Nvidia (NVDA +1.23%) made a name for itself by pioneering the graphics processing units (GPUs) that became the gold standard for rendering images in video games. The company adapted these chips to handle the rigors of artificial intelligence (AI), a field it now dominates. Unfortunately, being the leader means there’s always someone trying to take you down.
Alphabet’s (GOOGL +2.12%) (GOOG +2.20%) Google just unveiled two powerful new AI chips, the latest move in the company’s efforts to become a greater force in AI.
Image source: The Motley Fool.
A chip off the old block
In a blog post released on Wednesday, Amin Vahdat, senior VP and chief technologist for AI and infrastructure, unveiled the eighth generation of Google’s tensor processing unit (TPU), the company’s custom-built silicon for AI. In a move that surprised market watchers, he announced two distinct architectures — one for AI training and the other for AI inference.
“With the rise of AI agents, we determined the community would benefit from chips individually specialized to the needs of training and serving,” Vahdat wrote. To that end, Google introduced the TPU 8t and the TPU 8i. Not surprisingly, the TPU 8t was designed for “compute-intensive training workloads.” At the same time, the TPU 8i comes equipped with more memory to reduce the inherent latency (lag) in interactions between AI agents. Perhaps most importantly, Google notes that “specialization unlocks significant efficiencies and gains.”
Specifically, TPU 8t was built to reduce the time it takes to develop frontier models from months to weeks. It accomplishes this herculean task by delivering 3 times the compute performance, 10 times faster storage access, and double the chip data transfer rate than its predecessor.
Today’s Change
(2.12%) $7.03
Current Price
$339.32
Key Data Points
Market Cap
$4.1T
Day’s Range
$335.16 – $339.81
52wk Range
$147.84 – $349.00
Volume
527K
Avg Vol
32M
Gross Margin
59.68%
Dividend Yield
0.25%
On the other hand, the TPU 8i was built with inference in mind — the job an AI model does after it’s been trained. The age of AI agents will involve complex multi-step tasks, at times requiring multiple agents to collaborate to complete them more quickly.
The processor combines high-bandwidth memory (HBM) with 3 times the amount of static random-access memory (SRAM), reducing the lag caused by data transfers between chips. The TPU 8i also benefits from the addition of Google’s custom Arm-based Axiom CPU, which has been customized for the job.
Google say these innovations “deliver 80% better performance-per-dollar” compared to the previous generation, giving customers twice the volume for the same cost.
Will TPUs “chip” away at Nvidia’s lead?
It’s important to note that Alphabet remains heavily reliant on Nvidia’s GPUs and is among the chipmaker’s largest customers. In fact, Google introduced the first version of its TPU in 2016, and these custom chips have since become a cornerstone of the company’s cloud and AI strategy — while still depending on Nvidia for the heavy lifting.
Today’s Change
(1.23%) $2.46
Current Price
$202.34
Key Data Points
Market Cap
$4.9T
Day’s Range
$199.35 – $202.48
52wk Range
$103.11 – $212.19
Volume
3.1M
Avg Vol
174M
Gross Margin
71.07%
Dividend Yield
0.02%
These TPUs mark the latest in a long line of processors developed by competitors to decrease their reliance on Nvidia’s GPUS. Despite rising competition, Nvidia is estimated to control 92% of the data center GPU market, according to IoT Analytics. While Google’s TPUs may offer cost benefits for AI workloads, Nvidia’s GPUs still provide the greatest computational horsepower, powering both AI training and inference.
From a practical standpoint, these TPUs give Alphabet flexibility. The company can offer more affordable AI processing options for price-sensitive cloud customers, while also reducing its own energy consumption, thereby lowering costs and boosting profits. Moreover, at 31 times earnings, Alphabet stock is attractively priced compared to a multiple of 41 for Nvidia.
To be clear, I am a firm believer in Nvidia and Alphabet, seeing both as leaders in the AI boom — which is why I own both stocks.