Positron AI

Just 2 weeks after Mitesh Agrawal joined us as CEO, we are very excited to announce that Positron AI has closed $23.5M in seed funding, with new investors including Valor Equity Partners, Atreides Management, LP, and Flume Ventures. More than anything, we are proud to be shipping systems providing >3.5x performance per dollar and performance per watt advantage over NVIDIA Hopper *TODAY*, which we started delivery to customers prior to this latest round. And our chips and systems are designed, fabricated, manufactured, and assembled in the USA. #MadeInUSA Even more exciting announcements ahead!

Positron AI is pleased to announce the appointment of Mitesh Agrawal as our new CEO. With a background as the former COO and Head of Cloud at Lambda, an AI compute unicorn, Mitesh brings invaluable experience to our team. Having dedicated ourselves to developing cutting-edge chips for accelerating generative AI inference specifically for transformer models, we are thrilled to kick off our first multi-rack deliveries, which will be unveiled shortly. During his impressive tenure at Lambda, spanning nearly 8 years, Mitesh played a pivotal role in driving the company to a remarkable $2B valuation and generating over half a billion in revenue. Mitesh's arrival aligns perfectly with Positron's global expansion efforts, as we commence the distribution of our cards, servers, and comprehensive rack-scale systems to clients worldwide. Welcome aboard, Mitesh! (link in comments)

Positron is gearing up for some major announcements in the upcoming weeks. To kick off the new year, we're highlighting the benefits of #MadeInAmerica, showcasing our Altera FPGAs produced by Intel Foundry. The future of AI Inference Infrastructure is here in the USA! Read more about the advantages of manufacturing advanced chips in the USA: https://lnkd.in/g8gwPEHz.

Watch how Positron AI is turning 10 kW racks into 5✖️ better #AI computing capability, compared to traditional GPU systems, by using Altera's Agilex™ 7 #FPGA. View the live demo from #Supercomputing 2024 ▶️: https://lnkd.in/eJwvr34j Positron showcased its Atlas AI acceleration solution running #Meta’s #LLaMA3.1 8B model built on BittWare’s IA-860m cards, powered by Altera’s Agilex 7 M-Series #FPGAs. The results: ✔️ 2x the speed vs. NVIDIA H100 #GPUs ✔️ 3x more energy efficiency ✔️ 4x better performance per dollar

Check out my conversation with Chris Saad on The Startup Podcast, talking about Positron AI's *now shipping* hardware with greater performance per dollar and performance per watt versus NVIDIA. https://lnkd.in/ghVUHgXm

We're thrilled to share that Positron was listed #3 in the Computing category of The Information's 50 Most Promising Startups. https://lnkd.in/grz8BFmi You can also check out Scott Thurm's article covering the announcement here: https://lnkd.in/enhQn3C4

Positron AI was featured as #3 in the Computing category of The Information's 50 Most Promising Startups 2024!

Positron is proud to share our latest inference performance versus the GPU-based competition: ✅ 70% faster token generation on Llama3.1-8B ✅ 1/3 power usage on Llama3.1-8B ✅ 51% cost savings versus DGX-H100 💸 (Yes, less than half the cost.) Positron Atlas also supports Llama3.2 and Mixtral models at higher token rates and lower power consumption than on NVIDIA DGX-H100. (In this case, running on Lambda's Cloud.) Visit https://positron.ai to learn more.

Check out my keynote from the AI Hardware Summit 2024 and listen to how fundamental macroeconomic principals should have people thinking beyond the common "AI is a bubble" meme, and finally how my team at Positron AI are delivering the best performance per dollar and performance per watt for generative AI applications: https://lnkd.in/gzRrSUYN

It's time to consider alternatives to GPUs—bringing down the cost and power consumption of serving LLMs is now possible! Learn more from Thomas Sohmers in this week's issue of Cerebral Valley's newsletter: