Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders.<\/em> Subscribe Now<\/em><\/p>\n\n\n\n As demand for large-scale AI deployment skyrockets, the lesser-known, private chip startup Positron is positioning itself as a direct challenger to market leader Nvidia <\/strong>by offering dedicated, energy-efficient, memory-optimized inference chips aimed at relieving the industry\u2019s mounting cost, power, and availability bottlenecks. <\/p>\n\n\n\n \u201cA key differentiator is our ability to run frontier AI models with better efficiency\u2014achieving 2x to 5x performance per watt and dollar compared to Nvidia,\u201d said Thomas Sohmers, Positron co-founder and CTO,<\/strong> in a recent video call interview with VentureBeat.<\/p>\n\n\n\n Obviously, that\u2019s good news for big AI model providers, but Positron\u2019s leadership contends it is helpful for many more enterprises beyond, including those using AI models in their workflows, not as service offerings to customers.<\/p>\n\n\n\n \u201cWe build chips that can be deployed in hundreds of existing data centers because they don\u2019t require liquid cooling or extreme power densities,\u201d pointed out Mitesh Agrawal, Positron\u2019s CEO and the former chief operating officer of AI cloud inference provider Lambda<\/strong>, also in the same video call interview with VentureBeat. <\/p>\n\n\n\n The AI Impact Series Returns to San Francisco – August 5<\/strong><\/p>\n\n\n\n The next phase of AI is here – are you ready? Join leaders from Block, GSK, and SAP for an exclusive look at how autonomous agents are reshaping enterprise workflows – from real-time decision-making to end-to-end automation.<\/p>\n\n\n\n Secure your spot now – space is limited: https:\/\/bit.ly\/3GuuPLF<\/p>\n\n\n\n Venture capitalists and early users seem to agree.<\/p>\n\n\n\n Positron yesterday announced an oversubscribed $51.6 million Series A funding round<\/strong>\u00a0led by Valor Equity Partners, Atreides Management and DFJ Growth, with support from Flume Ventures, Resilience Reserve, 1517 Fund and Unless.<\/p>\n\n\n\n As for Positron\u2019s early customer base, that includes both name-brand enterprises and companies operating in inference-heavy sectors. Confirmed deployments include the major security and cloud content networking provider Cloudflare<\/strong>, which uses Positron\u2019s Atlas hardware in its globally distributed, power-constrained data centers, and Parasail<\/strong>, via its AI-native data infrastructure platform SnapServe. <\/p>\n\n\n\n Beyond these, Positron reports adoption across several key verticals where efficient inference is critical, such as networking, gaming, content moderation, content delivery networks (CDNs), and Token-as-a-Service providers<\/strong>. <\/p>\n\n\n\n These early users are reportedly drawn in by Atlas\u2019s ability to deliver high throughput and lower power consumption without requiring specialized cooling or reworked infrastructure, making it an attractive drop-in option for AI workloads across enterprise environments.<\/p>\n\n\n\n But Positron is also entering a challenging market. The Information<\/em> just reported that rival buzzy AI inference chip startup Groq<\/strong> \u2014 where Sohmers previously worked as Director of Technology Strategy<\/strong> \u2014 has reduced its 2025 revenue projection from $2 billion+ to $500 million, highlighting just how volatile the AI hardware space can be. <\/p>\n\n\n\n Even well-funded firms face headwinds as they compete for data center capacity and enterprise mindshare against entrenched GPU providers like Nvidia, not to mention the elephant in the room: the rise of more efficient, smaller large language models (LLMs) and specialized small language models (SLMs) that can even run on devices as small and low-powered as smartphones.<\/p>\n\n\n\n Yet Positron\u2019s leadership is for now embracing the trend and shrugging off the possible impacts on its growth trajectory.<\/p>\n\n\n\n \u201cThere\u2019s always been this duality\u2014lightweight applications on local devices and heavyweight processing in centralized infrastructure,\u201d said Agrawal. \u201cWe believe both will keep growing.\u201d<\/strong><\/p>\n\n\n\n Sohmers agreed, stating: \u201cWe see a future where every person might have a capable model on their phone, but those will still rely on large models in data centers to generate deeper insights.\u201d<\/strong><\/p>\n<\/blockquote>\n\n\n\n While Nvidia GPUs helped catalyze the deep learning boom by accelerating model training, Positron argues that inference \u2014 the stage where models generate output in production \u2014 is now the true bottleneck. <\/p>\n\n\n\n Its founders call it the most under-optimized part of the \u201cAI stack,\u201d especially for generative AI workloads that depend on fast, efficient model serving.<\/p>\n\n\n\n Positron\u2019s solution is Atlas, its first-generation inference accelerator built specifically to handle large transformer models. <\/strong><\/p>\n\n\n\n Unlike general-purpose GPUs, Atlas is optimized for the unique memory and throughput needs of modern inference tasks. <\/p>\n\n\n\n The company claims Atlas delivers 3.5x better performance per dollar and up to 66% lower power usage than Nvidia\u2019s H100, while also achieving 93% memory bandwidth utilization\u2014far above the typical 10\u201330% range seen in GPUs.<\/p>\n\n\n\n Launched just 15 months after founding \u2014 and with only $12.5 million in seed capital \u2014 Atlas is already shipping and in production.<\/p>\n\n\n\n The system supports up to 0.5 trillion-parameter models in a single 2kW server and is compatible with Hugging Face transformer models via an OpenAI API-compatible endpoint.