Nvidia And Micron: Recession Headwinds Are Tailwinds For AI (NASDAQ:MU)

What do you think of when I mention the word recession? Rising unemployment, perhaps? Demand destruction, maybe? Even layoffs, declining profits, and negative revenue growth as you shift toward thinking about your investments? Likely nothing positive, if I had to guess. But it takes counterintuitive thinking to find the positive in a recession story, so I don’t blame you. Data center and AI (artificial intelligence) names are at the center of my investment screen as the market and world stare down a US recession because of this counterintuitive thinking, however. By the end of this article, it will make a lot more sense how specific tech names will benefit from a recession.

Let’s take a walk.

We both know the pandemic pulled in cloud spending as everyone in the world needed to work, learn, and play from home. This pressured the data centers of the world to operate at a capacity never seen before – expansion came quickly. It further fueled an already lit fuse called the digital revolution – anything and everything was being moved to the cloud. Consumers, businesses, enterprises, and governments were all making increased use of this utility, mostly out of necessity at this juncture.

Then the pandemic began unwinding, and the pull-in of the last year-and-a-half worth of spending needed to be digested. But with the unwinding of the pandemic came with it an odd tug in the other direction: Workers weren’t returning to the workforce at the same rate they left it (forced out) at the start of the pandemic, at least not linearly. As a result, corporations and businesses now struggle to find workers, from convenience store clerks to highly skilled software engineers to data entry specialists to teachers to nurses. Some industries have had higher hire rates than quit rates, but the number of jobs needed remains higher than unemployed workers in many industries.

This is indeed a problem if the economy is to thrive.

And, as a quick aside, maybe it’s part of the reason GDP can’t keep up with inflation, putting us further in the kill zone for a recession.

Some companies are starting to work around the labor shortage, however.

This is a good time to sit before we continue walking.

AI Has Become Necessary

Companies in the last decade have been working toward replacing their workers. If AI and ML (machine learning) can do the same job as well or better than humans, why not put that cost in the cloud where it’s scalable and repeatable. Servers are less expensive than human resources in the long run, plus silicon depreciates!

Just think of McDonald’s (MCD) to start – order your meal yourself on those friendly little kiosks. It’s a simple way to replace the job of a worker, and it’s not a terribly complicated technological hurdle. Now, remember, this was being done years ago.

Looking at more recent and more complicated technology, Meta Platforms (META), for example, would love to replace the massive amount of content moderators it hired over the last several years. Naturally, it would cut down substantially on human resource costs and employee benefits (read: stock-based compensation).

And for the first couple of years working on [AI], we’re still at a relatively low recall rate where our AI systems had 10%, 15% of the content that we were addressing, we were dealing with proactively. But in recent years, the AI progress has been very impressive. We’re now above 90% of the content that we take an action on…

– Mark Zuckerberg, Facebook CEO, Q3 ’21 Earnings Call

To summarize, in the early AI days, only 10%-15% of the content was able to be caught by AI for moderation. This means a lot of manual human intervention had to be called upon to moderate the massive amount of content on Meta’s platforms. But today, over 90% of content moderation is being processed by AI, lessening the burden on humans to do the moderation.

The effects of these investments have allowed the company to slow content moderation jobs and further transition to machine learning and AI jobs. These highly technical jobs currently comprise just under 13% of all open positions Meta is seeking. But the volume needed for highly specialized jobs is far less than content moderation jobs, which has slowed the company’s headcount growth over the last four years.

Beyond content moderation, Meta also is utilizing AI/ML for advertising to serve up better ads in general but also more relevant and better content which can lead to targeting ads to different and, more importantly, better audiences for the advertiser.

…we are investing significantly in AI and machine learning investments to power ranking and recommendations for things like Ads, Reels and Feed. And so that does add to the CapEx intensity of the business. And we do think there is additional capital intensity of the business as we make significant investments in AI and machine learning on top of just additional capacity growth.

