When you buy through affiliate links in our content, we may earn a commission at no extra cost to you. Learn 
AI ‘superchips’ are at the heart of the artificial intelligence (AI) revolution. As Nvidia continues to dominate the global chip sector, and analysts assure everyone else is “just playing catch up”, Intel challenges the market with its latest advanced chip: the Intel Gaudi 3 AI Accelerator.
Intel claims the Gaudi 3 is 1.5 faster when training AI and 1.5 faster when running AI processes than the Nvidia GPU NV 100. This would put the company in the top ranking of AI chip performance.
Is Intel offering advanced technology and something new to the market? Is the AI chip war just starting to unfold? Did Intel design the chip to easily integrate with AI systems and infrastructure? Techopedia talks with experts to answer these and other questions as we dive into the new Intel technology.
The Launch of the Gaudi 3 AI Accelerator Chip
On April 9, during the Vision 2024 event, Intel announced the new Gaudi 3 AI chip along with a new AI strategy.
The company said the new Gaudi AI chip delivers 50% on average better inference and 40% on average better power efficiency than the Nvidia H100. While the official price has not been announced for this new Intel AI chip, the company says it will be available at a fraction of the cost of the Nvidia H100 — which costs over $40,000.
‘The AI Market is About to Change’
Kevin Surace, known as the “Father” of the virtual assistant and voice user interface, TED and TEDx speaker, and Chair of Token — an access security technology company — told Techopedia that the AI market is about to change.
“The market is gasping for alternatives to the $40,000 H100 from Nvidia. Nvidia has had a solid lead and earned a monopoly on GPUs for AI. That is about to change big time, and Intel’s offering is surprisingly strong.”
Michal Oglodek, CTO & co-founder of Ivy.ai — a leading provider of conversational, generative AI chatbots for higher education, health care, and government — agreed and said that challenging Nvidia’s AI dominion of the market was inevitable.
“It was only a matter of time before other powerhouses began to ‘chip’ away at Nvidia’s AI market share, so it’s hardly surprising that Intel made its move. The competition is now fully underway.”
A look at the new Intel Gaudi 3 chip. (Intel)
Oglodek explained that the AI explosion has propelled Nvidia’s stock to more than triple its value in the past year, attracting the attention of nearly every major company. “With the field becoming increasingly crowded, expect a surge of contenders as AI solutions and providers proliferate rapidly,” Oglodek added.
“While slow and steady (typically) wins the race, results from Nvidia (and soon Intel) show that there is little time to lose. The frontrunners are tackling the AI frenzy from all sides.”
Intel Seduces Original Equipment Manufacturer Industry
Intel also announced that the Gaudi 3 is now available for original equipment manufacturers (OEMs) including brands like Dell, Hewlett Packard Enterprise, Lenovo, and Supermicro.
Many of these leading global tech brands have released products compatible with, or powered by, Nvidia technology in the past year, despite having a strong commercial relationship with Intel that dates back many years.
With the availability of the Gaudi 3 chip, Intel attracts the attention of OEMs by offering new AI hardware and services that are more “powerful”, more “power efficient”, and more “cost-effective” than Nvidia’s portfolio.
The fact sheet of the Intel Gaudi 3. (Intel)
IBM, Bosch, NAVER, NielsenIQ Join Intel’s Gaudi 3 Front Line
Intel announced that big names in tech have already partnered up with the company to access the Gaudi 3 AI accelerator technology. These include Bosch, IBM, Landing AI, Ola, NAVER, NielsenIQ, Roboflow, Seekr, and others.
Intel also announced an open platform for enterprise AI, working with SAP, RedHat, VMware, and other industry leaders. The platform will focus on accelerating the deployment of secure generative AI (GenAI) systems, enabled by retrieval-augmented generation (RAG).
Pat Gelsinger, CEO of Intel spoke confidently about the company’s move into the AI market to the press during the release of the new AI chip.
“Intel is bringing AI everywhere across the enterprise, from the PC to the data center to the edge. Our latest Gaudi, Xeon, and Core Ultra platforms are delivering a cohesive set of flexible solutions tailored to meet the changing needs of our customers and partners and capitalize on the immense opportunities ahead.”
Intel’s AI Market Demands Analysis on Point
During the launch of the Gaudi 3 AI chip, Intel highlighted a December 2023 study by cnvrg.io — an Intel owned-company that helps organizations deploy GenAI and large language models.
The cnvrg.io 2023 ML Insider survey found that only 10% of companies have launched GenAI solutions in 2023, despite the AI hype. Three-quarters of those surveyed said their organizations have yet to deploy GenAI models to production.
Intel learned through this study that compliance and privacy (28%), reliability (23%), high cost of implementation (19%), and a lack of technical skills (17%) are the biggest roadblocks for companies implementing large language models (LLMs) into their businesses.
