This all-important metric is likely to signal that the euphoria surrounding artificial intelligence (AI) is beginning to fade.
Since 2023 began, no trend has been more responsible for lifting Wall Street’s major stock indexes to new heights than the rise of artificial intelligence (AI).
The allure of AI is the long-term capacity for software and systems to learn without human intervention. This gives AI-driven software and systems the ability to become more efficient at their tasks, and potentially evolve to learn new skills. With an addressable market that spans most sectors and industries, the analysts at PwC believe AI can add a jaw-dropping $15.7 trillion to the global economy come 2030.
Although dozens of companies have benefited from the AI revolution, none has been the poster child of success more so than semiconductor giant Nvidia (NVDA 4.55%).
Image source: Getty Images.
Nvidia is leading the next supposed leap forward in business innovation
In short order, Nvidia’s H100 graphics processing unit (GPU) became the go-to chip used by businesses to run generative AI solutions and train large language models (LLMs). With demand swamping supply, Nvidia has had no trouble meaningfully increasing the price of its H100 GPUs to between $30,000 and $40,000 per chip, or roughly two to three times what key rivals are charging for their AI-data center hardware.
The beauty of higher price points is that they have directly benefited Nvidia’s bottom line. Over the previous five reported fiscal quarters, ended April 28, 2024, the company’s adjusted gross margin has increased by close to 14 percentage points to 78.35%.
Nvidia hasn’t been shy about investing for the future, either. Its next-generation Blackwell platform, which is slated to hit the market next year, will accelerate computing capacity in six areas, including quantum computing and generative AI, and be more energy efficient than its predecessor. Meanwhile, in June, CEO Jensen Huang briefly teased the all-new Rubin GPU architecture, which will run on a different processor (known as Vera) and debut in 2026.
The final piece of the puzzle that’s helped Nvidia’s market cap grow by more than $2.8 trillion since the start of 2023 is its CUDA platform. This is the software platform developers use to build LLMs, and it’s working hand-in-hand with the company’s leading hardware to keep enterprise clients loyal to its ecosystem of solutions.
Although it’s been a seemingly perfect operating ramp, Wall Street is liable to see just how fallible Nvidia and AI as a technology are as a whole on Aug. 28.
This all-important figure from Nvidia could signal the bursting of the AI bubble
This coming Wednesday, Aug. 28, Wall Street’s AI darling will lift the hood on its fiscal second-quarter operating results.
Over the previous five quarters, Nvidia has done nothing short of obliterate even the loftiest analyst expectations. A combination of strong enterprise demand for its AI-GPUs, exceptional pricing power, and limited competition, has allowed the company to build up a backlog that would make any tech company envious.
However, headline revenue and profit figures aren’t going to tell the complete story come Aug. 28. Even if sales and profits sail past the consensus of analysts, another key figure can portend the end to AI euphoria. I’m talking about Nvidia’s adjusted gross margin. Nvidia’s “adjusted” gross margin excludes the impact of stock-based compensation, acquisition-related expenses, and a few other costs.
NVDA Gross Profit Margin (Quarterly) data by YCharts.
Following the release of Nvidia’s fiscal first-quarter results, Huang and his team offered adjusted gross margin guidance for the fiscal second quarter of 75.5% (+/- 50 basis points). This guidance implies a 235- to 335-basis-point decline from the first quarter.
While a median expected drop of 285 basis points in adjusted gross margin might sound like much ado about nothing considering the roughly 1,370 basis points Nvidia’s adjusted gross margin expanded by over the prior five quarters, it’s the reasons behind this forecast decline that are the real concern.
Nvidia’s compute advantages are unlikely to save it from the inevitable
Although demand has been undeniably strong for Nvidia’s H100 GPU, it’s the company’s pricing power that’s done most of the heavy lifting. Sales growth has handily outpaced an increase in cost of revenue, signaling that pricing power, fueled by persistent AI-GPU scarcity, is the company’s core driver.
The problem for Nvidia is that it’s not the only show in town. Advanced Micro Devices (AMD 2.16%) is ramping up production of its MI300X AI-GPUs, which are, on average, 50% to 75% cheaper than Nvidia’s H100. AMD also hasn’t been hindered by early stage chip fabrication supplier issues in the same way Nvidia has.
Furthermore, Nvidia’s four-largest customers by net sales — Microsoft, Meta Platforms (NASDAQ: META), Amazon, and Alphabet — are all internally developing AI-GPUs for their data centers. Even with these internally developed chips destined for complementary roles, they’re ultimately cheaper and more easily accessible than Nvidia’s hardware. These companies represent about 40% of Nvidia’s sales, and they’re all signaling a reduced future reliance on Wall Street’s AI darling.
To make matters worse, reports emerged a little over two weeks ago that Nvidia’s prized Blackwell chip would be delayed by “at least three months” due to design flaws and supplier constraints. Nvidia not being able to meet enterprise demand in a timely manner opens the door for AMD, Samsung, and Huawei to steal share.
Nvidia’s biggest gross margin lift has come from AI-GPUs being extremely scarce. But as new chips hit the market, and the company’s own top customers fill their valuable data center “real estate” with in-house chips, Nvidia will inevitably find that its pristine pricing power is eroding. The company’s median forecast of a 285-basis-point sequential-quarter drop in adjusted gross margin is evidence that AI euphoria is fading.
Image source: Getty Images.
When the AI bubble bursts, no company will likely be hit harder than Nvidia
Looking beyond Nvidia’s Aug. 28 report, history is another monkey wrench for the AI revolution.
Since the advent of the internet three decades ago, there hasn’t been a single innovation, technology, or buzzy trend with a mammoth addressable market that’s avoided an early stage bubble-bursting event. Without exception, investors always overestimate the use case(s) and consumer/enterprise uptake of a new technology or trend, which eventually leads to disappointment, euphoria fading, and a bubble-bursting event.
Including the internet, we’ve watched this play out with genome decoding, business-to-business commerce and networking, housing, China stocks, nanotechnology, 3D printing, cryptocurrency, cannabis, blockchain technology, virtual/augmented reality, and the metaverse.
To add to the point, you’ll find that few of the companies building out AI data centers have definitive plans for how they’re going to use the technology to increase sales and profits. For instance, Meta Platforms is investing more than $10 billion in Nvidia’s H100 GPUs, but has no immediate plans to profit from these investments in its AI data center.
The simple fact that most businesses lack a clear game plan when it comes to AI makes it crystal clear that we’re dealing with the next in a long line of bubbles.
This isn’t to say that artificial intelligence can’t, eventually (key word!), change the growth arc in a big way for corporate America — but there’s little question that the technology will need time to mature.
If the AI bubble does burst, as history suggests it will, there isn’t a company that’ll be hit harder than Nvidia. It’s adjusted gross margin in the coming week should provide confirmation that the beginning of this bubble-bursting event is underway.
