Bengaluru-based space tech company Grahaa Space announced it will launch its Solaras S2 nano-satellite by the end of November, following authorisation from the Indian National Space Promotion and Authorisation Centre (IN-SPACe).
The company plans to send the satellite from the Alcântara Space Centre in Brazil on the Hanbit-Nano rocket built by Korea-based Innospace. The mission will demonstrate the company’s satellite systems and mark its first orbital launch.
Solaras S2 is part of Grahaa Space’s plan to build a constellation of nano-satellites for near-real-time earth observation. The company signed an agreement with Innospace in 2024 for this technology demonstration mission. It aims to qualify its bus, platform and other subsystems before moving to later missions.
Founder and CEO Ramesh Kumar V said the Solaras S2 mission will help the company validate key systems. “It is a focused technical step that confirms our readiness for the next phase,” he said.
He added that support from STIIC at IIST Trivandrum and IN-SPACe has been central to the company’s progress. Grahaa Space is backed by the Viskan Group and incubated at the Space Technology Incubation and Innovation Centre (STIIC) at IIST.
The company said its subsequent missions are planned for early 2026 with Skyroot Aerospace.
These launches will test the communications module, collect geospatial data through the optical payload, and establish inter-satellite links. Kumar said the goal is to build “a reliable nano-satellite capability that can support various on-ground applications using near-real-time geospatial data.”
Grahaa Space builds payloads, inter-satellite links and onboard processing systems to support live geospatial data. Solaras S2 will serve as a qualification step before the company moves to a full constellation rollout.
The post Bengaluru’s Grahaa Space Prepares to Launch Nano Satellite by November End appeared first on Analytics India Magazine.
Every evening in Malad, a teenage Dhravya Shah would wander through Mindspace with a glass of Boost in hand, thinking about whatever he was building that week.
Just a kid who liked long walks and late-night code, who somehow ended up shipping 60+ products before 19, hacking for Cloudflare, and qualifying for an O-1 “extraordinary ability” visa at 20.
Today, that same kid sits in San Francisco in a “solo together” house, running one of the Valley’s most intriguing infra startups. Supermemory has already raised $3 million from Jeff Dean, Dane Knecht, Logan Kilpatrick, and a who’s who of AI founders betting on the memory layer of agents. Supermemory is simple to describe and brutally hard to build: a universal memory API that gives AI agents long-term, cross-app memory.
It ingests everything—files, documents, emails, chats, app streams—and converts them into a personalised, evolving knowledge graph that any agent can query instantly. It already powers AI video editors, assistants, agentic workflows, and multi-tool orchestration engines that need to remember users across months, not minutes.
Why Memory, And Why He Walked Past YC
After applying for the YC Summer Fellows grant, Dhravya got in. Shah told AIM that he walked away from Y Combinator because investors outside YC offered significantly better terms, deeper infrastructure expertise, and immediate strategic value for the company.
“I got the summer fellows grant, but then I decided to build Supermemory as a company instead,” recalled Shah, in an exclusive interview on Front Page by AIM Network. He added that “the valuation that they offered was already like six times more than YC.”
The investors he eventually chose, leaders from Google, Cloudflare and the broader infra ecosystem, now review architecture, guide research, open doors to design partners, and validate the technical rigour of Supermemory’s roadmap.
“They are some of the best people in this industry,” he said. “They act as a testament to how rigorous our own processes are in terms of infrastructure and research.” As a solo founder building in one of the hardest technical categories, particularly infra + memory, Shah said the right guidance mattered more than YC’s brand recognition. “I do think that maybe getting YC is better for recognition,” he said. “But that’s the wrong kind of recognition for a B2B SaaS guy like me. For me, this is the best thing that could happen.” He clarified that he never applied to YC’s traditional fundraising batch, only the Summer Fellows grant—and by the time the opportunity arose, external interest had already pushed Supermemory into a different orbit.
“We actually do industry partnerships as well with these companies,” he said, noting prior work with Cloudflare and ongoing collaboration with Google’s models.
The Noise, the Media, and Perplexity
Shah told AIM that most stories the Indian media publishes about him are simply wrong: the “IIT dropout” tag, assumptions about his upbringing, or that he comes from an elite network. “There’s multiple problems with the media narrative… none of it is true,” he said. “People think I’m from an elite rich family or that all my elite connections gave me money. I’m just a dude from Malad. My dad is a businessman. I know no one in tech.”
He said the oversimplification extends to his product too. “People say, ‘oh this is just RAG, I built this over a weekend,’ but they either don’t understand the product or only know the version from two years ago. This has taken years to build.”
