ET Explainer: What OpenAI's local data residency means for India
Tired of too many ads?
Remove Ads
Tired of too many ads?
Remove Ads
Tired of too many ads?
Remove Ads
ChatGPT-maker OpenAI earlier this month enabled local data residency in key Asian countries including India—its second largest market—and Japan, Singapore, and South Korea. This was in a bid to help organisations who want to leverage its ChatGPT Enterprise, ChatGPT Edu, and the OpenAI API (application programming interface) offerings, but also have data localisation requirements. ET explains what this move means for Indian businesses and whether data sovereignty is on the horizon.The feature allows 'data at rest' such as prompts, uploaded files, and chat interactions to be stored within India. But, models still reside in foreign servers and processing enterprise information at inference time (run-time) will need exchange outside India servers.Data localisation has until now prevented OpenAI from gaining market share in India as BFSI customers opted to host open models like Meta's Llama and DeepSeek on-premise.According to Aadya Misra, Partner at Spice Route Legal, 'OpenAI's residency option could allow financial institutions to deploy AI for use cases like payment processing while remaining compliant with existing requirements that require payment data to be stored locally.'She explained that the Reserve Bank of India does permit transient cross-border processing under certain conditions, 'so if implemented thoughtfully, concerns about data in motion could also be addressed. This move could shift reliance on self-hosted open-source models to enterprise-grade and centrally managed AI solutions.'The move may at best be seen as a first step towards compliance-enablement that could help companies to bolster contracts with responsible data handling clauses. It has fallen short of complete data sovereignty, however, experts said.'The architecture stores data 'at rest' locally, but not necessarily 'in transit' or during model inference. That data may still leave the country, exposing enterprises to regulatory scrutiny,' said Leslie Joseph, principal analyst at Forrester.Joseph noted that OpenAI has not announced local hosting of its GPT models or inference engines in India. 'There's no evidence of compute or model weights residing in-country. This is partial localisation at best, not sovereign AI,' Joseph added.He explained that although OpenAI has added AES-256 level encryption for data at rest and TLS 1.2+ for data in transit, without full model localisation, including inference compute, enterprises handling PII (personally identifiable information) will continue to face regulatory and data exposure concerns.'...There is no explicit indication that the underlying GPT models, including their inference engines, tokens, or trained weights, will themselves be hosted in India,' said Ankit Sahni, Partner at Ajay Sahni & Associates.Speculation remains that OpenAI may eventually bring full-stack model hosting to India, given its enterprise ambitions and steady competition from cost-effective open-weight models. For now, experts say companies must treat this as a 'compliance-forward gesture.'It could also mean opportunities for Indian data centre players.Although the company is likely to host local storage within its long-time exclusive partner Microsoft's data centres, sources told ET that OpenAI is hearing proposals from other colocation data centres in India as well.'Given OpenAI's shift to a for-profit structure and changing dynamics with Microsoft, we are actively seizing this opportunity to commit to a long-term relationship with them,' the senior executive at a leading data centre company told ET.Annapurna Roy contributed to this story.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
&w=3840&q=100)
Business Standard
a minute ago
- Business Standard
Timeline: How Rajasthan became India's top solar hub in just a decade
A decade ago, Rajasthan's vast desert stretches, scorching in the day and freezing at night, were seen as empty, unforgiving land, hostile to most forms of development. Today, those same empty plains sparkle with millions of solar panels, making the Thar one of the most powerful clean energy engines in the world. From the dusty villages of Jodhpur to the futuristic sprawl of Bhadla, the state has dramatically transformed–no one could have seen that coming. With great policy support, billions in investment, and overwhelming geography, Rajasthan has vastly climbed to the top of India's renewable energy map. What once was a story of survival in the desert has become one of scale, ambition, and leadership in the race to a