Nvidia Rival Etched Hits $5 Billion Valuation After Securing $1 Billion in AI Chip Sales

AI semiconductor startup Etched has been valued at $5 billion after raising more than $800 million in capital and securing more than $1 billion in customer sales commitments, making it one of the fastest-emerging competitors to chipmaking giant Nvidia in the red-hot AI hardware sector. The startup, which aims to build hardware that can power a range of artificial intelligence applications with high performance, has gained an enormous amount of traction in recent months amid massive investor enthusiasm for the AI industry. Etched has attracted nearly $850 million in total funding so far and secured commitments for over $1 billion in sales from customer clients, with its latest valuation catapulting it into unicorn status and solidifying its position as a potential leader in the rapidly growing inference segment of the AI chip market.

Unlike many AI chip startups, which are looking to replicate Nvidia's success in general-purpose graphics processing units (GPUs) used in the training of artificial intelligence models, Etched is focusing its efforts on developing highly specialized processors, custom silicon for performing trained AI models, better-optimized inference.

"AI inferencing the part of artificial intelligence that occurs when a model answers a query instead of just processing training data has quickly become one of the biggest revenue opportunities in AI," Etched cofounder and CEO Vinod Bakolia told VentureBeat. "We designed our processors with specific tasks for AI inference in mind. Unlike traditional chips that are general purpose, our architecture is fine-tuned to efficiently process a variety of AI models." Nvidia’s own high-end GPUs dominate not just the AI training chip space, but are also used for inference due to their immense processing power and programmability, despite their high cost and significant power consumption.

Many industry observers believe the inference market is ripe for competition as more enterprises look to deploy large-scale AI inference.

That’s due to the explosion of chatbots, AI assistants, and enterprise applications, each of which requires inference to answer questions and respond to prompts. “The demand for the chip, or the processing unit, will be huge going forward as more people move towards using more and more artificial intelligence tools and technology to improve the ways in which they work,” said industry analyst Mike Brown.

“InFERENCE has really taken off.”

Etched’s chips can run a variety of open-source large language models like Llama, Qwen, DeepSeek, and Mamba, the company says, indicating that its chips are well-positioned for common commercial applications in this space. The company plans to ship its first products later this year. Its growth is a strong indicator of the investor appetite for AI infrastructure: just about all of Etched’s recent sales commitments have come from companies either just beginning to explore their AI roadmaps or in the process of rolling out the initial applications of their AI endeavors.

“Our go-to-market strategy focused initially on those forward-thinking early adopters that are already developing and implementing the next generation of commercial AI applications, and the response has been outstanding,” Bakolia said.

“It indicates a widespread demand from enterprise for the kind of high-performance, cost-efficient, and efficient inference solutions Etched is providing.” The company is currently fabricating chips through partners in Taiwan and continues to ramp up its engineering and testing facilities in California to handle the increasing demand for its inference hardware, the company said. The rise of Etched marks another sign that the AI chip industry is becoming more competitive as companies scramble to address both the need for training massive models and deploying them for practical applications. While Nvidia has enjoyed immense success in providing GPUs for AI training, it’s been slower to gain dominance in the inference market, where the specialized processing requirements of deployed models are different from those for model training, making the opportunity ripe for competition and specialized hardware approaches.

It’s this demand that Etched is banking on.

“We don’t want to be everything to everyone. We’re going after that huge and underserved market,” Bakolia said. “There’s massive opportunity ahead.

Our mission is to accelerate the adoption of AI.