Shah has given up, he doesn’t fight the noise anymore. “I just ignore it… play into it… use it as marketing, even to boost ARR.” He also rejected the doom narrative around Perplexity with one of the sharpest lines of the interview. “Pessimists sound smart, but optimists make all the money,” he said.
People underestimate how hard it is to build a product that even ten people genuinely love, he added, which is why he respects what Aravind Srinivas has achieved. “I have huge respect for Perplexity and Aravind… they’ve built something millions of people love. That’s insanely hard.”
He believes Perplexity will continue to win because they have fundamentals and iteration velocity on their side. “Once a company gets to that stage, they can keep people happy… they have it in their blood. Perplexity will win no matter what people think.”
The post The 20-Year-Old Who Said No to YC appeared first on Analytics India Magazine.
2025 has been a pivotal year for AI, with technology finally moving from hype to actual deployment and models becoming more capable. However, a few AI startups couldn’t keep up with the race and had to shut shop due to lack of investor interest, poor product-market fit or other internal bottlenecks.
Here are seven AI startups that closed in 2025.
1. Builder.ai
Builder.ai entered insolvency in 2025 after nearly a decade of promising AI-powered app building. Founded in 2016 by Sachin Dev Duggal and Saurabh Dhoot, the company began as Engineer.ai. It was headquartered primarily in the United Kingdom and the United States, with subsidiaries in India, Singapore, and other locations.
The company positioned itself as a platform that could assemble fully functional applications with minimal human involvement, but investigations later revealed that much of the development work was handled manually by offshore teams. This mismatch between marketing claims and operational reality began eroding trust.
The financial unravelling came quickly, with revenue inconsistencies, increasing operational debt, and a major creditor dispute eventually pushed the company into insolvency proceedings.
2. CodeParrot
CodeParrot, a YC-backed AI developer-tool startup, shut down in mid-2025 after struggling to scale its flagship product. The startup, founded by Vedant Agarwala and Royal Jain in 2022, had raised $500,000.
The company focused on converting design elements, such as Figma screens and interface mockups, directly into React components or full-stack code. While the concept was strong and early demos impressed developers, the product failed to generate production-grade code that teams could confidently deploy. It struggled with a long period of ‘pivot hell’, frequently changing its product for a viable business model, resulting in diluted focus and confused investors.
As heavier competition emerged—including from GitHub Copilot, Vercel, Replit, and several LLM-powered coding agents—CodeParrot’s niche advantage began to shrink.
3. Astra
Astra, an AI-powered sales intelligence startup founded in Bengaluru, closed operations in late 2025 after being active for just a year. Backed by Aravind Srinivas, co-founder of Perplexity AI, the company aimed to solve one of the biggest problems in enterprise SaaS by improving slow and inefficient sales pipelines. Its platform analysed calls, emails, and CRM data to provide deal insights and automate outreach workflows.
However, internal differences among co-founders Supreet Hegde and Rajan Rajagopalan and a lack of market readiness led to an organisational split and delayed key product milestones.
It also faced challenges around working with large enterprises and navigating lengthy sales cycles, especially at a time when data privacy concerns and hallucination risks were receiving more regulatory attention.
4. Subtl.ai
Subtl.ai, a Hyderabad-based GenAI knowledge-automation startup, shut down in July 2025 after failing to raise additional funding. The company built tools that allowed employees to query internal documents, SOPs, and databases using natural language, speeding up workflows across support, ops, and sales teams.
While subtl.ai had strong early traction and a compelling problem statement, it struggled with product-market fit at scale. Many companies experimented with the tool but did not convert to long-term paid plans, often citing accuracy issues or complexities in integrating large document bases. Funding conditions tightened in 2025, and without renewed capital, the founders decided to shut down operations.
5. Humane
Humane, the highly hyped consumer AI-hardware startup founded by ex-Apple veterans, effectively shut down operations in 2025 after discontinuing its AI Pin device. The startup sold its AI Pin business to HP Inc. for $116 million, including most of its employees, software platform, and intellectual property, after poor reviews and dwindling sales.
The company imagined a future beyond smartphones, one in which an AI wearable could answer questions, project an interface on the user’s hand, and serve as a personal assistant.
Despite an enormous wave of publicity, the AI Pin faced problems immediately after launch: short battery life, overheating, inconsistent responses, and an unclear use case.
Reviews were overwhelmingly negative, and return rates soared. The hardware-AI integration proved far more complex than anticipated, requiring real-time inference that the device couldn’t reliably support.
