Indian start-ups say OpenAI’s pricing is prohibitive, eroding ROI compared with human labour
Founders urge OpenAI to replicate AWS,Google Cloud’s playbook of supporting startups with usage credits
Start-ups want sub-second response times and stronger vernacular support to cut costs and reach non-tech users
For years, India, with 886 million internet users as of 2024, has been a crucial market for Big Tech companies like Meta and Google, not because of revenue, but because of scale.
Artificial intelligence, however, is unfolding differently. Unlike social media platforms that flourished on free access, AI tools are built on monetisation models from the outset. Companies like OpenAI and Perplexity have leaned into subscriptions and paid APIs.
India has emerged as OpenAI’s largest user base, a fact not lost on CEO Sam Altman. He has already visited the country twice in the past year, and has publicly acknowledged India as his second-biggest market, with the potential to become its largest. His upcoming third visit underscores this priority, coinciding with the company’s decision to open its first corporate office in New Delhi later in 2025.
Ahead of the visit Outlook Business spoke to more than a dozen start-up founders, VCs and policymakers to understand their expectations from OpenAI. Here they are:
Reduce cost of tokens
Cost, in particular, has emerged as the biggest pain point. Today, OpenAI’s speech costs about $0.50 per minute, but in India the viable price is more like $0.05 per minute, a huge gap.
Take the example of CreativebitsAI, a start-up using GPT-4. In one of its healthcare outsourcing projects, processing a single page with GPT-4 costs about 14 cents, compared to 6 cents with human labour. The model also delivers lower accuracy, around 88% versus 94% by human coders. “If we can maintain the same quality but cut token costs by 40–60%, customers could see a positive ROI of 20–50%. That would make a big difference,” says Sumit Talwar, Sales Director at CreativebitsAI.
Experimental credits
Experts point out that using OpenAI isn’t free: every prompt and reply on GPT-4.5 carries a cost. For start-ups, even testing ideas can, therefore, quickly become expensive. They argue that OpenAI should offer free credits, much like Amazon Web Services (AWS) and Google Cloud (GCP). In simple terms, credits work like a prepaid balance that lets start-ups experiment with OpenAI’s services before dipping into their own funds.
“Support could mean help in building test or development systems. It could mean technical assistance when we hit roadblocks. It could also mean credits to use advanced models like Sora,” which is a “fantastic video generation model,” but very expensive. “So for companies like us, initial support in using Sora would make a big difference,” adds Sekhar Maddula, founder, Cartoon Movie.
Need for speed
Latency, or the time between command and response, poses another challenge for start-ups. In real-time conversations, even a few seconds of delay feels jarring. “It can go up to six–seven seconds, whereas for real-time conversation it needs to be under one second to be useful,” says Sudarshan Kamath, founder of smallest.ai, an Indian start-up that joined OpenAI’s meeting with entrepreneurs in San Francisco. Kamath adds that his platform is actively providing feedback to help improve these models.
Host data centres in India
Start-ups believe OpenAI must host data locally, so it stays within the country. Says Nikhil Ambekar, Co-founder and CEO of Turinton: “It would improve trust with customers, ensure compliance with data residency and privacy regulations, and enable adoption in sensitive sectors like finance and healthcare. It would reduce latency and make AI services more reliable for Indian start-ups.” Hosting data locally would eliminate the back-and-forth of messages to the US and back, making AI services far more responsive for Indian start-ups.
This call coincides with OpenAI’s reported talks with players like Yotta and Ctrl+S around plans to set up a large data centre in India with a capacity of at least 1 GW.
Start-ups in India stands to significantly benefit in other ways too should Altman decide to invest in local infrastructure not only for hosting but for running models (inferencing) and training new ones. Among the major beneficiaries would be start-ups working on multiple Indian languages and unique use cases, says Shweta Rajpal Kohli of the Startup Policy Forum. The more Indian data the models learn from, the better they get, which then helps start-ups build stronger products. This virtuous cycle would make AI more relevant in India while equipping local start-ups to scale and compete globally.
Improve Indian language support
OpenAI claims access to 11 Indian languages. But experts say only a handful (Hindi, Bengali, Marathi, Telugu, and Tamil) work reasonably well. This is a concern.
Even within Hindi, regional dialects (Delhi vs. Bihar vs. Punjab) are not captured accurately. Result? Founders end up stitching together multiple tools — OpenAI for speech-to-text, Sarvam for text-to-speech, and third-party software for tonality, trebling the cost of every workflow. OpenAI could help start-ups save all this time and money, they add.
“Companies like OpenAI have a responsibility. If they are choosing to operate in a developing country like India, they should ensure AI reaches common, non-tech people as well,” says Nishant Raj, Founder of Ailivate Ltd.
While it remains to be seen whether OpenAI can fully crack the Indian market, there is a clear sense of optimism among founders and investors. “OpenAI coming to India definitely means more focus on the ecosystem. Many of our portfolio companies already use OpenAI for parts of their stack, whether infrastructure or applications,” says Abhishek Srivastava, General Partner at Kae Capital.
He points out that AWS, Microsoft, and Google followed a similar playbook when they entered India: test the market, erect local data centres and start-up-focused divisions, offer credits, build communities, and invest.
“That ecosystem support was huge for start-ups, free credits allowed them to start from day zero without capital. Communities gave them access to scaled founders and global insights. And the cloud giants themselves invested in some cases,” Srivastava added.
Experts believe that if OpenAI adopts a similar approach, it could cement its place in India’s start-up ecosystem. But the prize is bigger than market share: it is the chance to shape how the world’s most populous