– Dave Wehner, Facebook CFO, Q1 ’22 Earnings Call

Keep that capital intensity commentary in the back of your mind because AI’s “end customer” isn’t the only one taking in the beautiful scenery on the AI train.

With that, let’s keep walking – we’ll need to stretch our legs to understand this next section.

Seeing It From The Other Side

As you know, AI and ML don’t just start producing answers, data, what have you – they require infrastructure, hardware, software... accelerators to do it faster. This means it needs fast, efficient, and scalable technology to provide the foundation for AI to do its job. This is where a company like Nvidia (NASDAQ:NVDA) comes in.

Nvidia is known for its cutting-edge GPUs. In the race for the next advanced graphics or accelerator technology, Nvidia is the leader. This is undisputed as the most recent push into real-time ray-tracing was pioneered by Nvidia, with AMD (AMD) following it up in a behind-the-leader manner. This is exemplified by AMD’s adherence to the DirectX12 ray tracing API, which Nvidia developed.

Don’t jump on me here just yet for putting AMD behind Nvidia in the GPU category – AMD will have its time here in a bit.

I point out Nvidia’s GPUs because they have become some of the best processors for performing AI/ML workloads. Nvidia’s Datacenter division is now its leading money maker in absolute dollars and growth rates. It appears it will be this way for the foreseeable future, at least on a revenue growth basis.

But the most interesting part is how Nvidia sees right through the revenue into the heart of the matter. It isn’t just selling blindly into the market. On its last conference call, CEO Jensen Huang mentioned the critical point regarding customer needs and led me to the recessionary thinking.

…live customer service being now supported by AI, conversational AI has an opportunity to enhance the customer service on the one hand. On the other hand, supplement for a lot of labor shortage.

– Jensen Huang, Nvidia CEO, Q1 ’23 Earnings Call

It was nearly a passing comment, but it was huge as I look around today and see businesses unable to hire to the level they need. fuboTV (FUBO) was one of the more recent ones I’ve come across, unable to meet advertising needs because of unfilled positions in a tight labor market.

Some of those delays were really based off the fact that obviously hiring right now is not an easy task. And so resource allocation internally to be able to develop that capability, I think it’s been a little bit slower than we had anticipated.

– Dave Gandler, fuboTV CEO, Q1 ’22 Earnings Call

And more recently, with conditions expected to worsen in the economy, some companies have already put hiring freezes in place while others are beginning to lay off employees almost preemptively. Companies have gone from being unable to fill their roster to enacting hiring freezes to outright layoffs rather quickly – all within two years.

Going back to Meta Platforms, the company implemented a hiring freeze a little over a month ago, according to an internal memo from the CFO given to 9to5mac.com. Four factors were involved in the decision, but none more important than the slowdown in revenue growth and macroeconomic uncertainties. However, even with the downbeat message to his employees, the CFO still put in one crucial piece that not everything will be frozen:

…we feel optimistic about the opportunities to re-accelerate growth, particularly through unlocking new revenue streams through Reels monetization, [and] incorporating AI into our ads system to overcome signal loss from iOS changes…

In other words, even with everything happening macro-wise and its struggle to overcome iOS targeting changes, AI is still on the list of things to continue to invest in, and, in most cases, accelerate investments in. This memo encompasses what any major technology company is trying to do: Reduce expenses while utilizing AI to compensate for a lost workforce. But not just make up for a lost workforce and reduce costs, but chief among the reasons is to produce better and more efficient growth than it would sans AI.

AI is not just a replacement for an uncooperative workforce environment but the answer to producing better results, even if nothing else was driving it.

AI is key to a recessionary environment. It can make a company leaner and produce better results. It’s hard to achieve that outcome by simply cutting your workforce. Sure, you’ll reduce expenses, but it’s at the expense of undercutting your product. AI is needed to push your company and product forward, but it becomes necessary when cutting your workforce to avoid being left behind during the worst days.

A recession will become the “necessity” for AI investments just like the pandemic was for moving to the cloud.