6. Wuri
YC-backed AI startup Wuri shut down after struggling to achieve sustainable growth. Founded as an enterprise AI startup, Wuri faced intense challenges, including high customer acquisition costs, difficulties in scaling products, and fierce competition from similar AI wrapper applications that lacked strong differentiation or proprietary technology.
The startup’s founder, Akshay Megharaj, described the fast pace of AI development as a major hurdle, noting that rapid change made it impossible to rely on past experience or traditional strategies.
7. Locale.ai
Locale.ai, an operations-intelligence and geospatial analytics startup, ceased operations in 2025 after struggling to scale its enterprise pipeline. The company built AI tools that helped businesses analyse logistics, supply chain anomalies, and rider or delivery patterns in real time.
Founded by Aditi Sinha and Rishabh Jain in 2019, the company faced critical challenges despite generating decent revenue and securing international customers.
After raising about $5 million and navigating several difficult economic periods, including the COVID-19 slowdown, the co-founders became severely burned out after years of nonstop work. Even though they saw new AI opportunities in sales automation, they chose to close the company responsibly by returning money to investors and helping customers move to other options.
The post 7 AI Startups That Failed in 2025 appeared first on Analytics India Magazine.
Karnataka announced six new letters of intent (LoIs) and a set of skilling partnerships at the Bengaluru Tech Summit 2025 on November 19. The projects include new units in electronics manufacturing, EV systems, battery recycling, biotechnology and a dedicated drone testing site. The government said the initiatives will add jobs across key sectors and strengthen efforts to expand growth beyond Bengaluru.
The IT minister, Priyank Kharge, said the agreements form part of the state’s wider plan to support innovation clusters and deepen the Local Economy Accelerator Programme (LEAP). The announcements under LEAP include Elevate Next, beginning January 1, 2026, and Elevate Beyond Bengaluru, starting February 1, 2026. He revealed that LEAP is a five-year, ₹1,000 crore programme designed to strengthen clusters outside Bengaluru through infrastructure, funding and sector support.
As many as 40 startups will receive grants ranging from ₹50 lakh to ₹1 crore under Elevate Next, while 50 startups will receive grants of up to ₹50 lakh under the Elevate Beyond Bengaluru programme. The minister further said that the expansion will help “decentralise opportunities” and strengthen the state’s innovation network.
The first LoI covers a drone testing facility to be set up on a 20-acre site in Chintamani by the Drone Federation of India. The government will provide the land, while the federation will bring companies to use the site. This has been provided with an outlay of ₹25 crore to ₹100 crore.
In another major agreement, Global HDI will establish a multi-layer PCB manufacturing plant on an 84-acre plot in Tumakuru with an investment of ₹1,500 crore. The state expects the project to create 1,000 jobs. Moreover, Elleve Solutions will set up a PCB and electronics manufacturing unit in the city, with an investment of ₹250 crore.
Meanwhile, TSUYO Manufacturing has signed an LoI of ₹250 crore to set up an EV powertrain unit in Dharwad. The company plans to produce powertrain systems and generate 700 jobs. Another facility worth ₹350 crore will be established by MiniMines Cleantech Solutions for critical mineral refining. That apart, Eyestem Research has signed an LoI for its vision-restoration work using cell-based therapies.
Alongside the investment announcements, the state also introduced a series of skilling initiatives. Marvell Semiconductor and the Electronics Sector Skills Council of India have signed an MoU to train 90 women in VSLI design and embedded systems. The programme will be offered in Tier 2 and Tier 3 cities and has been integrated with graduation-level courses.
The state also launched the Nipuna Karnataka initiative to train 4,000 youth in fields such as AI, cybersecurity and data science. Officials said the first phase will open access to 2,800 job opportunities with partner companies. The programme will run through four training aggregators working with corporate employers in the technology and finance sectors.
Moreover, in its Bengaluru Innovation Report, the government highlighted the city’s rise as a global tech hub. Bengaluru now ranks as the world’s fifth-largest AI hub, is India’s leading unicorn capital, and accounts for 58% of the country’s total AI funding.
Beyond Bengaluru, emerging clusters in Mysuru, Mangaluru and Hubballi-Dharwad are gaining momentum, backed by new infrastructure, rising investor interest and strong state support.
The post Karnataka Announces Initiatives Worth ₹2,600 Crore, New Skilling Partnerships appeared first on Analytics India Magazine.
Today’s AI, despite its power, is still primitive in usability, much like the early command-line days of personal computing. “We are in the MS-DOS era of AI,” according to Abhishek Mathur, VP of engineering at Figma, a San Francisco-based software company that provides a cloud-based design and prototyping platform.
Mathur told AIM, “every prompt, every agent, every new modality is a design waiting to happen,” underscoring that we are only seeing the early stages of what AI can do and become.
