11 Months On, Open Cloud Compute Still Awaits Launch

The Open Cloud Compute (OCC) network is a decentralized network comprising 10,000 small-sized data centers with the purpose of offering low-cost computing power to businesses in order to deal with the rising requirements of AI. But it hasn't launched yet, despite having been announced eleven months ago.

Insiders indicate that the network still needs to define its key objectives and determine whether it has sufficient computing capacity to deal with AI tasks.

"We must do more research to know what type of AI applications can be supported by this network," a source who asked not to be named said. "Currently, we are focused on this, and we are not adding any new computing vendors."

OCC was established by the AI non-profit group people+ai based in Bengaluru. The vision was to have an interconnected cloud computing platform where standalone providers of different GPUs could collaborate on a single platform so that customers could easily discover and utilize their services.

When it was announced, 24 firms, including the American semiconductor firm Advanced Micro Devices (AMD), Oracle, Dixon Technologies, Tata Communications, E2E Networks, IBM, and Dell Technologies, indicated that they would join OCC.

But others feel that the recent launch of the IndiaAI Compute Portal with more than 14,000 GPUs and competition from big cloud players may render OCC less attractive.

One of the biggest challenges for this consortium of micro data centers is that they must be able to provide sufficient computing capacity to be able to adequately serve AI workloads.

Individual micro data centers would have to spend a lot of money to buy GPUs, something they may not be able to afford, said Narendra Sen, the founding CEO of NeevCloud and one of the consortium partners.

The central government-led IndiaAI Mission has addressed the issue of computing resources by introducing the IndiaAI Compute Portal on March 6. The new platform will offer access to more than 14,000 GPUs at discounted hourly prices. "I think OCC may be attempting to do something similar, and that is why they are reluctant to proceed," Sen said in an interview with ET.

Vishnu Subramanian, CEO and co-founder of GPU cloud platform Jarvis Labs, said that they did a pilot program with OCC at PES University last October, but he hasn't heard from the network recently.

"We are in the OCC consortium. Had the network launched, we would have been one of their partners providing GPUs to consumers," Subramanian told ET. "Since we have not heard back from them for a while, chances are that they are not going to launch." Firms like people+ai, AMD, Oracle, and Dell Technologies declined to comment.

Dixon Technologies, Tata Communications, E2E Networks, IBM, and other members of the consortium, including Vigyanlabs, also did not reply to ET's query for comments until publication time on Friday.

Neil Shah, a partner at Counterpoint Research and an analyst in the tech industry, gave his views to ET, stating, "Distributed computing is a wonderful concept until it is implemented, which is why it has failed everywhere." He further stated that bigger firms, or hyperscalers, have always emerged victorious in this race.

Shah explained that there are two primary factors: cost and scale. The bigger companies can scale up, which enables them to reduce costs and provide more competitive pricing for computing resources. "On the other hand, microdata centers and distributed workloads are more effective when they are placed in certain locations where latency is important," he said.

From a scaling point of view, Shah does not think this model will work in India, because bigger hyperscalers will always be able to provide cheaper prices. 

People+ai is an arm of the EkStep Foundation, which was co-founded by Nandan Nilekani, co-founder of Infosys.