Intel doesn’t want to fight NVIDIA. The plan is different

The current leader in the artificial intelligence market is NVIDIA. Their AI accelerators are the most efficient solutions available, and they are used by giants like Google, Meta, Oracle Whether OpenAI. Thus, the company managed by Jensen Huang is the second highest valued in the world, losing only to Apple.

Intel says we will be moving away from large models

Of course, everyone is trying to get something out of it a market worth hundreds of billions of dollars. The most frequently mentioned companies include: AMD, Intel and Huawei. However, the Blues revealed that they have a different strategy than the competition, and their latest arrangements are proof of this.

The pure performance race with NVIDIA is extremely difficult. Both in terms of the Greens’ many years of experience and their R&D budget. That’s why Intel wants to focus on offering the best possible price-to-computing power ratioand thus target smaller companies.

These are not just empty words. Accelerator Intel Gaudi 3which debuted on the market at the end of September, compared to NVIDIA H100 systems delivers 80% better performance-per-dollar. And in tests based on the Llama-2 model, this difference is even twofold.

Americans want to target startups and individuals. Additionally, Intel claims that in the long run we will have to move away from large models and move to smaller, specialized ones. Currently popular, huge data centers have serious problems when it comes to the cost of building and powering them. The latter in particular becomes clearly visible when companies like Oracle and Google turn to nuclear reactors.

Intel Gaudi 3 accelerators have so far been warmly welcomed by the industry and are already used by… IBM Cloud, Hewlett Packard Enterprise and in the company’s data centers Dell. Of course, the focus is still on the newest family NVIDIA Blackwellwhich has already sold out for the next 12 months.