Intel has introduced its latest Gaudi 3 AI chip, named , amid a frenzied race among chip manufacturers to develop semiconductors capable of efficiently training and deploying large AI models.

The Gaudi 3 chip boasts remarkable efficiency, claiming to be over twice as power-efficient as Nvidia’s H100 GPU while achieving AI model execution speeds that are one-and-a-half times faster. Offering versatility, it is available in various configurations, including a package of eight Gaudi 3 chips on a single motherboard or a card designed to integrate seamlessly into existing systems.

During testing, Intel evaluated the Gaudi 3 chip using models like Meta’s Llama and the Falcon backed by Abu Dhabi. Results indicated its suitability for both training and deployment purposes across a spectrum of AI models, encompassing applications such as Stable Diffusion and OpenAI’s Whisper model for speech recognition.

One significant selling point emphasized by Intel is the reduced power consumption of its chips compared to Nvidia’s offerings. Despite Nvidia’s dominant position, commanding an estimated 80% share of the AI chip market with its GPUs, Intel remains confident in its ability to compete. The Gaudi 3 chips are slated for availability to customers in the third quarter, with major companies like Dell, HP, and Supermicro committed to integrating them into their systems. However, specific pricing details for the Gaudi 3 chip were not disclosed.

Intel’s Vice President of Xeon Software, Das Kamhout, expressed confidence in the competitiveness of their new chips, citing factors such as pricing, an integrated network on chip, and the utilization of industry-standard Ethernet.

As the demand for AI capabilities in data centers continues to surge, there is ample opportunity for competitors to challenge Nvidia’s dominance. Cloud providers and businesses are investing in infrastructure to support AI software deployment, driving market growth and creating space for alternative chip suppliers.

Cost considerations are paramount for companies engaged in AI initiatives, particularly with the expense associated with running generative AI and purchasing Nvidia GPUs. Consequently, there is a growing interest in diversifying suppliers to mitigate costs.

While Nvidia has experienced significant stock growth attributed to the AI boom, Intel’s stock performance has been comparatively modest. Nonetheless, Intel remains optimistic about its position in the AI chip market and is actively expanding its offerings.

AMD, another major player in the semiconductor industry, is also seeking to capitalize on the burgeoning demand for AI chips. With notable customers like Meta and Microsoft, AMD has introduced its data center GPU, the MI300X, and is poised to compete with Nvidia and Intel.

In response to Nvidia’s proprietary software advantage through CUDA, Intel is collaborating with industry giants such as Google, Qualcomm, and Arm to develop open software solutions. This collaborative effort aims to offer flexibility to software companies and reduce dependency on proprietary technologies.

Furthermore, the construction of Gaudi 3 on a five-nanometer process suggests Intel’s engagement with external foundries for chip manufacturing. In addition to designing advanced AI chips like Gaudi 3, Intel plans to leverage its manufacturing capabilities to produce AI chips for external partners, as outlined by CEO Patrick Gelsinger’s recent statements regarding a forthcoming Ohio factory.