Open Cloud Compute: A UPI moment for AI?
But compute is awfully expensive. For a country like India, which needs scale and cheaper compute, the cost is only going to get increasingly higher, making it inaccessible to large sections.
Open Cloud Compute (OCC), an initiative of People+AI is looking to solve this very issue. It is founded by EkStep Foundation, which Infosys cofounder Nandan Nilekani started in 2015 and is trying to offer an alternative for larger players by creating a network of micro data centres.
Tanuj Bhojwani, who heads People+AI, told ET that a key challenge is the lack of options for big companies. “Today, the entire internet is paying taxes to large cloud companies but 90% of applications don’t need them,” he said. “Unfortunately, we don’t have an alternative.”
Deploying open source isn’t the solution either as this is equally expensive. That’s where OCC could help since its mission is to make compute more freely and cheaply available as demand rises inexorably. Vishnu Subramanian, founder of JarvisLabs.ai, a GPU rental and cloud platform, said it was uneconomical to deploy even open-source models on the scale that India needs. “This means we need a diverse set of compute providers, who offer different sets of GPUs (since not all tasks will require expensive high-end GPUs), and data centres,” he said. “The idea of OCC is, why should there only be one or two hyperscalers. We can have many so that wealth creation is also diversified.” GPUs or graphics processing units are chips with the power to run AI applications. In the long run,
