Nvidia's (NVDA 1.06%) stock has nearly tripled since the beginning of 2023 and is currently hovering near its all-time high. The chipmaker's stock caught fire as the explosive growth of the artificial intelligence (AI) market -- driven by the rise of "generative AI" platforms like ChatGPT and DALL-E -- boosted its sales of high-end data center graphics processing units (GPUs).
The bullish thesis for Nvidia is easy to understand. Nvidia's top-tier GPUs are used to accelerate the world's most complex machine learning and AI tasks, so their sales will continue to rise as those markets evolve and expand.
The global AI market could still grow at a compound annual growth rate (CAGR) of 37% from 2023 to 2030, according to Grand View Research, so Nvidia's data center business (which brought in 60% of its revenues in its latest quarter) could still have plenty of room to run. Nvidia also remains the world's largest producer of high-end gaming GPUs -- and that market should continue to expand as more graphically demanding video games hit the market.
Image source: Nvidia.
Analysts expect Nvidia's revenue and adjusted earnings to grow 59% and 133%, respectively, in fiscal 2024 (which ends next January). Those growth rates are stellar for a $1 trillion company that already generated $27 billion in revenues in fiscal 2023.
All those strengths made it tempting to hop aboard the bullish bandwagon. But before pulling the trigger, shrewd investors should carefully review these four red flags for Nvidia's future -- and why they indicate its stock could be getting overheated at these record levels.
1. Its high valuations
At $420 per share, Nvidia trades at 54 times this year's adjusted earnings. Its enterprise value of $1.04 trillion is also 24 times higher than this year's estimated revenues. These valuations are very high relative to its industry peers.
AMD, Nvidia's main competitor in discrete GPUs, is growing at a slower rate but trades at 40 times forward earnings and eight times this year's sales. Intel, which is trying to challenge Nvidia and AMD with its own discrete GPUs, trades at 114 times forward earnings (due to a recent drop in its profits related to the expansion of its foundries), but just three times this year's sales. In other words, a lot of growth has already been priced into Nvidia's stock at these levels. Those valuations could be popped if it falls short of the market's high expectations.
2. Its insiders are selling
If Nvidia's insiders expected it to double in value again and become a $2 trillion company in the near future, they would probably be eagerly scooping up more shares. But over the past three months, they sold about 93 times as many shares as they bought. Over the past 12 months, they sold more than four times as many shares as they bought. All those sales suggest that Nvidia's valuations are stretched and its near-term upside is limited.
3. A more efficient AI processing method might exist
Nvidia's data center GPUs are considered the industry-standard chips for accelerating AI tasks. But for many years, a British start-up called Graphcore has repeatedly claimed it could beat Nvidia in the AI market. To understand why, we should review the key differences between Intel's and AMD's central processing units (CPUs), Nvidia's GPUs, and Graphcore's intelligence processing units (IPUs).
CPUs process a single piece of data at a time through scalar processing, while GPUs process a wide range of integers and floating point numbers simultaneously via vector processing. That's why Nvidia's GPUs are better equipped to handle complex AI tasks than Intel's CPUs. Graphcore's IPUs, however, simultaneously crunch all of the data mapped out across a single graph. It claims this "graph processing" method is faster and more cost-efficient than scalar and vector processing.
Graphcore is still tiny compared to Nvidia, but it recently got a big vote of confidence when Meta Platforms poached Graphcore's entire AI chipmaking team to accelerate the development of its own in-house chips. If Meta develops its own graph-processing IPUs for its AI tasks, it could potentially stop using Nvidia's GPUs altogether and show other large tech companies that they can develop their own superior chips.
4. Google also thinks it can make better chips
Alphabet's Google is another tech giant which believes it can make faster and more power-efficient chips than Nvidia. In April, Google published a research paper suggesting its fourth-generation TPUs (Tensor Processing Units) -- which are used to train its own AI models -- were 1.2 to 1.7 times faster than Nvidia's A100 chips and up to 1.9 times more power efficient.
Google didn't directly compare its latest TPUs to Nvidia's latest H100 chips, but these head-to-head tests suggest the tech giant wants to develop its own AI chips to sharpen its competitive edge and reduce its dependence on third-party chipmakers.
Do these red flags make Nvidia a risky investment?
Of these four red flags, I'd pay the most attention to Nvidia's frothy valuations and insider sales than Meta and Alphabet's development of first-party chips. Nvidia still has a bright future, but you shouldn't pay the wrong price for the right company -- so investors should carefully separate the hype from the facts before they pull the trigger.