Alright, that was a lot of walking. Let’s sit down and discuss who benefits here and what stocks to buy for the AI tailwind.

AI Tailwind Beneficiaries

Nvidia

I’ve already gone into one of the primary beneficiaries of this accelerating AI train: Nvidia. A lot of AI is possible because of Nvidia’s accelerators, such as the A100 and now the H100 using the latest Hopper architecture. Because of this, Nvidia has a greater than 80% market share of the data center for AI workloads. Therefore, it already has a large foundation of customers using its hardware and software stack.

You might say Nvidia is susceptible to competitors stepping on its turf and taking away customers. This is possible, and if Nvidia slacks on any one front in the AI game, it deserves to lose to its competitors. However, don’t forget what I just mentioned about its CEO’s focus – it knows not only how to produce fast, efficient AI/ML hardware and software but also why it’s doing it and why its customers need it. This kind of leadership allows Nvidia to work with its customers directly to give them what they need. It’s the beauty of owning the hardware and software stack to deliver a complete system – it can tailor the product to the requirements.

AMD might produce good and perhaps even better hardware, but it lacks in the software department. This is the significant difference between the two companies. Nvidia’s software team is what has given it the market share it has. But its latest secret is its implementation of Mellanox’s networking directly into its systems allowing for six times faster network connectivity and NVLink, which allows up to 256 H100s to be connected at nine times wider bandwidth.

With H100 now being delivered, customers who already have implemented A100 systems can upgrade to produce near instantaneously faster results and larger workloads without losing their progress in software. Nvidia isn’t slacking on getting its customers the latest and greatest, and just at the right time as AI investments ramp up into the necessity of recession.

According to electronicsb2b.com, Nvidia claimed $3.2B of the industry’s $4B for the global artificial intelligence processors in the cloud and data center market in 2020. By my calculations, Nvidia is set to grow its AI revenue by 50% in 2022 after growing it by 59% in 2021. Furthermore, Nvidia is working within an industry expected to grow nine times to $37.6B in 2026 from 2020. So even if Nvidia’s 80% market share drops to 60% by 2026, it will generate $22.6B in AI revenue. This would bring its AI revenue to 36.4% of its estimated 2026 total revenue compared to just 22.5% of its expected 2022 (FY23) revenue.

While many concern themselves with the effects of crypto mining on Nvidia’s revenue growth and performance, AI will be its leading source of growth over the next several years. This is especially true as its overall Data Center division continues to grow at outsized rates compared to the rest of the company.

After its release in 2020, the A100 contributed significantly to Nvidia’s Data Center growth in calendar 2021 (FY22), “led by hyperscale customers fueled by continued rapid adoption of Ampere Architecture Tensor Core GPUs, for both internal and external workloads,” according to the company’s CFO in last year’s FQ3.

And, if the company’s CEO couldn’t be prophetic enough, in November of last year, he was well ahead of the curve, calling 2022 an AI year by saying, “…we expect next year the cloud service providers to scale out their chip learning and their AI workload really aggressively.”

Right on cue for the H100 to take the reins.

The H100 will start to ramp up Data Center revenue again in FY24 as companies take delivery of the new product. Add in the added build-out for a recession scenario, and Nvidia will likely be tapped out of data center accelerator supply for the next year and a half.

However, the most important piece to Nvidia’s success is not its hardware but its software. This is where the company has beaten its competition many times over. The latest instance of this is Intel’s (INTC) Arc GPU, where its recent China release is having cold water splashed on it. Initial reports are showing it is less impressive than initially thought, and the driver situation appears to be at the center of it:

At the same time, there are already genuine concerns around Intel’s Arc desktop GPUs, and particularly reports regarding the graphics driver, which seemingly can be wonky in some cases (with certain games suffering from serious woes or not even running at all). This is the main theory as to the reason for the delay of Arc desktop…

Nvidia’s ability to produce solid software and drivers is how it can take good or great hardware and make it perform exceedingly well. AI is dependent on stable, high-performance drivers and software. If for nothing else, I’m a buyer of Nvidia based on its software engineering team alone, especially as AI becomes a growing need in a recession.

