Lambda Labs’s Mitesh Agrawal has left to lead the AI hardware startup Positron.lambda, focuses on the development of cloud infrastructure and has been evaluated over $ 2 billion.
Lambda Labs, a NVIDIA partner, has lost the highest executive officer in hardware construction hardware for a well -known company for the AI industry.
MITESH AGRAWAL told Business Insider that it played a new role as a Positron CEO this month. Positron builds hardware for inference for transformer models. This is a way for chatbots like Chatgpt to respond to user demands.
The departure of Agrawar is important in consideration of his role, which forms Lambda into one of Silicon Valley’s most funded startups.
In the Series C round last February, the company was evaluated as about $ 1.5 billion. Agrawal refused to share the accurate evaluation of the company, but has grown to more than $ 2 billion.
Agrawal told BI that when he joined Lambda in 2017, the company focused on the construction of a machine for image generation models. Five years after the twin brothers Steven and Michael Baraban have been established as a company developing facial recognition technology. Shortly after Agrawal arrived, the company has shifted the focus, designed infrastructure for full -scale data centers, and pivoted to cloud services.
He stated that LAMBDA’s business is currently focusing on the development of cloud infrastructure and focusing on renting a server equipped with a NVIDIA graphic processing unit. It also provides software that includes APIs for inference and mechanical learning libraries for customers.
Agrawal stated that his movement to Positron has been in the ability to apply for reasoning and the AI model has been applying training to new data.
“Technology for inference is increasing,” between chatbot, such as Chatgpt and XAI’s GROK and O1 of OPENAI, “Agrawal is increasing.” He added that he is thinking about many things and how to solve these models in as much efficiency as possible and how to do it.
He believes that Positron is suitable for taking on the challenge.
Positron was founded in 2023 by Thomas Somers, which Agrawar met in 2015. The two overlapped with lambda during Somers states in 2020 and 2021. The simplest conditions were “building hardware to compete with NVIDIA.”
Positron says that its hardware exceeds NVIDIA’s H100 and H200 GPUs, which promoted AI races before releasing more powerful black welty chips at performance, power, and reasonable prices.
Facing a giant like NVIDIA, which overtains Apple as the most valuable company in the world last week, is not an easy job for a promising company. However, SOHMERS stated that by providing hardware to the reasoning of the transformer model, Positron can be distinguished from competitors.
Trans Model -Learning the context and meaning of data and the new data is behind several of the most popular AI applications. Compared to folding neural networks, which have supported the progress of machine learning over the past few decades, the need for memory models has been enhanced. Somers said he had seen the opportunity to use these demands.
“I think we think there is a better way to do things because we started Positron,” said Somers. “As a large company that focuses on many other products, NVIDIA really optimizes a specific niche that we focus on, a transformer model, and did not focus. “
Agrawal is also confident in Positron’s hardware performance and energy efficiency. He stated that compatibility with various transformer models would help attract customers from competitors.
“NVIDIA has a very powerful ecosystem in the AI model world. I’ve heard about their CUDA MOAT and have heard about software moats,” he said.
“What Positron did was to completely remove this friction,” Agrawar said. He added that a company adopted a trained model on the NVIDIA GPU, which means “executing the model with a Positron card, just as running on the NVIDIA GPU.”
Agrawar said that a jump to young startups, such as a established player like Lambda, has presented an exciting challenge.
“You can compete with veterans in the industry and in a very large field,” he said.