Users interact with AI through prompts the way people once typed commands into MS-DOS, effective, but far from intuitive or accessible. The real transformation will come when AI moves into its “Windows moment,” where visual, multimodal, and agent-driven interfaces replace text-based prompting and make AI truly seamless.
Just as graphical operating systems unlocked the mass adoption of computing, the next phase of AI will be defined not just by better models, but by better design.
Previously, chief product officer (CPO) at Microsoft, Aparna Chennapragada also told AIM that she thinks this is “the DOS-to-GUI moment of AI.” She added that Microsoft is pursuing two paths—integrating AI into familiar tools like Word, Excel and PowerPoint, and developing AI-first interfaces such as the M365 Copilot app and new AI editors.
The early days of computing were text-based, command-driven, and intimidating to the average user. Then came the graphical user interface (GUI), a breakthrough that made computing accessible to everyone.
Figma, Mathur believes, will play a similar role in AI-era creativity. “The next generation of designers will define how humans interact with AI. We’re only just getting started, ” he said.
Bengaluru as the Command Line
With its world-class talent across engineering, product, and design, Bangalore is uniquely positioned to shape this future. The city needs to now translate this depth into global leadership in AI product design.
The next generation of AI-native products will require multidisciplinary teams that combine model capability with deep understanding of workflows and user behaviour.
Leslie Joseph, principal analyst at research and advisory firm Forrester, told AIM that if the ecosystem continues to invest in product leadership, design education, and cross-functional collaboration, the city can evolve from being a delivery hub to being a birthplace of world-leading AI-native products.
India is Figma’s second-largest market, home to one of its most vibrant user communities. “Every year, 35 million design files are created in India,” Mathur said. “That’s 35 million ideas, products, and innovations, and it’s growing exponentially.”
The Friends of Figma community, made up of designers, engineers, and product leaders, is thriving across cities like Bengaluru, Pune, and Hyderabad.
Figma is constantly expanding its product ecosystem, from FigJam, the company’s collaborative whiteboard, to Figma Make, its latest prompt-to-application product. “Figma Make allows you to build applications inspired by your company’s design systems,” said Mathur. “So when you turn an idea into an app, it’s instantly on-brand, functional, and production-ready.”
The enthusiasm isn’t limited to design. India also leads globally in the adoption of Figma Make, the AI-powered product that turns text prompts into functioning prototypes.
“The highest number of makes in any country have been created in India,” Mathur shared. India is increasingly a development hub for Figma. The company is hiring across design, engineering, and developer advocacy roles in Bengaluru, with plans to build global products from India.
Role of AI in the Next Phase
The next leap in AI isn’t just about better models—it’s about the rise of agentic AI, where systems don’t merely respond but act. In this phase, AI agents will extend the reasoning abilities of underlying models into real-world products
“What distinguishes meaningful products is the way they translate ambiguous human intent into structured tasks, guide users through decisions, and create trust through clear boundaries and feedback,” Joseph added.
Talking to AIM, Siddon Tang, SVP of engineering and product, chief architect, and GM of the Asia Pacific region for TiDB, an open source database, stated that “Great product design is what makes AI feel magical.”
Tang further added that as we move into the era of AI agents, design becomes even more critical. Agents may run on top of LLMs, but they still depend on high-quality APIs, guardrails, permissions, memory structures, and interaction models.
All of that is product design.
This is the reason why “Product-led growth (PLG) becomes even more potent in the AI era,” Tang said, adding, “think of products like Dropbox, Lovable, Manus, and others—they scaled because the experience was so good that users naturally shared them.”
The biggest breakthroughs ahead won’t come from slightly smarter models but from products that translate model intelligence into real-world capability, clarity, and confidence.
Tools like Canva, Figma, Notion, and a wave of vertical SaaS platforms are embedding AI everywhere. And as users bring more ideas, workflows, and creative inputs into these tools, AI itself improves.
AI’s role in design, Mathur argued, is twofold: it lowers the barrier to entry while raising the ceiling of what’s possible.
“As an engineer or product manager, you can now be an effective designer,” he said. “And as a designer, you’re freed from repetitive tasks, localisation, renaming layers, filling mockups — to focus on craft, quality, and vision.”
Figma, in his words, is not about replacing designers, but about amplifying creativity across roles. “Designers will now be in the business of codifying quality,” he added. “They’ll define the brand’s DNA in design systems — and everyone else can build on it.”
The post ‘We Are in the MS-DOS Era of AI’ appeared first on Analytics India Magazine.
The steady drumbeat of data centre announcements across India provokes a critical question: Does the country actually need all of them?
