Ola will save about INR 15 crore a year by switching to Krutrim AI Cloud from Amazon and Microsoft Azure cloud infrastructure. “Cloud fees are absurd. In an exclusive interview with AIM, Sasank Chilamkurthy, the founder of Qure AI and Von Neumann AI (the business developing a personal AI server JOHNAIC), stated that Bhavish Aggarwal “basically used to burn INR 5 lakh per day, which is quite a bit of money.”
After quickly doing some number-crunching, Chilamkurthy estimated that the cost of purchasing servers on Azure that would support Ola’s present load would be between INR 8.2 and 10 crore. The cost over three years would have been INR 54.74 crore if they hadn’t moved.
This works out to about INR 45 crore, which is a significant savings of nearly INR 15 crore each year (INR 54.75 crore minus INR 8.2 crore).
According to Chilamkurty, if a business reserves an AWS g4dn.4xlarge instance for three years, the cost of the instance will be INR 10–25 lakh.
Estimates and statements made by the Ola CEO on X suggest that this change may result in a daily revenue loss of INR 5–25 lakh for Microsoft Azure and INR 30–40 million for AWS. This might equal about INR 18.25-91.25 crore for Azure and INR 109.5-146 crore for AWS on an annual basis.
Furthermore, Chilamkurthy claims that Ola paid roughly INR 85 lakh to Azure for egress charges. The fee that cloud providers charge for moving or transferring data from the cloud storage where it was previously uploaded is known as the egress cost.
Chilamkurthy said to AIM that he met Aggarwal at a cafe recently, when he gave him a pitch for Krutrim AI Cloud’s JOHNAIC. According to him, Aggarwal expressed some interest in JOHNAIC, in part because it makes use of Intel GPUs as well. He went on to say Ola Krutrim will purchase an Intel Gaudi 2 in order to construct the Krutrim AI Cloud.
Over 2,500 developers have signed up for Krutrim Cloud services, according to a recent post by Aggarwal, and they will all be onboarded in the upcoming weeks. Aggarwal most recently traveled to Taiwan to see Arm chief Rene Haas.
Cloud in a Box: JOHNAIC
Recently, Von Neumann AI introduced JOHNAIC, a “cloud in a box” solution. Scalability and flexibility are among the advantages of cloud computing that are offered, but in a local on-premises setting. With this hybrid approach, local data control and cloud capabilities coexist in perfect harmony.
The following hardware specifications are included: 64 GB of RAM, an Intel Arc 770 16 GB GPU, an Intel i5 12400 CPU, a 1 TB SSD, and an optional 1100 VA UPS.
Chilamkurthy remarked, “We’ve sachetized AI into little boxes so you can personalize them and take them wherever you want.” With built-in SaaS and AI capabilities to manage SMEs and startups, it promises to reduce AI costs by 85%.
JOHNAIC supports oneAPI, runs Meta Llama, and includes Ubuntu, Matrix, and ERPNext. Chilamkurthy stated, “JOHNAIC is a one-time investment of about INR 2 lakh, resulting in 80-92% savings.” Users do not need to purchase an Apple MacBook, which is comparable in cost, in order to run AI applications.
With JOHNAIC, Chilamkurthy is presently aiming for the inference market. “For the majority of people and startups looking to do something interesting with OpenAI GPT’s APIs, it is very reasonable and affordable,” he stated.
People+AI is JOHNAIC’s first client; it uses the technology for its own AI needs while protecting the privacy of its data. Because JOHNAIC has a very affordable pricing structure when compared to public cloud services like AWS, it is a desirable choice for companies trying to cut expenditures on AI infrastructure without sacrificing functionality.
“It’s actually not possible to buy an NVIDIA A100 alone on AWS; instead, you have to buy it as a bundle of eight GPUs, which is rather pricey in my opinion. One A100 would be plenty for the majority of people, who would only use it for a week, according to Chilamkurthy.
Intel has been going after India hard lately in an attempt to take the market share that the non-availability of NVIDIA GPUs has caused. “I think NVIDIA overcharges us, so I don’t utilize them. They encourage big server sales rather than providing tiny packages like sachets, according to Chilamkurthy.
