Mythic bet big on analog AI but has run out of cash
Veep still says 'technology will eventually be very successful'
A startup that bet on the old concept of analog chips to provide energy-efficient AI computing has run out of capital.
Texas-based Mythic "ran out of runway with the investors before we could get to revenue," said Ty Garibay, the startup's vice president of engineering, in a Tuesday LinkedIn post. Garibay and several other Mythic employees said on the social network that they are looking for new jobs.
Garibay said Mythic had developed a "unique new technology" that promised to deliver performance-per-watt and performance-per-dollar gains for edge AI. However, that apparently wasn't enough to give the 10-year-old company the traction it needed to keep investor money flowing.
Mythic CEO Mike Henry did not respond to a request for comment.
Mythic has operated in a crowded AI chip market that remains largely dominated by general purpose chips (GPUs) versus custom cores in the datacenter and at the edge. It hasn't helped that the global economy is weakening, raising the prospect of a recession in the near future.
These factors are testing the financial stability of AI chip vendors, who raised billions of dollars from venture capitalists in 2021 as the economy started to slow down.
For instance, UK-based Graphcore said in September that it was planning job cuts due to an "extremely challenging" macroeconomic situation. The next month, The Times reported that investors had slashed its valuation by $1 billion in the face of financial woes, including a terminated deal with Microsoft.
Also in October, Habana Labs, an AI chip company owned by Intel, reportedly laid off roughly 10 percent of its workforce, which amounted to around 100 jobs being cut.
Future of Mythic's analog chips up in the air
Garibay told The Register that Mythic's management team has decided to not discuss the company's status at this time. However, he left open the possibility that there could be a future for Mythic's Analog Matrix Processors (AMP), for which he said there are no "comparable solution in the industry."
"I believe that Mythic's technology will eventually be very successful, and I am deeply personally disappointed to not be able to continue with that development myself," he said in an email. "For now, my focus is on placing the great engineers in my teams in good new positions as quickly as possible."
Mythic has pitched its AMP chips as the best option for edge AI devices in a variety of markets, arguing that the digital chips made by most AI chip designers can't keep up with the "demanding requirements of size, performance, and power consumption."
The startup's chips come in a variety of form factors, including a stand-alone processor as well as an M.2 PCIe card and a standard PCIe card. A year ago, the upstart said its MP10304 PCIe card, powered by four M1076 AMP chips, was capable of delivering up to 100 trillion operations per second of AI performance while only using less 25 watts of power.
Mythic turned to analog chips — used for computers decades ago before digital circuits become all the rage — in trying to figure out how to bring processing elements as close as possible to the memory, a challenge many chip designers are trying to solve to improve performance and efficiency.
- Startups bag billions to fill gaps left by chip world giants
- AI caramba, those neural networks are power-hungry: Counting the environmental cost of artificial intelligence
- Intel takes on AMD and Nvidia with mad 'Max' chips for HPC
- Chipmakers are crippling products to evade US China ban
By using analog chips instead of digital chips, Mythic was able to take the concept of "compute-in-memory to an extreme" and "compute directly inside the memory array itself," according to the company.
"This is possible by using the memory elements as tunable resistors, supplying the inputs as voltages, and collecting the outputs as currents," Mythic said on its website.
Garibay said Mythic proved its point at the AI Hardware Summit in September, where the company showed off analog chips that could run the YOLOv5 object detection algorithm on high-resolution video at 60 frames per second while only consuming 3.5 watts.
"I am extremely proud of what our team achieved in taking a concept that almost everyone else in the industry thought was impossible to implement and building it in to a real working silicon product," he said.
As of last year, Mythic had raised a total of $165.2 million from investors, which included Hewlett Packard Enterprise and New York investment giant BlackRock. ®