clements

FlexAI plans to combat artificial intelligence's growing compute power A startup promising “universal AI compute” has joined France’s thriving artificial intelligence sector. Paris-based FlexAI emerged from stealth today with a €28.5mn war chest and an ambitious plan to “unlock the future of machine intelligence.” To pick that lock, FlexAI is rearchitecting compute infrastructure. The startup’s centrepiece is an on-demand cloud service. Set to launch later this year, the product will enable developers to build and train AI applications with heterogeneous compute architectures. An integrated software layer will abstract and orchestrate the workload to the optimal architecture. As a result, users can apply multiple hardware configurations without complex code changes. The goal is to maximise efficiency and simplicity. By providing easy access to a wide range of workloads and compute, FlexAI aims to expand access to AI. The startup’s founders, Brijesh Tripathi and Dali Kilani, told TNW that their ultimate goal is enabling AI solutions to the world’s biggest problems. Their hope is that more access leads to more innovations. “We believe in AI’s transformative power to solve some of humanity’s biggest problems, but it will require a 1000x magnitude more compute to be able to realise this vision,” said Tripathi, FlexAI’s CEO. “The availability of AI compute today is limited to a select few. Our vision is to unlock access to compute for the many.” AI’s big compute problem FlexAI arrives amid growing alarm about AI’s voracious appetite for computing power. As the field advances, the breakthroughs are becoming increasingly unaffordable, inaccessible, and unsustainable. FlexAI promises solutions to all these problems. For startups, the service cuts the vast costs of compute, which can consume 80% of a company’s capital. By automatically assigning the compute required for a given workload, FlexAI could ensure that customers only pay for what they actually use.  The approach would also reduce AI’s expanding carbon footprint. By 2030, artificial intelligence is set to consume more power of India, according to the CEO of chip giant Arm. As models increase in size and complexity, the need for greater energy efficiency will only grow. FlexAI’s CEO Brijesh Tripathi (right) and CTO Dali Kilani co-founded the startup after spells at Apple, Intel, Nvidia, Tesla, Lifen, and Zoox. Credit: FlexAI The cloud service also aims to mitigate a growing skills shortage. Current AI infrastructure requires highly-specialised expertise to run. By building a software layer that simplifies the process, FlexAI hopes to enable more startups and developers to build AI apps.  This trio of solutions have already attracted $30mn (€28.5mn) of seed funding. The round was led by Alpha Intelligence Capital (AIC), Elaia Partners, and Heartcore Capital. Antoine Blondeau, Managing Partner at AIC, highlighted the pressing need for FlexAI’s service. “Compute is lagging, not leading, AI models’ increasingly potent capabilities,” he said. “To deliver on AI’s potential, the industry has to solve the issue of compute costs and that of stack complexity — both of which dramatically increase development costs and time to market.” “Brijesh and Dali, uniquely, have the hardware and software skill sets to architect and build the ecosystem for the AI compute orchestration layer.” Enhancing France’s AI sector The investors also praised the growing power of France’s AI ecosystem. Led by Parisian unicorn Mistral AI, the country has become a European hub of artificial intelligence. With a strong scientific base, a growing roster of startup founders, and increasing access to global capital, France’s AI sector has “global ambitions,” Blondeaue said. The sector’s rise will also uplift Emmanuel Macron. The French president’s mission to create “a startup nation” has thus far attracted both admiration and mockery. In the thriving AI scene, he now has evidence that his vision can become a reality.