Mythic Raises $125 Million to Challenge Nvidia with 100x More Efficient AI Chips

AI Chip Startup Mythic Raises $125 Million in Bid to Take on Nvidia

Chip startup Mythic Inc. has raised $125 million in a new funding round, money that will help support its attempt to challenge Nvidia Corp. in the lucrative market for artificial intelligence processors. Mythic, which tapped Nvidia veteran Taner Ozcelik as its chief executive officer last year, raised the money from a group led by venture firm DCVC, according to a statement Wednesday. The investors also included New Enterprise Associates, Atreides Management, SoftBank Group Corp., Honda Motor Co. and Lockheed Martin Corp.

With $125 Mn haul, Mythic Wants to Take on NVIDIA GPUs with 100x Energy-Efficient AI Chips | AIM

Mythic plans to deploy its chips across data centres, automotive systems, robotics, and defence. Mythic, a Palo Alto-based AI chip company, has raised $125 million in a funding round to develop analog processing units designed to cut AI energy use by up to 100x compared with GPUs. The round was led by deep tech-focused venture capital firm DCVC and will support Mythic's product development, software, and commercial scale-up efforts. NEA, Atreides, Future Ventures, Softbank KR, S3 Ventures, Linse Capital, One Madison Group, and Catapult, along with Honda Motor and Lockheed Martin, also joined the round. The company said the raise follows a restructuring under chief executive officer Taner Ozcelik (former NVIDIA VP and GM), and focuses on addressing power constraints in AI computing. "Energy efficiency will define the future of AI computing everywhere," Ozcelik said in a statement. Mythic plans to deploy its chips across data centres, automotive systems, robotics, and defence. The 13-year-old company will also use the funding to rebuild its architecture, software stack, and strategy. The company also introduced Starlight, a sub-one-watt sensing platform that integrates its chips into image sensors. Mythic said the system improves signal extraction in low-light conditions and targets defence, automotive, and robotics use cases. Mythic's chips are manufactured in the United States and allied countries using standard semiconductor processes. The company plans to use the new capital to expand production, mature its software development kit, and pursue commercial deployments in AI inference markets. Mythic's chips use analog in-memory computing, which combines memory and processing in a single plane. The company said this design reduces energy loss during data movement, which it claims accounts for most of the power consumption in current AI systems. According to Mythic, its current architecture delivers 120 trillion operations per second per watt. Ozcelik said the company aims to complement GPUs rather than replace them. "Much as GPUs became the accelerated computer of choice next to CPUs, our APUs will become the accelerated computer of choice next to GPUs," he said. Mythic said its chips can run large language models with up to one trillion parameters without requiring high-speed interconnects used by GPU clusters. Internal benchmarks cited by the company show higher tokens per second per watt compared with current high-end GPUs. Aaron Jacobson, partner at NEA, said the platform "collapses today's limits on energy and cost" and gives the company scope to scale. Steve Jurvetson of Future Ventures said Mythic's approach unifies computation and memory "as in the brain," improving efficiency.

Compute-in-memory chip startup Mythic raises $125M round - SiliconANGLE

The venture capital firm was joined by NEA, Softbank KR, Honda Motor Co. and a long list of other backers. The investment brings Mythic's total outside funding to more than $175 million. Standard processors represent information as a series of individual electrical signals that each correspond to 1 or 0. Information processed in this manner is known as digital data. Mythic has developed a so-called analog processing unit, or APU, that takes a different approach. It represents data as fluctuations in a single, continuous electrical signal rather than as multiple individual signals. Analog chips are mainly used for tasks such as distributing power to a computer's components and filtering Wi-Fi inference. Mythic's APU, in contrast, is designed to run artificial intelligence models. The company claims that its chip can run AI models with 100 times the performance per watt of traditional graphics cards. Mythic's APU is based on a design that it describes as a compute-in-memory architecture. The chip is made of memory circuits that not only store information but also process it. Calculations are carried out using resistors, tiny electronic components that inhibit the flow of electrons. An AI model is a collection of artificial neurons, code snippets that are organized into groups called layers. Each layer carries out a small portion of the calculations involved in analyzing a prompt and then passes its output to the next layer, which repeats the process. Mythic says that its APU can run an AI model's layers in parallel rather than one after one to speed up processing. The company is productizing its technology with a device called Starlight. According to Mythic, it contains multiple APUs that cumulatively consume less than one watt of power. Starlight can be embedded in edge systems such as robots to enhance the quality of the data collected by their image sensors. Mythic also sees customers deploying its silicon in data centers. Ahead of today's funding announcement, the company tested the APU's ability to run large language models. It determined that APU-powered servers can process up to 750 times more tokens per second than graphics cards. Mythic provides a software toolkit that makes it easier for developers to adapt their LLMs to its chips. The toolkit optimizes AI models using quantization, a method that compresses neural network parameters to reduce their memory footprint. Mythic's software can further boost an LLM's performance by retraining it for the APU.