMC wasn’t paying close enough attention to that and now they’re being fined.
Conclusion
So takeaways here is software going to eat the world. In other words, is the hardware that everybody’s buying – do they really need to be buying so much of that? And startups are where you’re going to presumably see such advancements.
I was talking to a firm in Silicon Valley that operates similarly to how CoreWeave does and they’re seeing technologies come out on the software side that are rather remarkable that sort of bypass some of the tech that NVIDIA has been relying on as a competitive advantage. Now everyone assumes that customers are going to be lining up to demand NVIDIA’s latest and greatest and maybe it’s time that we focus on maximizing the return on investment from existing spend.
There’s been some very critical commentary from Goldman Sachs around how there isn’t exactly visibility into the ROI for all this money that’s being spent. This time is not different. We’ve seen hype before and it remains to be seen if Gartner’s hype cycle is going to rear its ugly head in response to all the hype that’s been surrounding AI, right?
So there’s certainly there, there and there’s a lot of potential but as is always the case with advanced technologies, there is a hype phase that eventually will wreak havoc. Now necessity is the mother of invention and when hype gets disrupted itself and all the money stops flowing, then companies are going to start doing more with less especially if we stop giving them everything they demand. This invest-at-any-cost attitude is eventually going to come to an end. This time’s not different.
NVIDIA’s Future
So in considering NVIDIA’s future, is the strong demand for AI compute driving NVIDIA’s high margins? Absolutely. Is this demand going to persist for the remainder of time? Absolutely not. Eventually, we’re going to see pressure on NVIDIA’s margins. And there are multiple catalysts that can and will, as we’ve talked about today, lead to a decrease in demand.
Cut the Chinese off, guess what they’re going to do? They’re going to build their own substitutes. Now can they do that without TSMC? Well, I don’t know. There’s some good conversations to be had around that. The Financial Times says that China’s 7-nanometer and 5-nanometer process technologies are 40-50% more expensive than TSMC’s but their yield rate is less than 1/3. So they have a long way to go. But the more time goes by, the more likely someone starts to infringe on NVIDIA’s success story.
And that may bring up AMD. What about AMD? Well, we’re going to take a look at AMD very soon so make sure you’re subscribed for that video. In the meantime, I’m going to leave you with this great relevant piece that we did on what’s a fair price for NVIDIA stock. Thanks so much for taking the time to watch this video today.
Share