Micron

Now you might be wondering why I’d go from a fabless software house with hardware to a “commodity” memory player to take advantage of AI in a recession. This is simple, really. Micron (NASDAQ:MU) doesn’t just sell chips into the open market. No, if you are honest about your homework, you’ll know Micron has moved heavily into a complete or nearly complete solutions market where its customers are buying chipsets and not merely DRAM chips.

Moreover, Micron is the only supplier of X-type graphics memory. It’s the sole supplier of GDDR6X memory and is the memory used in all of Nvidia’s products above the RTX 3070, including the RTX 3070Ti.

But, moving into a broader sense of how Micron benefits, it comes down to the company’s ability to shift its needs toward the demand and do it quicker than at any time in its history. It did this in early 2020 when mobile demand cratered due to the pandemic while server needs were in extremely high demand to expand cloud capacity. This year’s amount of heads up is significant in terms of being able to remain on a growth trajectory while everyone and their mother talks about PC and mobile weakness in 2022 while server and data center will remain tight.

I find it interesting the focus has been on PC, consumer, and mobile demand while server and graphics have higher margins, comparatively, and in this AI-driven scenario, will sustain these categories of DRAM prices. Trendforce, which has been wildly inaccurate over the last year while Micron’s revenue and earnings have continued to grow, is expecting a 0%-5% decline in server DRAM for Q3 ’22. Those are not concerning numbers! That means it may not even change! A 1%-2% decline won’t even affect Micron as it produces the leading DRAM node in the world, meaning it has the best cost benefits out of its two other competitors.

The final reason Micron stands to benefit the most is where its stock is in relation to what a typical memory cycle would encounter. When memory is on the slide – actually on a slide, not 0%-5% memory declines – its stock will retreat to one time book value. With the latest dump in the market, Micron has moved to 1.33 times book value.

The valuation is already within 20% of this one-to-one book value without memory declining (anyone remember the DRAM Winter?). This doesn’t mean when memory prices actually decline Micron will suffer another 50% decline. No, it means the market is so far past its skis that it’ll begin pricing in any upturn in DRAM prices well ahead of when they happen. In short, the downside in Micron is 20%, while the upside is 93% (2.4x book value), as is historically true time and time again.

Furthermore, the recession fears won’t be nearly as impactful to Micron as they would otherwise because not all end markets will decline. AI will continue to be invested in even as all other line items get cut at major tech companies. Auto manufacturers learned this lesson firsthand and continue to pay for it to this day. The diversification Micron has moved toward over the last decade would not only sustain it comfortably during an economic recession, but an end market that will buck a recession trend won’t cut Micron’s estimates as far as the market thinks.

Ready to get to our destination? Let’s finish our walk.

Not Your Father’s Recession

The bottom line is this recession won’t see all markets and all walks of technology impacted equally. Instead, we’re going to see and already seeing an investment in technology to combat the headwinds of an economic slowdown, specifically in AI/ML. The use of AI is more than just a cool gadget or artificial lifeform to improve your lifestyle. Instead, it’s a business advantage to replace a workforce it can’t hire or a workforce needing to be laid off to reduce costs. Not only can AI be depreciated on the income sheet, but it’s also many times more efficient than a single human and will truly bring on the next chapter of workforce creative destruction.

This AI movement has beneficiaries, including the hardware and software companies developing the backbone. There are more than just Nvidia and Micron, too. In fact, AMD and Arista Networks (ANET) are my next two favorites able to benefit from the increasing move toward AI. While I didn’t get into them in this article, they, too, have a significant role to play in the expansion of the AI cloud and its uses.

For now, with Nvidia and Micron at valuations not seen except during economic slowdowns, it’ll be an even greater reward when the fallout isn’t as impactful to them as many other parts of the economy. AI will be the investment during this downturn, let alone a crossed-off budget item.

Hopefully, you enjoyed our walk.