Industry projections suggest the demand is real. India’s data centre demand is expected to surge from 1.3 GW in FY2025 to 4.7-5.7 GW by FY2030, attracting a significant influx of investment from both domestic and global players.
Recent industry data, expert perspectives and insights from the draft prospectus of Sify Infinit Spaces (Sify)—India’s first pure-play data centre company heading to IPO—provide clarity on whether this boom is justified or overblown.
The underlying drivers are strong. Internet usage has expanded dramatically, and according to a September report by Kotak Mutual Funds, India generates 20% of global data but stores only 3% of it locally.
Two forces are now working to close this gap: the growing demand for lower latency and the government’s push for data localisation.
But there’s also a third force, one that created this gap in the first place and continues to amplify the need for localisation and speed: the explosive growth of wireless internet usage in India, which underpins everything from social media to AI-driven services.
The Wireless Explosion
India has transitioned from a low-data market to one of the world’s heaviest users, driven by the affordability of data and a rapid shift towards video-heavy consumption.
Source: Kotak Mutual Fund
One indicator of AI’s growing demand is India’s rapid adoption of tools like ChatGPT.
Depending on the source, India now ranks either first or second in global usage. India is also a significant market for a wide range of AI products.
Recognising this, AI companies are planning data centres equipped with the GPU-based infrastructure needed to support these workloads.
Furthermore, the IndiaAI Mission, approved in March 2024 with a ₹10,372 crore outlay, aims to expand national AI infrastructure through a public-private partnership model.
Over 34,000 GPUs have been committed so far, with 17,000 already deployed by data centre providers such as Yotta, Netmagic and Sify, who supply compute power to startups and researchers via a shared platform.
This large-scale rollout not only fuels AI research but also accelerates the expansion of data centre capacity across India.
“India consumes a huge amount of digital assets and generates a massive amount of data that can be used to train models,” explained Amit Agrawal, president of Techno Digital.
This dual role positions India as an attractive location for AI companies looking to build and train their models.
Sify states that AI-related workloads will surge from less than 1% of total data centre workloads in India in Fiscal 2025 to 15-20% by Fiscal 2030.
Regulatory mandates also aid the demand.
The Reserve Bank of India’s 2018 directive mandated that all payment data be stored exclusively within India. SEBI’s 2023 mandate extended this requirement to stock exchanges, brokers, mutual funds, depositories and KYC agencies.
The upcoming Digital Personal Data Protection Act, 2023, added another layer of data localisation requirements across sectors.
Other policies, such as Digital India, BharatNet and National Digital Communications Policy, are all, in one way or another, promoting increased internet usage in India.
The Economics That Make It Work
But demand alone doesn’t build data centres. The economics must work—and in India, they do, often in ways that surprise people familiar with developed markets.
Data-centre development costs in India average just $7 per watt, among the lowest globally. Electricity costs are 20% lower than in the United States.
For infrastructure players with deep roots in the power sector, these numbers are especially compelling.
The country has solved the infrastructure constraint that would typically limit this kind of build-out. India transitioned from a power-deficient to a power-sufficient status, reaching a total installed capacity of 452.69 GW by October 2024, with essentially zero power deficit.
“Wherever there is power availability, the data [centres] will go there, not the other way around,” Agrawal pointed out. This inverts the traditional model where you build a data centre and then figure out power.
In May last year, the country reached its peak demand of around 250 GW and has met 242 GW so far in 2025. Renewable capacity reached 203.18 GW, accounting for 46.3% of total power.
Sify Infinit Spaces estimates that data centres currently consume less than 0.5% of India’s total power generation capacity, which means there’s headroom for massive expansion without straining the grid.
India also serves as a critical hub for global data transfers. The country hosts 17 international subsea cables, which land at 14 stations across five coastal cities: Mumbai, Chennai, Kochi, Tuticorin and Thiruvananthapuram.
In an earlier interaction with AIM, Sify’s CFO MP Vijay Kumar said these cities, especially Mumbai and Chennai, function like “international airports for data”.
These subsea cables enable high-speed, high-capacity data transmission, reducing latency and disruptions. Their proximity to landing stations makes these regions prime locations for large-scale data centre development.
Beyond these fundamental factors, both state and central governments are providing substantial incentives to establish data centre facilities.
Tamil Nadu offers a three-year waiver on electricity duty and stamp duty concessions. Uttar Pradesh provides an electricity duty exemption for 10 years and a stamp duty exemption on the first transaction. Meanwhile, Maharashtra offers 100% stamp duty exemption along with a permanent exemption from electricity duty.
