Track market-moving stories before they get noisy

Nvidia (NVDA 0.38%) has been one of the biggest winners from the artificial intelligence (AI) infrastructure build-out. The stock has advanced more than 1,300% since January 2023. But most Wall Street analysts still believe Nvidia is deeply undervalued.

In fact, the consensus target price has increased from $265 per share to $295 per share in the last 90 days, according to LSEG. That implies 42% upside from the current share price of $209.

Here's what investors need to know.

Image source: Getty Images.

Nvidia is gaining market share in AI inference workloads Nvidia graphics processing units (GPUs) are the industry standard in artificial intelligence (AI) accelerators, chips that assist CPUs by handling repetitive mathematical tasks. Nvidia accounts for more than 80% of AI accelerator sales, but some analysts expected the company to lose significant market share as the industry shifted toward inference.

To elaborate, AI training is a discrete event in which models learn to perform certain tasks, but AI inference is a continuous process wherein models are used to generate outputs. Inference accounts for about two-thirds of AI workloads today, up from about one-third in 2023, and the shift will only intensify in the future as more models are deployed.

Companies like Alphabet and Amazon have designed custom AI accelerators in an effort to reduce their dependence on Nvidia GPUs. In certain scenarios, those custom chips are actually more efficient, but Nvidia's inference market share still increased eight percentage points to 74% over the past year, according to The Information.

Why? GPUs are general-purpose accelerators, while custom chips are designed for specific workloads. That makes them very efficient in certain situations, but it also means they are much less flexible (i.e., they run fewer algorithms). Venture Beat explains, "If a new AI technique is invented tomorrow, a GPU will run it immediately." That is not necessarily true for custom AI accelerators.

Beyond that, Nvidia has a competitive advantage in its vertically integrated business. The company not only designs GPUs but also CPUs, networking, and software that together form a turnkey solution for AI infrastructure. That translates into cost savings for customers. "Nvidia compute is not just the highest performance AI infrastructure, it is the most economic," says CEO Jensen Huang.

Nvidia is gaining market share in other categories of AI infrastructure While Nvidia is best known for its GPUs, the company is actually gaining share in other AI infrastructure categories. Networking revenue has at least doubled in each of the last three quarters, and it nearly tripled in the most recent quarter, because customers want tightly integrated systems. Nvidia recently became the largest networking company in the world.

Meanwhile, demand for Nvidia's next-generation Vera CPU is already immense ahead of its launch later this year. Vera is twice as efficient as x86-based alternatives (CPUs designed by AMD and Intel). CFO Colette Kress recently told analysts, "We have visibility to nearly $20 billion in total CPU revenue this year, setting us up to become the world-leading CPU supplier."

Today's Change

(

-0.38

%) $

-0.76

Current Price

$

199.28

AI infrastructure spending is projected to quadruple by the end of the decade To summarize, Nvidia is gaining share within the inference category of the AI accelerator market. That's important because inference has already surpassed training in terms of workload volume, and it will become an even larger part of the market in the future.

Meanwhile, Nvidia is also gaining share in networking equipment and CPUs as customers prioritize tightly integrated systems. Collectively, that puts the company in a good position. CEO Jensen Huang thinks AI infrastructure spending could hit $4 trillion annually by 2030, up from about $1 trillion today. Grand View Research has published similar numbers.

Here's the big picture: Multiple industry experts expect AI infrastructure spending to grow by 36% annually through the end of the decade. Nvidia is gaining share across multiple categories in that market, suggesting its earnings could grow even faster than 36% annually. That makes the current valuation of 32 times earnings look cheap. Patient investors should feel comfortable buying a small position today.