low-carbon future. Here is how the journey unfolded. The journey of solar power production over a decade The foundation for Rajasthan's solar success was laid in December 2014, when the Centre launched the Solar Parks and Ultra Mega Solar Power Projects Scheme. Rajasthan was among the earliest states to sign up, with Bhadla identified as the flagship site for large-scale development. Just a few years later, in 2017, the Bhadla Solar Park in Jodhpur district began to take shape, attracting major developers such as ACME, Adani, and Hero Future Energies through Solar Energy Corporation of India Limited (SECI) auctions. The same year proved decisive for India's tariff story as well. In May 2017, Bhadla's competitive auctions discovered a record-low solar tariff of ₹ 2.44 per unit, a benchmark that redefined expectations for solar pricing nationwide and continued to shape auctions through 2018. By 2020, Rajasthan's installed solar capacity had grown to about 5.1 GW, a figure that positioned it for the surge to follow. That surge came between 2021 and 2022, when Rajasthan added capacity on an unprecedented scale. In 2022 alone, the state brought online 6.7 GW of new solar and wind power, about 43 per cent of all renewable capacity added in India that year. By June 2022, Rajasthan's large-scale solar had already reached close to 13 GW, thanks largely to central auctions conducted by SECI. The growth story soon began drawing global capital. In 2023, funds such as Brookfield poured money, financing projects in Rajasthan through a ₹3,380-crore pool. Masdar, the Abu Dhabi-based developer, tied up with Hero Future Energies on a 2.2 GW India platform, with 1.5 GW of that capacity linked to projects in Rajasthan and Haryana. Indian companies like ReNew were also continuing to create their presence across the solar clusters in the state. By early 2024, Rajasthan is estimated to be above 20 GW, even tracker companies placing its operational solar base above 21 GW, and this is a lead which none of the other states can match. This trend continued till mid-2025, and official MNRE data demonstrated that Rajasthan had achieved 32.32 GW of solar capacity by July 31. Not only does this establish Rajasthan as India's undisputed leader in renewables, but it also calls itself home to the world's largest solar park in Bhadla at 2,245 MW installed capacity. Why Rajasthan leads in solar power production Rajasthan's dominance in solar energy is rooted in its geography, policy, and scale. The state enjoys consistently high solar irradiation of around 5.5-6.5 kWh/m²/day, coupled with vast stretches of arid land across its western districts that are ideally suited for gigawatt-scale solar parks. The early decision to participate in the Centre's Solar Park Scheme gave Rajasthan a head start, while a clear state framework for aggregating land and setting up park-specific special purpose vehicles reduced developer risk and shortened project timelines. Scale has also been central to Rajasthan's rise. The Bhadla Solar Park, spread across about 56 square kilometers and was developed in multiple phases starting from 2015, became the world's largest single solar park with an installed capacity of 2,245 MW and a showcase of pooled land, ready infrastructure such as roads and pooling stations, and competitive auctions that pushed tariffs to record lows of under ₹2.50 per kWh. Alongside this, investments in transmission under the Green Energy Corridor ensured that power could be evacuated both within and outside the state. Phase II of the corridor, approved in January 2022, is on track for commissioning by March 2026, promising even greater grid flexibility. What's next for renewable energy in Rajasthan Rajasthan's clean energy journey is expanding beyond solar parks and is being taken one step further. The Centre's National Green Hydrogen Mission has already named the state one of two (the other being Ladakh) leading candidates for large-scale hydrogen production in the country. The rationale is simple and logical - Rajasthan has available land and substantial solar potential, making it a suitable area for electrolyser-based projects, resulting in competitively priced green hydrogen. Additionally, the next wave of hybrid projects that integrate solar, wind, and storage technologies is being accommodated with additional transmission corridors under Green Energy Corridor-II. With over 32 GW of solar installed already, the state should comfortably remain India's top state for solar throughout this decade, which is key for the 500 GW renewable energy commitment by 2030, with hybrid parks and larger storage projects leading the next wave of growth.