Why Infrastructure Players are Flooding In
The macro story makes sense on paper, but understanding why established infrastructure players are flooding this market requires looking at their actual deployments.
Telecom giants like Airtel, broadband specialists like Sify, and conglomerates with expertise in power and real estate, such as Adani, have all entered the market, each leveraging their core strengths.
The same pattern holds across the industry, even for mid-sized and emerging players.
Take Techno Digital, which carries a 40-year legacy in power infrastructure through its parent, Techno Electric & Engineering Company. In October 2024, the company announced a $1 billion investment to build 250 MW of data centre capacity across India.
The rollout commenced with a 36 MW facility in Chennai, which launched in July, followed by an 18 MW facility in Noida and a 13 MW facility in Kolkata. “Tamil Nadu is one of the most conducive states in terms of doing business,” Agrawal noted, alluding to the policy backing the company received.
Techno’s foundations in engineering, procurement and construction, coupled with decades of experience delivering mission-critical power infrastructure, give the company a distinct advantage in designing, building and operating data centres closely integrated with power supply.
As these operators leverage their power and infrastructure strengths, data centres are expanding well beyond Tier-1 hubs.
Mumbai acts as the financial and global gateway, Noida benefits from government proximity and land availability, and Hyderabad offers a growing IT ecosystem with cost advantages.
Land economics drive this diversification. Premium micro-markets, such as Powai, West Hyderabad and Gurugram, are expensive; mid-range areas, including Thane-Belapur Road, Noida and Ambattur, offer a more balanced approach. Meanwhile, emerging zones like Panvel and South Hyderabad provide low entry costs and room for growth.
Latency is also a crucial technical factor driving this geographic expansion. “India is not a very big country, but it is a sizable country where east-to-west coverage, or north-to-south coverage, can take anywhere between 30 milliseconds and 45 milliseconds in the connected world,” Agrawal explained.
He pointed out how AI, which promises a variety of real-time applications and use cases, might lose its importance if latency increases to a second. This is pushing infrastructure towards Tier 2 and Tier 3 cities, where power is available and land costs are lower.
Read More: India’s Most Powerful AI Data Centres by Capacity
The Global-Local Partnership Model
A conducive ecosystem is forming between local and global players. Global hyperscalers like Google aren’t building everything from scratch; they rely heavily on Indian partners for land, power, construction and regulatory navigation. In turn, local firms gain access to cutting-edge AI and cloud platforms.
The Indian conglomerate Adani Group, for instance, was chosen to supply renewable energy to Google’s Indian cloud and data centre operations. In Jamnagar, Gujarat, Reliance Industries Limited (RIL) and Google Cloud are collaborating to build a dedicated AI cloud region.
RIL will design, build and power the facility, while Google Cloud supplies its global-standard AI infrastructure, compute, software stacks and services.
This data advantage, combined with competitive costs, power availability, a regulatory push and strategic partnerships, suggests that India’s data centre boom is a response to genuine market forces.
Inside a Major Operator: The Sify Model
In addition to these on-the-ground insights, Sify’s Draft Red Herring Prospectus (DHRP) offers deeper insight into the quantitative aspects of how a large-scale player operates in India.
It ranks among the top three colocation service providers in India, with a 15.26% market share by built IT capacity as of March 31.
The company continues to expand aggressively, including through large-scale projects such as its 130 MW AI-ready data centre campus in Siruseri, near Chennai.
Sify operates 14 operational colocation data centre facilities with a power capacity of 188.04 MW as of June 30.
These facilities are spread across six cities: Mumbai, Chennai, Noida, Hyderabad, Bengaluru and Kolkata.
The core capacity serves enterprise clients across banking, financial services, insurance, fintech and media, as well as hyperscaler customers, all of which require over 99.99% uptime and carrier-neutral hyperconnectivity.
For specialised AI workloads, Sify has deployed three new NVIDIA-certified DGX-Ready campuses engineered for high-density computing.
Sify Infinit Spaces is currently profitable. For the financial year ended March 31, revenue from operations was ₹1,428.36 crores and profit after taxes was ₹126.36 crores.
This explains why state electricity duty exemptions matter so much. A permanent exemption from electricity duty in Maharashtra doesn’t just improve margins by a few basis points; it fundamentally changes project economics. Various state policies also ensure that data centres receive an uninterrupted power supply.
The capital expenditure breakdown reveals where money actually goes when building a data centre. Land and the building shell account for 25-30% of project costs. Power-based infrastructure, such as substations and high-voltage systems, accounts for 4-6%. The remaining 60-65% goes to power-optimised design fit-outs covering electrical, mechanical, and cooling systems.
