If there's one certainty on Wall Street over the past three decades, it's that investors always flock to next-big-thing investments. There have been no shortage of new innovations and game-changing trends that have captivated investors' attention, including the advent of the internet, businesses-to-business commerce, genome decoding, 3D printing, blockchain technology, cannabis, and the metaverse, to name a few.

At the moment, no innovation is garnering more attention than artificial intelligence (AI). AI involves the use of software and systems to oversee tasks typically assigned to humans. Incorporating machine learning allows software and systems to grow smarter and evolve over time. This evolution is what gives AI utility in virtually every sector and industry.

A closeup of an Nvidia A100 graphics processing unit.

The Nvidia A100 GPU is a cornerstone of high-compute data centers. Image source: Nvidia.

Based on a report from PwC that was released earlier this year, AI has the ability to add $15.7 trillion to the global economy by 2030, with the biggest gains being seen throughout North America and China. Any multitrillion opportunity is going to attract investors -- and it's why graphics processing unit (GPU) kingpin Nvidia (NVDA 6.18%) has seen its share price ascend to the heavens.

Nvidia is poised to be the infrastructure backbone of the AI movement

Over the past two quarters, Nvidia has completely blown the doors off of Wall Street's consensus sales forecasts. Prior to reporting its fiscal first-quarter results (ended in late April), analysts had been expecting $7.2 billion in forecast sales for the fiscal second quarter (ended in late July). Nvidia offered guidance of $12 billion.

For the fiscal second quarter, it proceeded to surpass its own revenue forecast of "$12 billion" by $1.5 billion ($13.51 billion in fiscal Q2 sales), then issued sales guidance of $16 billion for the fiscal third quarter, which was about $3.5 billion above Wall Street's projection. No matter how high Wall Street set the bar, Nvidia has had absolutely no trouble surpassing it.

Nvidia's strength in numbers is almost entirely due to strong sales of its AI-driven A100 and H100 GPUs, which are used in high-compute data centers. AI systems require split-second decision-making and ultra-fast data processing, which is where Nvidia's high-powered GPUs come into play. The $10.32 billion in data-center revenue Nvidia tallied in its recently reported quarter is a cool 171% higher than the comparable period in fiscal 2023

Based on analyst expectations, Nvidia accounts for around 90% of the GPUs currently being used in high-compute data centers. With AI utility still just scratching the surface, the sky would seem to be the limit for Wall Street's top-performing megacap stock in 2023.

This figure should have Nvidia shareholders thinking twice about their investment

While most investors have been jaw-dropped by Nvidia's stellar growth rate, its rapidly expanding margins, and the company's phenomenal pricing power, there's one figure in the company's report that stands out as a potentially glaring problem -- and not a single person appears to be talking about it.

Metric Three Months Ended July 31, 2022 Three Months Ended July 30, 2023 Six Months Ended July 31, 2022 Six Months Ended July 30, 2023
Revenue $6,704 $13,507 $14,992 $20,699
Cost of revenue $3,789 $4,045 $6,646 $6,589

Data source: Nvidia fiscal second-quarter 2024 report. Figures are in millions of dollars.

What you see in the chart is a very simple breakdown of Nvidia's total sales and cost of revenue over the three- and six-month periods of fiscal 2024 and fiscal 2023. To most investors, this looks great. Sales are soaring, and margins are rocketing higher.

But note that cost of revenue has actually declined through the first six months of the current year. What this tells us is that Nvidia's data center sales growth is almost entirely driven by pricing power and relative GPU scarcity, not an increase in GPU volume.

I know what you're probably thinking: "Isn't strong pricing power a good thing?" The short answer is yes -- in the very short run. However, with competition from Advanced Micro Devices (AMD 2.37%) and Intel (INTC -9.20%) on the horizon, new high-compute GPUs entering the market will weigh heavily on Nvidia's pricing power.

AMD expects to begin shipping its MI300X AI-driven GPU to select customers by later this year, with shipment volume of its Nvidia-challenging GPU really ramping up into 2024.  Meanwhile, Intel intends to bring its Falcon Shores high-compute data center product to market by 2025. While Intel may be late to the party, Falcon Shores should still have plenty of ability to grab market share in AI-accelerated data centers.

Nvidia's gains on the pricing side also suggest that data center AI-GPU orders are being driven by FOMO (the fear of missing out).

Historically, every next-big-thing investment has gone through an initial boom-then-bust phase. This is a period where big numbers dance in investors' heads. Unfortunately, the uptake of every next-big-thing innovation takes time. Even AI companies will need time to mature and assess their needs. As new competition enters Nvidia's space, demand could quickly taper.

A person using a pin to pop a bubble containing a green money sign.

Image source: Getty Images.

Meeting otherworldly expectations could prove impossible

If there's one shining light for Nvidia bulls, it's that the world's largest chip fabrication company, Taiwan Semiconductor Manufacturing (TSM 1.26%), more commonly known as TSMC, is set to double its chip on wafer on substrate (CoWoS) capacity. CoWoS is effectively a mainstay in high-compute data centers due to its reasonably low cost and the need for high-bandwidth memory integration. By next year, Nvidia should be able to meaningfully expand its AI-GPU data center deliveries, which can drive sales even higher.

But even upping its own high-compute GPU output is liable to have a deleterious impact on its pricing power. With AMD's MI300X also set to become mainstream in 2024, we could soon witness a complete reset of AI-GPU pricing.

Nvidia has, thus far, been able to sustain an outsized move higher by blowing away Wall Street's sales expectations. However, history shows that outsized gains in megacap stocks of the magnitude that Nvidia has delivered in 2023 are rare.

With the company now commanding a multiple of 96 times trailing cash flow, it's effectively priced for perfection. Based on what I'm seeing from the company's cost of revenue, coupled with an inevitable uptick in competition, the volume simply isn't there to continually deliver the "perfect" reports that'll be needed to support the stock's phenomenal move higher.