Cerebras & Perplexity Disrupt $100B Search Market
Cerebras Systems and Perplexity AI have joined forces in a groundbreaking partnership aimed at disrupting traditional search engines. Their collaboration introduces Sonar, an AI-powered search model that operates at unprecedented speeds, marking a significant step toward AI-first search experiences.
Redefining AI Search with Speed
Sonar, Perplexity AI’s latest model, leverages Cerebras’s specialized AI chips to achieve an astounding 1,200 tokens per second. Built on Meta’s Llama 3.3 70B foundation, this model represents a major shift in how AI search engines operate—prioritizing speed without sacrificing accuracy. According to Perplexity’s internal testing, Sonar outperforms major competitors like GPT-4o mini and Claude 3.5 Haiku in user satisfaction metrics.
Denis Yarats, Perplexity’s CTO, emphasized the importance of the collaboration:
“Cerebras’s cutting-edge AI inference infrastructure has enabled us to achieve unprecedented speeds and efficiency.”
The Rise of Specialized Hardware in AI
This partnership underscores an emerging trend: AI companies are increasingly turning to specialized hardware to gain a competitive edge. Cerebras recently showcased its DeepSeek implementation, proving to be 57 times faster than conventional GPU-based solutions. By making AI inference dramatically faster, Cerebras is positioning itself as the preferred provider for companies looking to optimize AI-powered applications.
Andrew Feldman, CEO of Cerebras, argues that advancements in AI hardware don’t shrink the market but expand it:
“Every time compute has been made less expensive, they [investors] have assumed that made the market smaller. And in every single instance, over 50 years, it’s made the market bigger.”
Can Perplexity Challenge Big Tech in Search?
Perplexity’s push into AI search comes at a time when traditional search engines are struggling to integrate AI seamlessly. By offering near-instantaneous responses with high factual accuracy—Sonar scores 85.1 out of 100 in factuality tests, outperforming GPT-4o and Claude 3.5 Sonnet—Perplexity aims to attract enterprise users who value both speed and reliability.
However, questions remain about scalability and cost. While Cerebras’s chips offer superior performance, their cost-effectiveness compared to GPUs is yet to be determined. The challenge will be convincing large-scale enterprise customers that the benefits justify the investment.
What's Next
Perplexity plans to roll out Sonar to its Pro users first, with broader availability expected soon. As AI search evolves, speed and efficiency could become the primary differentiators, rather than just model size. This partnership signals a shift toward AI-first search experiences, where traditional search engines may need to adapt—or risk being left behind.