<\/p>\n\n\n\n Positron is now preparing to launch its next-generation platform, Titan, in 2026. <\/strong><\/p>\n\n\n\n Built on custom-designed \u201cAsimov\u201d silicon, Titan will feature up to two terabytes of high-speed memory per accelerator and support models up to 16 trillion parameters<\/strong>. <\/p>\n\n\n\n Today\u2019s frontier models are in the hundred billions and single digit trillions of parameters, but newer models like OpenAI\u2019s GPT-5 are presumed to be in the multi-trillions, and larger models are currently thought to be required to reach artificial general intelligence (AGI), AI that outperforms humans on most economically valuable work, and superintelligence, AI that exceeds the ability for humans to understand and control. <\/p>\n\n\n\n Crucially, Titan is designed to operate with standard air cooling in conventional data center environments, avoiding the high-density, liquid-cooled configurations that next-gen GPUs increasingly require.<\/p>\n\n\n\n From the start, Positron designed its system to be a drop-in replacement, allowing customers to use existing model binaries without code rewrites. <\/p>\n\n\n\n \u201cIf a customer had to change their behavior or their actions in any way, shape or form, that was a barrier,\u201d said Sohmers.<\/strong><\/p>\n<\/blockquote>\n\n\n\n Sohmers explained that instead of building a complex compiler stack or rearchitecting software ecosystems, Positron focused narrowly on inference, designing hardware that ingests Nvidia-trained models directly.<\/p>\n\n\n\n \u201cCUDA mode isn\u2019t something to fight,\u201d said Agrawal. \u201cIt\u2019s an ecosystem to participate in.\u201d<\/strong><\/p>\n\n\n\n This pragmatic approach helped the company ship its first product quickly, validate performance with real enterprise users, and secure significant follow-on investment. In addition, its focus on air cooling versus liquid cooling makes its Atlas chips the only option for some deployments. <\/p>\n\n\n\n \u201cWe\u2019re focused entirely on purely air-cooled deployments\u2026 all these Nvidia Hopper- and Blackwell-based solutions going forward are required liquid cooling<\/strong>\u2026 The only place you can put those racks are in data centers that are being newly built now in the middle of nowhere,\u201d said Sohmers.<\/p>\n\n\n\n All told, Positron\u2019s ability to execute quickly and capital-efficiently has helped distinguish it in a crowded AI hardware market.<\/p>\n\n\n\n Sohmers and Agrawal point to a fundamental shift in AI workloads: from compute-bound convolutional neural networks to memory-bound transformer architectures. <\/p>\n\n\n\n Whereas older models demanded high FLOPs (floating-point operations), modern transformers require massive memory capacity and bandwidth to run efficiently.<\/p>\n\n\n\n While Nvidia and others continue to focus on compute scaling, Positron is betting on memory-first design. <\/p>\n\n\n\n Sohmers noted that with transformer inference, the ratio of compute to memory operations flips to near 1:1, meaning that boosting memory utilization has a direct and dramatic impact on performance and power efficiency.<\/p>\n\n\n\n With Atlas already outperforming contemporary GPUs on key efficiency metrics, Titan aims to take this further by offering the highest memory capacity per chip in the industry. <\/p>\n\n\n\n At launch, Titan is expected to offer an order-of-magnitude increase over typical GPU memory configurations \u2014 without demanding specialized cooling or boutique networking setups.<\/p>\n\n\n\n Positron\u2019s production pipeline is proudly domestic. The company\u2019s first-generation chips were fabricated in the U.S. using Intel facilities, with final server assembly and integration also based domestically. <\/p>\n\n\n\n For the Asimov chip, fabrication will shift to TSMC, though the team is aiming to keep as much of the rest of the production chain in the U.S. as possible, depending on foundry capacity.<\/p>\n\n\n\n Geopolitical resilience and supply chain stability are becoming key purchasing criteria for many customers \u2014 another reason Positron believes its U.S.-made hardware offers a compelling alternative.<\/p>\n\n\n\n Agrawal noted that Positron\u2019s silicon targets not just broad compatibility but maximum utility for enterprise, cloud, and research labs alike. <\/p>\n\n\n\n While the company has not named any frontier model providers as customers yet, he confirmed that outreach and conversations are underway.<\/p>\n\n\n\n Agrawal emphasized that selling physical infrastructure based on economics and performance\u2014not bundling it with proprietary APIs or business models\u2014is part of what gives Positron credibility in a skeptical market.<\/p>\n\n\n\n \u201cIf you can\u2019t convince a customer to deploy your hardware based on its economics, you\u2019re not going to be profitable,\u201d he said.<\/p>\n Daily insights on business use cases with VB Daily<\/strong><\/p>\n If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI.<\/p>\n \n\t\t\t\t\tThanks for subscribing. Check out more VB newsletters here.\n\t\t\t\t<\/p>\n An error occured.<\/p>\n<\/p><\/div>\n
\n<\/div>
\n\n\n\n
\n<\/div>Entering a challenging market that is decreasing AI model size and increasing efficiency<\/h2>\n\n\n\n
\n
Atlas is an inference-first AI chip<\/h2>\n\n\n\n
From Atlas to Titan, supporting multi-trillion parameter models<\/h2>\n\n\n\n
Engineering for efficiency and compatibility<\/h2>\n\n\n\n
\n
Memory is what you need<\/h2>\n\n\n\n
U.S.-built chips<\/h2>\n\n\n\n
What\u2019s next?<\/h2>\n\n\n\n
![\"\"\/](\"https:\/\/venturebeat.com\/wp-content\/themes\/vb-news\/brand\/img\/vb-daily-phone.png\")
\n\t\t\t\t<\/div>\n<\/p><\/div>\n<\/div>\t\t\t<\/div>\r\n