Nvidia has surged this year with 241% gains YTD, which has more than doubled the returns of the FAANGs. This is no small feat considering it’s widely understood Big Tech is holding up the broader market. Valuations are stretched and leadership is only narrowing; to say there’s pressure going into Nvidia’s report this evening is an understatement.
The outsized demand for the H100 has led to historic moments as Nvidia is expected to exit this fiscal year with quarterly data center revenue of $14 to $15 billion compared to $3.6 billion per quarter at FY2023 exit. Should these estimates be correct (we will get the official guide this evening), Nvidia will end the year with a bang with approximately 300% growth in the final fiscal quarter.
Wow, what a year. Investors may not truly appreciate what Nvidia accomplished given a global pandemic and shelter-in-place orders fueled triple digit growth in tech stocks three years ago. Yet, what Nvidia accomplished was entirely due to product-market fit and design prowess with no end of the world scenario needed. It’s rare what Nvidia did, which was to ignite demand of enormous magnitude.
It’s well known my firm was early to this move in Nvidia with a bold analysis that claimed Nvidia will surpass Apple in valuation by 2026. You can look forward to my firm updating the long-term thesis in the coming weeks with details on how Nvidia will close-in on the next trillion in market cap. But in the near-term, Nvidia investors face what makes or breaks a portfolio, which is the inevitable moment of when Nvidia will top and sell off, how to handle these enormous gains, and if Nvidia can surprise the market again now that it was the defacto leader in the Nasdaq’s historic rally this year.
My firm strongly believes that simply picking a stock is akin to playing a fantasy sport, whereas discussing how to manage the stock is what separates fantasy from the live game. On Nvidia, we’ve been quite clear that we were net buyers in 2022 and we have been trimming the position to take gains in 2023. Meanwhile, Nvidia has remained our largest position until very recently when we put a different stock as first place and Nvidia as second place. Although we typically reserve our trades for our research members, we’ve been open about our strategy of active portfolio management with this spectacular, winning position. Judging by filings by famous hedge fund managers, we are in good company with this strategy.
Going into this highly anticipated report, I’d like to provide my readers with more information on how we are managing our Nvidia position and what to expect from the earnings report. This is a near-term analysis whereas our long-term thesis that Nvidia will surpass Apple in valuation is still firmly intact.
Neck-Breaking Release Cycle: H200 is Hopping Ahead
Nvidia has a near-monopoly in data center GPUs, and one of its strategies to protect its moat is to upgrade GPUs quickly to where it’s hard for AMD, Intel or custom silicon to catch up. The release cycle from the H100 to H200 is neck-breaking, as a typical cycle is two years whereas the H200 will ship in volume one year following the H100. The B100 based on the Blackwell architecture is expected to hit the market at the end of calendar year 2024 with the X100 following soon after.
Source: NVIDIA INVESTOR’S PRESENTATION
If 2023 was the year AI accelerators made their importance known to Wall Street, then 2024 will be the year that memory and HBM3/HBM3E makes its importance known as the competition is going head-to-head at memory capacity and bandwidth per GPU rather than compute performance. This further translates to mean the AI race is more focused on inference for the next generation of GPUs as the neural network can be run entirely in memory without the need to move data back-and-forth with the external memory. The H200 is the first GPU with HBM3e for 141 GB of memory and 4.8 TB/s bandwidth. This will result in 1.6X to 1.9X better inferencing performance than the H100.
To drive the point further as to how important memory will be in the next generation of GPUs, the compute performance from the H100 to the H200 is not changing much. According to what the industry has seen so far from Nvidia’s GPU HGX 200 systems, there will be “32 PLOPS FP8” performance, which would be achieved through eight H100s with 3,958 teraflops of FP8 each. The translation is that Nvidia’s H200 upgrade is strategically focused on memory, which also translates to Nvidia feeling pressure from AMD as the MI300X will be the first GPU to hit the market with the memory capacity and bandwidth offering full utilization to increase LLM inferencing performance.