Economic Times
4 minutes ago
- Economic Times
Three-lenders cap makes micro borrowers fall in line
Synopsis Borrowing by vulnerable economic groups is changing. The number of borrowers with many lenders decreased significantly. This reflects stricter lending practices. Data shows a drop in borrowers using more than three lenders. However, some older loans still pose a risk. The microfinance sector is now focusing on stability. The market size has contracted. Getty Images The number of bottom-of-the-pyramid borrowers taking loans from four or more lenders has come down 44% from a year ago, reflecting the impact of prudential lending in the most vulnerable economic strata where mounting delinquencies dented the performance of Indian lenders in the last five number of such borrowers queuing up at more than three financiers fell to 3.1 million at the end of the June from 5.7 million a year ago, showed the CRIF High Mark data, illustrating the efficacy of a three-lender exposure cap that took effect from the start of this fiscal. The latest count represents about 4% of an active borrower base of 79.8 3.1 million customers, who could be considered 'higher risk' have ₹35,712 crore of outstanding borrowing between them at the end of June. That amounts to about a tenth of the sector's outstanding loan portfolio of ₹3.6 lakh portfolio relating to borrowers transacting with more than three microfinance lenders remains very vulnerable to slippages, especially after the guardrail 2.0 regime limiting fresh loans from more lenders to such borrowers, Crisil Ratings' director Malvika Bhotika said a fortnight ago. "As per our estimates, almost a fourth of the portfolio with more than three lenders is in the PAR (portfolio at risk) 31-180 days bucket as on June 30, 2025," she said. The portfolio exposure of borrowers with high lender associations declined from 19.2% of the book over the past 12 months, CRIF data this transition has lowered the exposure of MFIs to over-leveraged borrowers, slippages from the legacy portfolio will keep asset quality under pressure, experts said. The issue of borrower overleveraging has been at the heart of the high delinquencies over the past one years leading to severe asset quality stress and loss in income for lenders. It's ironic that lending to the micro borrower segment, which was once known for their robust repayment behaviour, is now being considered risky. "The industry is prioritising risk management and portfolio stabilisation over aggressive growth, reflected in subdued lending activity and a stronger focus on liquidity," CRIF said. This moderation aligns with the implementation of guardrails by self-regulatory organisation Microfinance Institutions Network to mitigate overleveraging risks. This led to a 17% year-on-year contraction in the size of the microfinance market to ₹3.59 lakh active loan accounts fell to 132 million from 159.3 million, with the live customer base declining to 80 million from 86.6 million between June 2024 and June 2025, data compiled by CRIF sector's portfolio at risk for more than 90 days past due remained elevated at 15.54% of the total on- and off-balance sheet book. This means the gross non-performing assets without excluding the technically written-off loans stood at ₹55,820 crore at the end of June. Loans are classified as NPA when borrowers fail to pay interest or principal for more than 90 days.
Time of India
8 minutes ago
- Time of India
'Very worried about China', says OpenAI CEO Sam Altman with a warning for America and the new 'China-Safe' chips policy
OpenAI CEO Sam Altman has issued a stark warning that America may be underestimating the complexity and seriousness of China's advancements in artificial intelligence. During a rare on-the-record briefing with a small group of reporters, Altman said that relying on export controls alone is likely not a reliable solution to the geopolitical AI race. "I'm worried about China," Altman stated, speaking over Mediterranean tapas in San Francisco's Presidio, last week. According to a report in CNBC, he emphasized that the US–China AI race is deeply entangled, encompassing more than just who is ahead on a simple scoreboard. Altman detailed the various layers of the AI landscape, including inference capacity, research, and product development. "I don't think it'll be as simple as: Is the U.S. or China ahead?" he said, adding that China might be able to build inference capacity faster. 'China-Safe' chips policy not enough Despite the escalating U.S. export controls on semiconductors, Altman expressed skepticism that these policies are keeping up with the technical reality. When asked if it would be reassuring to have fewer GPUs reaching China, he responded, "My instinct is that doesn't work." by Taboola by Taboola Sponsored Links Sponsored Links Promoted Links Promoted Links You May Like TV providers are furious: this gadget gives you access to all channels Techno Mag Learn More Undo Altman elaborated on the challenges of a purely policy-driven approach, suggesting that export controls on one specific item might not be effective. "You can export-control one thing, but maybe not the right thing… maybe people build fabs or find other workarounds," he noted, referring to semiconductor fabrication facilities. "I'd love an easy solution," he added, "But my instinct is: That's hard." Altman's comments come as Washington continues to adjust its policies aimed at curbing China's AI ambitions. While the Biden administration initially tightened export controls, President Donald Trump went a step further in April by halting the supply of advanced chips altogether. However, just last week, the U.S. created an exception for certain "China-safe" chips, allowing sales to resume under a controversial agreement that requires chipmakers like Nvidia and AMD to give the federal government a percentage of their China chip revenue. The result is a complex and often contradictory set of rules that may be easier to navigate than to enforce. While U.S. companies deepen their reliance on chips from key American firms, Chinese companies are actively developing alternatives from domestic suppliers like Huawei. This raises questions about whether the strategy of cutting off supply is having its intended effect, or simply accelerating China's push for self-sufficiency in AI technology. AI Masterclass for Students. Upskill Young Ones Today!– Join Now