Besides, the state government’s support directly translates into Sify’s bottom line, in ways that go beyond tax exemptions.
The company’s Chennai 02 facility is situated on 4.90 acres of land leased from the State Industries Promotion Corporation of Tamil Nadu, a state government entity.
The terms include a 93-year lease, an upfront payment of ₹22.13 crores, and annual rent of ₹1 per year for 92 years—one rupee per year.
This type of arrangement makes projects pencil out faster by eliminating land cost escalation risk for nearly a century.
The Customer Concentration Reality
The top three clients of Sify, all hyperscalers, accounted for 67.04% of revenue in Q1 fiscal 2025.
While Sify has not explicitly revealed any names, it states that hyperscalers refer to companies such as AWS, Microsoft, Google and Oracle, which are consuming approximately 55% of Indian data centres’ IT megawatt capacity as of March 31.
That said, with hyperscalers building more of their own capacity in India, insourcing remains a real competitive risk for colocation providers.
Sify serves over 500 clients in total, including three of the top four global hyperscaler companies, seven of the top 10 Indian banks and four of the top 10 Indian insurance companies.
Contract economics show the long-term nature of the business. As of June 30, 67.04% of Sify’s revenue came from contracts with terms of at least seven years.
The average relationship length with the company’s top five clients was seven years. Most contracts include built-in rental increases of 2-4% annually, providing predictable revenue growth.
Sify’s utilisation profile shows steady absorption of the capacity it brings online. The gap between installed and operational capacity is narrow, which signals that most of what it builds is quickly backed by customer commitments.
Among Sify’s 188.04 MW of built capacity, 131.88 MW is installed, meaning 70.2% is usable today. 113.67 MW is operational, which is 60.4% of the total built capacity.
India’s data centre surge reflects genuine digital demand and strong economics, but it shouldn’t eclipse broader priorities.
As operators scale, the real test is whether growth can coexist with responsible resource use—and whether the country can extend reliable power, water and connectivity to households with the same urgency it brings to hyperscale infrastructure.
The post Unpacking India’s Data Centre Boom appeared first on Analytics India Magazine.
As AI continues to reshape industries and business models, Polestar Analytics is positioning itself not just as a service provider but as a strategic partner by helping organisations align data, decisions and transformation.
In a conversation with Polestar Analytics’ co-founders, Chetan Alsisaria (CEO), Amit Alsisaria (COO) and Ajay Goenka (CFO), AIM discussed how they envision AI evolving, what enterprise readiness truly means and why convergence is the next frontier for data-driven businesses.
Rethinking Enterprise AI
Reflecting on how enterprise AI has changed over the last two years, Chetan said the biggest realisation has been that success in AI is as much about defensibility as it is about innovation.
“The real gap lies between AI ambition and enterprise readiness,” he explained. “Many organisations still operate in silos across teams, technology and processes, leaving no clear path from pilot to production. The need of the hour is alignment, convergence, ownership and trust—not just algorithmic brilliance.”
He added that as hyperscalers and startups pour billions into AI, true differentiation will come from domain context, agility and trust by building IP that is industry-aware, outcomes-driven and closely aligned with the business.
Ajay, on the other hand, believes the next wave of AI leadership will require both technical fluency and human intelligence. “The next generation of AI leaders will need to connect data science with empathy, ethics and domain fluency,” he said. “We’re moving from a world of coders to a world of contextual thinkers.”
Betting on Convergence
Meanwhile, Amit described Polestar Analytics’ strategic bet on what he calls “the great convergence of data and process”.
He pointed out that while 80% of enterprise data is unstructured, most organisations still design their AI strategies around the structured 20%. “That’s like trying to understand human behaviour by only reading spreadsheets,” Amit said.
Amit explained that the real winners in the next decade won’t be the ones building the most advanced models, but those who can turn unstructured data—like emails, documents, conversations and videos—into actionable and operational intelligence.
AI Adoption in India
According to Ajay, Indian enterprises are at an inflexion point in their AI journey.
“A few years ago, the conversation was around AI experimentation. Today, it’s about how to scale responsibly and drive measurable outcomes,” he said.
Across the world, industries are no longer evolving through incremental change. They’re reimagining entire systems with AI that is contextual, cost-efficient and outcome-first.
“The biggest opportunity lies in convergence,” Ajay added. “True transformation will happen when data, decisions and delivery operate in one connected ecosystem.”
From Services to Platforms
Polestar Analytics recently raised new funding to accelerate the development of its 1Platform, an enterprise-grade AI and data convergence stack.