By adding HBM3 and HBM3e memory, the compute engines get a performance boost, albeit at a higher cost as HBM3 costs 5–6 times more than typical DRAM. Fewer GPUs will be needed so the cost does not translate to an equal increase in total cost of ownership. GPUs with HBM3 and HBM3E will run compute-intensive large language models with fewer GPUs than is required with the H100s due to offering roughly double the memory. The need for fewer GPUs is accomplished by running LLMs in the memory. The H200 with 141GB of memory compared to the H100’s 80GB will reduce the number of GPUs required for running popular large language models.
If you read between the lines on the H200, then Nvidia is a bit nervous about AMD’s MI300X with the H200 serving as an attempt to bridge the H100 and the B100. AMD’s design more than doubles the memory of the H100 with 192GB HBM3 memory and 5 TB/s of bandwidth, and most importantly, will be out a few months prior to the H200. The MI300X was the first to run a 40B parameter large language model on a single GPU.
AMD should feel satisfied that it forced the near-monopoly leader to hurry toward releasing the H200 with HBM3e as an answer to the MI300X. We covered this in a deep dive for our premium members in July and reiterated it again in August when we covered our favorite memory stock.
What to Expect in the Upcoming Earnings Report:
The very quarter that Nvidia began reporting double digit negative revenue growth of (-16.5%) was the best buying moment. Near the bottom a year ago, our firm wrote for Forbes that Nvidia Was Ready to Rumble with the RTX 40 Series and the H100 GPUs. Notably, Nvidia is up 200% YTD yet is up over 300% since the October low, which is why timing matters.
One year later, and Nvidia is unrecognizable from where the company was exactly one year ago. For the October quarter, Nvidia is expected to report YoY growth of 169.6% to 171% for $16 to 16.1 billion and growth of 190.6% YoY growth for the December quarter. According to current estimates, the December quarter is peak growth.
Source: I/O FUND
Pictured Above: The very quarter that Nvidia bottomed in fiscal Q1 was the quarter that the stock was had its highest short interest since the Covid low as the product thesis was little understood at the time.
A beat is very important for Nvidia given the spotlight on this company. Demand is certainly there, and what instead is in question (into the foreseeable future) is supply.
Here is what the CFO stated on the last earnings call:
“We expect supply to increase each quarter through next year” and also “Demand for our Data Center platform where AI is tremendous and broad-based across industries on customers. Our demand visibility extends into next year. Our supply over the next several quarters will continue to ramp as we lower cycle times and work with our supply partners to add capacity.”
Where the market was a tad disappointed last quarter was when the CFO declined to elaborate on what percentage increase in supply she was expecting to see. The translation is that these are hard comps to compete with, and without a substantial increase in supply, the growth rate may have an inherent constraint given supply has already increased triple digits YoY.
The soaring demand for GPUs is evident in Nvidia’s growth rate. Per the Financial Times, Nvidia is planning to ship 1.5M to 2M GPUs next year compared to a target of 500,000 this year. Given this outsized demand, the hiccup is more likely to happen on the supply side. For this reason, we detail Taiwan Semiconductor’s chart below.
When you strip out data center revenue, what you have is an even higher growth rate for the data center segment of 239% to $13 billion expected this quarter. So, the question remains — — can supply continue to grow at these elevated percentages?
Source: NVIDIA IR
The data center segment is clearly the thesis but it doesn’t hurt that gaming has rebounded, as well, with 22% growth last quarter.
Source: NVIDIA IR
Last quarter, the gross margin improved significantly to 70.1% compared to 64.6% in Q1 and 43.5% in the same period last year. This was the best gross margin in Nvidia’s history due to higher average sales prices and some contribution from the increased mix of software.
Per the CFO: “software is a part of almost all of our products, whether they’re our Data Center products, GPU systems or any of our products within gaming and our future automotive products.” Separately, the standalone software business is worth “hundreds of millions of dollars annually.” As seen with our note on the H100 release from last year, its important investors are early to a tipping point. This is why we’ve been adamant that Nvidia’s true AI moment was in 2020 with the A100. If you bought the stock for the H100, you likely missed this year’s power move. The same will be true for Nvidia’s software revenue.
