ElastixAI, a Seattle-based AI infrastructure startup founded by former Apple and Meta machine learning researchers, emerged from stealth with $18 million in seed funding to address systemic inefficiencies and high costs in generative AI inference.
The company is launching a software platform that transforms off-the-shelf FPGA-based servers into highly efficient AI supercomputers. ElastixAI says its software, ML, and hardware co-design approach can reduce the total cost of ownership for large-scale language model inference by up to 50x and reduce power consumption by 80% compared to traditional GPU-based systems.
According to the company, the AI inference market is expected to reach $255 billion by 2030, but existing infrastructure remains fundamentally unsuitable for generative AI workloads. LLM inference is memory-bound, whereas standard GPUs are designed for compute-bound tasks such as training, resulting in low compute utilization, wasted capital, and excessive energy consumption. The company also noted that custom silicon development cycles can exceed three years and often lag behind rapid advances in machine learning technology.
ElastixAI positions its platform as a drop-in replacement for traditional GPU workflows, enabling denser execution of LLM operations while maintaining compatibility. The company says this approach eliminates “dark silicon” by activating only the circuitry needed for inference, enabling cutting-edge AI implementations on current hardware without waiting for next-generation chip cycles.
The platform is currently available to select enterprise partners, data center operators, and AI model providers.
Important quote:
“The industry is currently sabotaging orders of magnitude performance because the hardware cannot keep up with ML advances. We are moving away from ‘one size fits all’ hardware. By applying unique post-training optimizations to the FPGA, we adapt the hardware to the model rather than forcing the model to struggle on the hardware.”
Dr. Mohammad Rastegari, Co-Founder, ElastixAI