“Our fundraiser is a strategic step towards transforming from a services-led organisation into a platform-driven AI company,” Chetan said.
The company plans to deploy capital across three areas: IP development, enterprise expansion and global growth. “We’re doubling down on the convergence of data, decisions and automation, helping enterprises scale faster with governance and measurable impact built in from day one,” he said.
Amit cited a recent project with a consumer goods company where 1Platform unified production, sales and supply chain data into a single intelligent layer.
“Earlier, leaders were making decisions on outdated dashboards,” he said. “With 1Platform, they can now ask, ‘What’s putting my Q4 targets at risk?’ and get real-time, contextual answers.”
Collaborating Across the Ecosystem
On partnerships, Amit shared that Polestar Analytics’ collaborations with hyperscalers, startups and academia go beyond traditional alliances.
“With hyperscalers like Microsoft and Databricks, we’re doing co-creation—building joint solutions using Azure and Databricks stacks,” he said.
Polestar Analytics also partners with institutions like IIM Calcutta for the PGDBA programme. “We help universities align with real-world industry needs while tapping into fresh thinking,” Amit said.
On startups, he pointed out, “It’s about speed and specialisation. Startups bring focused innovation; we bring market access and implementation expertise.”
When asked what sets Polestar Analytics apart, Chetan said enterprises should seek partners that understand both business context and technical complexity.
“The right AI partner connects data, workflows and outcomes,” he said. “They take a neutral, interoperable approach and commit to ROI and governance from day one. That’s exactly what we’ve built with 1Platform.”
He added that the company’s differentiation lies in “speed to value, contextual intelligence and measurable business transformation.”
Balancing Ambition and Responsibility
For Ajay, responsible scaling is central to Polestar Analytics’ growth philosophy.
“It’s not about how quickly you can innovate, but how responsibly you can scale,” he said. “Every AI solution we build touches data, people and decisions. That comes with immense responsibility.”
Polestar Analytics embeds governance and ethical checkpoints into its delivery model. “Responsibility means ensuring every solution reflects our values as much as it drives value,” he added.
Through its centre of excellence, the company tests every emerging technology internally before deploying it for clients. “When we walk into a client meeting, we’re not speaking theoretically; we’ve already used it ourselves,” Amit said.
Looking Ahead to 2030
According to Chetan, many enterprises struggle to derive measurable value from GenAI due to poor governance and a lack of alignment.
“My advice: start small, but start with purpose,” he said. “Treat every pilot as if it’s going to scale. Build governance and LLMOps from day one.”
He shared that GenAI creates real value only when it’s built on an organisation’s own data, workflows and context, and when success is measured through meaningful metrics such as adoption, accuracy and tangible business impact.
Chetan described Polestar Analytics’ long-term goal as enabling organisations to operate as truly AI-first enterprises, where data, intelligence and execution function as a single, seamless system.
“If we’ve done our job right, enterprises will operate with greater agility, lower friction and higher trust,” he said. “Success is when AI isn’t a project anymore—but the default way enterprises think, decide and act.”
The post Inside Polestar Analytics and the Future of Converged Data Ecosystems appeared first on Analytics India Magazine.
xAI, the AI lab led by Elon Musk, released the Grok 4.1 AI model on November 17. The model is claimed to bring improvements in creative writing and emotional intelligence.
“It is more perceptive to nuanced intent, compelling to speak with, and coherent in personality, while fully retaining the razor-sharp intelligence and reliability of its predecessors,” the company claimed.
On the LMArena Text leaderboard, which evaluates AI models on text-generation quality by a blind test by human voters, Grok 4.1 Thinking stands at the #1 spot with 1483 points, and Grok 4.1 stands second at 1465 points.
On EQ-Bench, which evaluates the emotional intelligence capabilities of AI models, Grok 4.1 models occupied the top two positions.
Even on the Creative Writing v3 benchmark, Grok 4.1 Thinking and Grok 4.1 were among the top three models tested. The model is also claimed to bring lower hallucinations.
To achieve the above results, xAI stated, “We used the same large-scale reinforcement learning infrastructure that powered Grok 4 and applied it to optimise the style, personality, helpfulness, and alignment of the model.” The company also ‘silently’ deployed preliminary Grok 4.1 builds to users to gauge their preferences. “Compared to the previous production model in traffic, Grok 4.1 is preferred 64.78% of the time.”
Recently, CNBC reported that xAI is raising $15 billion in a Series E round; however, Musk denied the development in a post on X.
False
— Elon Musk (@elonmusk) November 13, 2025
The post Elon Musk’s Grok 4.1 Is the Best AI Model on LMArena Text appeared first on Analytics India Magazine.