Regarding this quarter’s gross margin, management expects it to expand to 71.5% in the upcoming quarter. The operating income grew by an incredible 1,263% YoY to $6.8 billion, which shows the cyclical nature of semiconductors. The operating margin was 50.3% compared to 7.4% in the same period last year. The management guidance for the next quarter is 53.1%. Typically, Nvidia’s operating margin is in the 30% range.
Source: NVIDIA IR
This has flowed through to the bottom line with Nvidia’s adjusted EPS up 429% YoY for $2.70 compared to 481.3% growth expected this quarter for EPS of $3.37.
Source: SEEKING ALPHA
Nvidia has the strongest cash flow margins among mega cap stocks. The operating cash flow margin is 47% with a free cash flow margin of 44.8%. In addition to higher revenue helping the cash flow, there was also $1.25 billion in customer payments received ahead of the invoice date.
The company has cash and marketable securities of $16.02 billion with debt of $9.7 billion. Last quarter, there was $3.28 billion shares repurchased. The Board of Directors approved an additional $25 billion in stock purchases with $4 billion authorized remaining at the end of Q2.
Data Center Assumptions
I/O Fund Analyst Notes on Nvidia’s Data Center Segment
The magnitude of Q4 guidance will be very important given heighted expectations. Assuming Nvidia meets its Q3 guidance of $16B +/- 2%, we’ve put together a simple scenario analysis to parameterize the different outcomes anticipated based on Nvidia’s potential Q4 guidance.+/- indicates anticipated stock positive or negative price performance on the next trading day based on that scenario.
Source: I/O FUND
At $40,000 per H100, that equals $28B in H100 sales alone, and when you add the A100 and other data center sales at a current run rate of $14B, the Data Center segment could report total revenue of $42B in FY24 (CY23). When you equal this out across the upcoming quarters, it looks something like this based on our estimates and Piper Sandler estimates.
Nvidia’s Data Center segment could report total revenue of $42B in FY24 (CY23)
Source: ESTIMATES FOR DATA CENTER REVENUE FOR Q3 AND Q4 FROM PIPER SANDLER
We believe the market will reactive negatively if Nvidia provides F4Q24 (Jan-Q) guidance that is in-line or lower than consensus growth of 11% Q/Q for the Jan-Q.
On the flip side, Nvidia will likely need to provide guidance of at least greater than 20% Q/Q growth for a significant positive reaction. This is because consensus will need to make upward revisions to their earnings for the remainder of FY24 (CY23) and FY25 (CY24). This is critical to support the current valuation with NVDA trading at ~45x NTM Non-GAAP P/E in-line with its 5 year average of ~45x NTM Non-GAAP P/E as of Monday November 21, 2023.
Our base case assumption is that Nvidia’s F4Q24 (Jan-Q) guidance will estimate Q/Q growth of at least +20%. Recall, H100 was only introduced to the market toward end of CY22. The Apr-Q was the very first quarter when Nvidia was beginning to see the impact of AI and demand for the H100. Piper Sandler believes Nvidia will close out the year with data center revenue of $42B and 2H23 Data Center revenue ~88% greater than 1H23 revenue.
Furthermore, we believe if Nvidia maintains its ~35% beat that it had for the Jul-Q for the rest of FY24 (CY23), Nvidia can potentially do $52B in Data Center revenue for FY24 (CY23).
Looking ahead to FY25, we believe Nvidia can do ~$92B in Data Center revenue based on our estimates for % beats for Actual Data Center revenue vs. Estimates for Data Center Revenue (Piper Sandler).
Source: I/O FUND
Exploring Scenarios for the Upcoming Earnings Report:
The neutral-case scenario is that Nvidia reports in line, but can’t give the Street what it wants, which is a raise on already impressive growth to help sustain the market leader’s gains this year.