Surace from Token explained how Intel with its new Gaudia 3 Chip efficiently reads the current AI market demands and delivers accordingly.
“Model creators and users need to have faster processing for AI, lower energy usage, and lower costs. Intel’s offering outperforms Nvidia in major model tests at a lower price point.”
“This is the start of a number of new tensor and GPU offerings that will challenge Nvidia’s dominance.”
With the new Gaudi 3, an AI platform, and new services, Intel now supports GenAI, from pilot to production. Intel also highlighted that the Gaudi 3 chip will have no availability issues and is designed to address complexity, fragmentation, data security, and compliance requirements.
Did Intel Check All the Right AI Boxes?
Intel’s presentation of the Gaudi 3 AI chip is almost utopian. At a fraction of what a Nvidia chip costs, what company or developer would not want a more energy-efficient and powerful chip? But did Intel check all the right boxes?
Rodolfo Rosini, co-founder and CEO of Seattle and London-based Vaire, a company focusing on using reversible computing to create near zero-energy silicon chips, told Techopedia that the chip market no longer plays with the same rules that existed in the early computer days.
“Intel’s hardware is not bad, but the true battlefield is more than that.
“The market for AI PCs might take a while to materialize, and right now it’s still unclear from a use case perspective which cohort of users will be the early adopters.
“It’s not like in the early days of personal computers where business users needed Excel, or the early days of data centers when developers wanted to use the relatively cheap Intel chips (compared to what was running on Unix) and install Linux.
“The market moved to be more vertically integrated, with more alignment between hyperscalers, AI research labs, and chip companies.”
During the AI chip release, Intel outlined its strategy to help organizations deploy scalable AI systems, including hardware, software, frameworks, and tools. Intel also reassured its customers that its AI strategy and approach enables a broad, open ecosystem of AI players to offer solutions that satisfy enterprise-specific GenAI needs, from equipment manufacturers to database providers, systems integrators, software, and service providers.
However, Rosini from Vaire said that Intel should be more aggressive with mergers and acquisitions and acquire AI labs. “There are several candidates like Stability AI, Sakana, or Mistral,” Rosini added.
“AMD famously spun out their fabs to finance the acquisition of ATI to enter the GPU market. Possibly Intel needs to copy that strategy to go all in on AI.”
Intel’s AI, Open Ethernet, Compatibility, and Open-Source Approach
Another area where Intel draws a hard line in the sand is compatibility. Intel’s new AI Gaudi chip is not compatible with Nvidia GPUs. Surace from Token broke down the Gaudi 3 compatibility issue.
“Intel Gaudi does not run CUDA as that is proprietary to Nvidia. However, they are using Ethernet as the network protocol instead of Nvidia’s proprietary networking and have formed an open alliance with Google Qualcomm Arm, and others to have an open-source offering that competes with or replaces CUDA.”
Integration and Compatibility: Open Ethernet
Intel announced the Ultra Ethernet Consortium (UEC) to drive integration and compatibility. Through UED, Intel is leading the way in using open Ethernet networking for AI fabrics. The company introduced an array of AI-optimised Ethernet solutions, including the AI NIC (network interface card) and AI connectivity chipsets.
The Ultra Ethernet Consortium (UEC) breaks away from traditional AI models and operations.
Instead of proprietary solutions, Intel is now promoting the use of ethernet for communication between AI chips and other devices. This would make it easier for companies to integrate Gaudi 3 into existing systems and scale with ease.
Through UEC, Intel fosters collaboration to ensure interoperability between different ethernet-based AI components and take bigger bites into the market.
Intel’s tools like AI NICs (network interface cards) and AI connectivity chiplets are specifically designed for efficient data transfer within AI systems using ethernet. The benefits of using ethernet for data transfer needs are straightforward. Ethernet offers a good balance of speed and reliability. However, the performance, privacy, and security of AI high-speed heavy data transfers through ethernet still require research and testing on a case-to-case basis.
However, by straying from closed proprietary solutions that build custom communication protocols for specific needs and moving towards open ethernet, Intel seeks to attract more customers who are leaning towards more open technologies.
The Bottom Line
While Intel’s Gaudi 3 chip boasts impressive capabilities and open communication strategies for easier integration, its true disruptive potential of Gaudi 3 hinges on its performance, cost, and adoption within the AI community.
Intel has managed to remain not only relevant but of value by introducing a serious competition challenge to Nvidia. Additionally, Intel’s AI chip market move and AI strategy are just beginning. Future AI products and services should be expected.
Is Intel’s Gaudi 3 AI chip a market disruption tech, or is it too little, too late? Only time will tell, but two things are certain. Intel’s history of innovation and business reach shouldn’t be underestimated, and like Surace from Token told Techopedia, “It’s too early in the AI chip game — first inning.”
