If there is one stock that exemplifies the potential to profit from artificial intelligence (AI), it would have to be Nvidia (NVDA -0.32%). The company pioneered the graphics processing unit (GPU) that rendered lifelike images in video games. What makes these chips so effective is parallel processing, which takes computationally complex jobs and breaks them up into smaller, more manageable chunks. Nvidia was able to adapt this technology to handle the rigors of AI, making it arguably the most successful company in the AI chip space.
However, as demand for AI has accelerated, numerous technology stalwarts have been working furiously behind the scenes to come up with viable alternatives to Nvidia's industry-leading processors. This week saw a number of announcements as rivals unveiled the fruit of these efforts.
Let's take a look at three.
Image source: Getty Images.
1. Intel
Earlier this week, Intel (INTC -1.78%) unveiled its latest AI chip, dubbed Gaudi 3, in a direct challenge to Nvidia's GPU dominance. The company claims that Gaudi can train AI models nearly twice as fast as Nvidia's H100 flagship data center processor and 50% faster for inference -- running AI models that have already been trained. The company also said Gaudi delivered all that performance with 40% better power efficiency than the H100 "at a fraction of the cost," according to Intel.
It's important to note that Nvidia recently unveiled the successor to its H100, its Blackwell B200 AI GPU, which will begin shipping later this year. While Gaudi wasn't compared to the B200, the company said, "We do expect it to be highly competitive" against Nvidia's latest processor, calling Gaudi "a strong offering."
Intel is likely not looking to outperform Nvidia's most recent AI chips but rather offer enterprises a cheaper alternative. It's also worth noting that there's been an ongoing shortage of Nvidia's H100 and H200 chips, creating a bottleneck in AI adoption. Intel's latest offering could help alleviate the backlog.
2. Meta Platforms
Not to be outdone, Meta Platforms (META -2.83%) announced the Meta Training and Inference Accelerator (MTIA) v2, a custom AI chip that runs in the company's data centers. Meta announced that the latest version of its MTIA "more than doubles the compute and memory bandwidth" of its predecessor. Early results suggest the new processor performs three times better than the first version.
Meta noted that it's trying to provide "the right balance of compute, memory bandwidth, and memory capacity" with its latest chip. The company isn't using these chips for AI processing; at least, not yet. Meta is using the processors to improve the speed and efficiency of its ranking and recommendation systems, used to surface relevant content on its social media sites and target advertising.
Meta is one of Nvidia's biggest customers, so if it is successful in its AI chip efforts, the company could reduce its dependency on Nvidia's AI processors.
3. Alphabet
At the Google Cloud Next event this week, Alphabet (GOOGL -3.37%) (GOOG -3.30%) announced the availability of Cloud Tensor Processing Unit (TPU) v5p, the latest version of its AI-centric processor. The company said its Cloud TPU can train large language models (LLMs) -- which underpin AI systems -- three times faster than version 4.
As is generally the case, Google's TPUs won't be sold but will be available on Google Cloud later this year.
However, in a blog post, Google noted that it's expanding its Nvidia GPU capabilities with the addition of the A3 Mega Supercomputer, which is powered by Nvidia H100 GPUs. Google Cloud customers will also have access to virtual machines (VM) featuring Nvidia's Blackwell B200 and Grace Blackwell GB200 GPUs.
Google also unveiled Axion, its first data center central processing unit (CPU), designed in conjunction with Arm Holdings. Google says Axion is 30% faster than the leading CPU, with 50% better performance and 62% greater efficiency than x86 chips from Intel and Advanced Micro Devices (NASDAQ: AMD).
An important takeaway for Nvidia investors
While Intel intends to compete directly with Nvidia for a piece of the AI chip market, Meta Platforms and Alphabet are developing their own AI chips for specific purposes, either in-house or for cloud customers. Furthermore, while Intel claims its AI-centric chips can compete head-to-head with Nvidia's flagship processors, Nvidia has already moved on and will soon be shipping the next version of its state-of-the-art AI processors.
This helps highlight Nvidia's enviable position in the industry -- its rivals are always playing catch-up. They claim to best Nvidia in terms of performance, but their bragging rights don't last, as Nvidia's next-generation chips are already on the way.
It's also important to point out that the market for generative AI is growing at a rapid clip, increasing the market for these cutting-edge processors. As a result, Nvidia will likely maintain its industry-leading position, even as rivals enter the fray.