If investors are being realistic, a raise is best left to next quarter when the company typically offers a fiscal year outlook. The question is not whether Nvidia is a top AI stock, and has a promising future (of course it does). The question at hand is whether Nvidia can produce a report that pushes buyers off the sidelines. These are two different matters, and are often in opposition after a large run-up in price.
The best-case scenario is that Nvidia’s been downplaying its supply (just a touch) and there will be a beat for the fiscal Q4 guide. Nvidia’s story is quite clear, which is that the data center segment is producing historic growth and the bottom line is so beautiful, you have to squint to make sure it’s real. If this happens, we could see the price go into the mid-$500s before technicals are predicting that buyers will be exhausted. As a reminder, that’s only a 7% move from where the stock is trading now.
Piper Sandler has a data center estimate for fiscal year 2024 of $42 billion, which translates to $14.9 billion in data center revenue if we assume $13 billion this quarter. We detail below the price targets we are eyeing to take more gains should Nvidia report a beat on Q4FY24.
The topping-out scenario is that Nvidia’s buying is exhausted, and there isn’t one fundamental analyst on earth that can help investors figure out when this will happen. That is best left to somewhat-esoteric technical analysis. As you’ll note, I am not calling this the bear-case as there is not a bear case for Nvidia. Even if the company loses China entirely due to restrictions, it’s likely that demand gets absorbed. However, there is a bear case for the semiconductor sector, of which Nvidia is exposed to, and I detail this for you below.
Regarding the topping-out scenario, it’s unlikely Nvidia has a major negative surprise to the downside as semiconductors have strong visibility compared to, say, an ad-tech company. The management team should be going to great lengths to be consistent and accurate with Wall Street given the long golden roadway in front of them. Therefore, the topping-out scenario is aptly named as a 200% gain means you’ve got to impress the Street to keep those gains, and Nvidia may need to refuel for a quarter or so until we can get to a new fiscal year guide next quarter.
The Red Scare
What’s not to be forgotten in the excitement of the product road map is China, which has been the predominant risk for semiconductor stocks dating back to 2018. Last year, the government restricted Nvidia from selling its two most powerful chips to China, the A100 and H100. To circumvent these restrictions, Nvidia designed slightly less powerful chips called the A800 and H800. As reported by Reuters, the H800 has as much computing power as the H100 in certain settings. For the United States, these chips are important to block as they strengthen China’s military.
Last month, the U.S. Department of Commerce announced updated rules focuses on computing performance by removing the bandwidth parameter and focusing exclusively on how powerful a chip is, as well as performance density, which will prevent companies from working loopholes. According to an official who spoke to Reuters, “the U.S. will require companies to notify the government about semiconductors whose performance is just below the guidelines before they are shipped to China.”
Although this is a medium-term issue for Nvidia, analysts believe the demand is high enough today that the company shouldn’t have any issues absorbing the 20% to 25% loss in its data center segment from tighter export restrictions to China. Looking further out for FY2025, Keybanc sees a $5 impact to Nvidia’s $25.62 EPS estimate, and up to a $20B impact to its data center segment with current estimates at $101B for the data center in FY2025.
Eventually, demand may settle — especially as more competitors step up — and investors should pencil-in losing China revenue as a risk that is materializing now, with the revenue impact likely to be felt in FY2025.
The Topping Out Scenario
Nvidia continues to push to all-time highs, which is a scenario that was outlined in our prior free report on NVDA in September of this year. In the last analysis, the I/O Fund Portfolio Manager stated: “as long as we hold $340, Nvidia has the potential for one more swing higher into year-end/early next year.”
The primary scenario presented had the $545-$574 region as the target for the next swing higher. As of today, we are about 7% away from this target in what appears to be the final 5th wave in an uptrend off the October 2022 low.
I have laid out two scenarios that I continue to see playing out in the coming weeks-months:
- The topping-out scenario has NVDA in a complex topping pattern. We would see a sharp reversal from current levels that would ultimately break below $435-$419 support region. This would signal that the top is in, and we would then setup our downward targets to start accumulating again.
- The bull scenario has us already in the final swing higher. Our targets are between $545 — $575 for this move. If we end up seeing a gap and continuation higher from the earnings report, then we would get a direct path to these targets. We would use this strength to continue to trim. The other scenario is that we see a slight pullback that holds the $435-$419 region, which would set us up for a push into higher targets in the coming months.
Source: I/O FUND
Semiconductor Industry May Be the Achilles Heel
Nvidia could certainly miss, yet it’s less likely given the company has outsized demand and visibility on supply. Within this context, it is easier to see the level of risk with interrelated stocks. One chart that is quite concerning, which has ramifications for all of tech, is Taiwan Semiconductor (TSM)
Source: I/O FUND
The bounce from the October, 2022 low is clearly an overlapping and messy move higher. This is common of B waves. What’s concerning is that the drop from the July high is a 5 wave pattern that broke through the major trendline. This would be wave 1 of the larger © wave.
What followed is a 3 wave retrace, so far, which would be wave 2. If the next drop is a 5 wave pattern that takes us below $89, it will be a strong warning. On the other hand, if we can see a vertical move over $104, then it will shift the odds away from the red count above, and suggest that we could see a larger swing higher into early 2024, which would be the green count.
We do not own TSM as we closed this position, yet one reason we are watching this chart is to help manage our semiconductor positions as a break below $89 is concerning enough to have a read-through to our other positions. In this case, we will likely hedge the semiconductor stocks that we have identified as those we want to own in a downturn.
A break below $89 could also be concerning given TSM is in the crosshairs with China, and the United States recently tightened export restrictions to effectively cutoff AI chips. China has made it known they are pursuing domestic silicon, and if so, TSM may become stuck in a tug-o-war on which country gets 3nm, 4nm and 5nm supply.
The Broader Semiconductor Sector (PHLX)
The PHLX Semiconductor Index is a popular index of the broader semiconductor sector. It currently has the same downside setup that we are seeing in TSM. However, it is moving up into major resistance and into a cycle that suggests a reversal is likely to follow.
Source: I/O FUND
The fan placed at the October, 2022 low represents a series of important angles that the PHLX has been using in its push higher. The red 1x1 line is a true 45 degrees off the low, and is the most important angle in defining an uptrend. Note how price broke below it and is now testing this angle as resistance.
Furthermore, those symbols above price represent cycles that we see within the PHLX. Note how price tends to reverse the trend that is moving into them. So, regarding these cycles, how we trend into them is the most important thing. We are currently trending up, into the current cycle, while testing the major angle in red.
It is likely that the broader semi sector sees a reversal soon, and until the PHLX can retake the red 1x1 angle, the pressure and risk remain to the downside.
Nvidia’s earnings outcome is not easy to read in the tea leaves. This is because the fundamentals are the best in the S&P 500 and the CFO has been clear that she has strong visibility for this quarter and into next year. It’s possible the company misses, but not probable (outside of something China related). Rather, Nvidia’s issues are sector-wide as semiconductor indexes and the bellwether TSM are looking weak on technicals. This would signal even if Nvidia beats/raises and the stock goes up, that its peer group may weigh on the company’s price action in the near-term. There’s also immense pressure that Nvidia raises, which may not be realistic given constraints on supply.
We’ve been crystal clear in both August and September that Nvidia has a move to $545 to $570 and this could mark the top. We continue to believe this is the price target where our firm will again take gains. If we don’t get there this evening, and price breaks down, then we will also take gains. In our opinion, this is the only way to procure a win-win scenario with a stock that holds a leading allocation in a portfolio that has extended 200% in one year.
Meanwhile, you can look forward to an update on how, exactly, Nvidia will surpass Apple’s valuation by 2026 in the coming weeks.
Our premium members will receive our post-earnings analysis this evening after hours. If you own Nvidia stock, or are looking to own NVDA, we encourage you to attend our weekly premium webinars, held every Thursday at 4:30 pm EST. Next week, we will discuss our plan following NVDA’s earnings, as well as a handful of other AI plays for 2024 — what our targets are, where we plan to buy as well as take gains.