Oct 5, 2026
As Andhra Pradesh races to build one of India’s largest data-centre and AI infrastructure clusters, Rohit Kashyap’s Maytree AI is weighing a role that runs from backing a local startup to a stake in the state’s emerging Data Park.
By Staff Correspondent
AMARAVATI / NEW DELHI: The most valuable layer of the artificial intelligence economy is rarely the one consumers see. It sits underneath the chatbots and copilots, in the compute, power, data and institutions that let technology companies scale.
Andhra Pradesh has decided to compete for that layer. The state wants large-scale data-centre capacity, sovereign AI infrastructure, a startup pipeline, a skilled workforce and, ultimately, the jobs that should follow. It wants all of them at once.
Maytree AI, founded by entrepreneur Rohit Kashyap, is now exploring where it might fit. Its immediate interest is modest: supporting an Andhra Pradesh-based startup. In parallel, the company is studying opportunities around the state’s planned Data Park and wider AI infrastructure push.
The pairing is what makes the move worth watching. It is less the story of one company entering a new state than a test of a question India will face repeatedly over the next decade: can the country turn its AI infrastructure build-out into companies, research and jobs, or will it remain a landlord for other people’s servers?
The plumbing beneath AI
Public debate about AI tends to begin with models. Every model, however, depends on an expensive physical stack: computing capacity to train and run it, data, cloud platforms, high-speed networks, reliable electricity, cooling, engineers and, finally, a market willing to buy what gets built. That dependency is turning the global AI race into an infrastructure race.
Andhra Pradesh has placed a deliberate bet on that stack. Its Data Centre Policy 4.0 (2024–29) seeks to attract new capacity and makes specific provision for AI-enabled data centres and large data-centre parks. A separate startup policy runs alongside it, offering a framework for innovation and entrepreneurship within the state.
The ambition lies in combining the two. Data centres supply the infrastructure; startups supply experimentation; AI is the technology layer; talent is the human capital; and government can supply real problems and early demand. If those pieces mature together, the whole could be worth considerably more than its parts.
The Lokesh thesis
The strategy is closely identified with Nara Lokesh, the state’s Minister for IT, Electronics and Human Resource Development. What sets the current push apart is less the size of the investments under discussion than the attempt to treat technology as an ecosystem rather than a list of projects.
The clearest expression of that idea is the AI Living Labs initiative. The programme bills itself as a statewide, governed AI co-creation platform resting on three assets: shared compute, governed data and future-ready skills. It is meant to bring citizens, students, startups, researchers and government departments onto common ground, with shared GPU capacity, data sandboxes, skilling and certification, innovation challenges, a marketplace for AI solutions and research partnerships. Lokesh has pitched it as a launchpad for innovators to build and deploy solutions with real-world impact.
The model inverts the usual sequence of government technology. Conventionally, a department defines a need, buys a product and deploys it. Under the Living Labs approach, the state identifies a problem, opens up infrastructure and data, lets startups and researchers experiment, validates what works and only then deploys it.
Whether that holds up at scale will depend on execution, governance and adoption by departments. As a design for public innovation, though, it is more sophisticated than most.
6.5 gigawatts, and a cluster to match
The numbers explain the industry’s attention. In September, Lokesh said the state had secured agreements covering its targeted 6.5 GW of data-centre capacity and that the projects were moving towards execution. By Indian standards, that is an extraordinary figure.
Capacity, however, is not an ecosystem. The questions that matter come afterwards: how many companies emerge, how many engineers are trained, how much research happens locally, how many public problems the infrastructure helps solve, and how much of the value it generates stays in Andhra Pradesh. The answers will decide whether the state becomes a technology producer or merely a host.
Lokesh has said the state will develop data centres in planned clusters rather than letting them scatter, describing the approach in a September interview as systematic. The logic is sound. Concentrated facilities can share power and fibre, a pool of skilled labour, vendors, security and maintenance services, training institutions and research partners. Silicon Valley proved the power of clustering decades ago; Bengaluru repeated it for India’s software economy. Andhra Pradesh is betting it can build the AI-era equivalent, with compute as the anchor.
Visakhapatnam is at the centre of that geography. Its industrial base, coastline, connectivity and access to energy have made it the focus of the state’s plans, and the large AI and data-centre projects proposed there have drawn national and international interest.
The city’s real gain, though, will depend on what grows around the hyperscale sites. A data centre employs a relatively small, specialised workforce. A functioning ecosystem creates demand for software developers, AI engineers, data scientists, cybersecurity and cloud specialists, power engineers, researchers, founders and a long tail of service businesses. That multiplier, not the direct payroll, is the prize.
Lokesh has framed it in precisely those terms. In a September 2026 interview, he called jobs the central priority of the government’s technology agenda and tied the data-centre strategy to that goal. The aim is not simply to build computing capacity but to use it as an economic catalyst.
The hard part: power, water and sustainability
No honest account of the opportunity can skip its physical cost. AI infrastructure consumes enormous amounts of electricity, needs water or other resources for cooling, and requires land, transmission lines and resilient connectivity. Those pressures grow with every additional megawatt.
They are already part of the state’s political debate. In August 2026, Lokesh fielded questions in the Assembly about the environmental and water demands of the proposed facilities. He said the 6.5 GW target had been secured, that water for the data centres would come from industrial allocations rather than the city’s drinking supply, and pointed to newer cooling technologies and environmental safeguards. More recent reports indicate the state is moving towards a framework that would require new data centres to draw a substantial share of their power from renewable sources.
That shift matters, because the AI build-out cannot be separated from the energy transition. The winning locations will not be those with the cheapest land. They will be those that can offer reliable power, clean power, water security, connectivity and regulatory certainty together. Andhra Pradesh’s record on those fronts will count as much as its success in signing investors.
From sovereign AI to local capability
A further layer of the strategy is sovereign, locally deployable AI. In February 2026, the state announced collaborations with IBM, BharatGen and NxtGen on a proposed sovereign AI framework that would combine models, data, cloud and GPU infrastructure, and applications for government, startups, academia and industry.
The rationale extends well beyond procurement. Governments want control over sensitive data, businesses want secure infrastructure, and countries want compute on home soil. AI that works in Indian languages also needs datasets and models built around local linguistic and social realities.
That gives Andhra Pradesh two distinct goals: to become a place where the world’s AI infrastructure is built, and to develop the local capability to put that infrastructure to use. The second may prove the more valuable.
Where Maytree AI fits
This is the territory Maytree AI appears to be studying. The company does not seem to view Andhra Pradesh as simply another market for AI services; its interest sits closer to the point where infrastructure meets entrepreneurship.
The gaps at that point are easy to identify. An AI startup may struggle with the cost of compute. A government department may hold valuable data but lack the capacity to turn it into a product. A university may have researchers but no access to production-grade infrastructure. An enterprise may have a problem and no team to solve it. A Data Park designed as more than a cluster of server halls could become the physical and commercial meeting point for all of them.
Maytree’s startup engagement offers a way into the local innovation ecosystem; the Data Park exploration represents the infrastructure end. Between the two lies a large and largely unclaimed middle layer: AI applications, startup acceleration, enterprise adoption, skilling, applied research, government technology, innovation centres, data services and deployment.
For a company built around AI and entrepreneurship, that middle layer may be strategically more interesting than becoming one more operator in the data-centre business. It also suggests the opportunity should not be read as a new office, an investment or a single project, but as a chance to help shape an ecosystem while it is still forming.
For Kashyap, the question has shifted from what Maytree can sell in Andhra Pradesh to what role an AI and entrepreneurship company can play in the system the state is trying to build. There is no answer yet, and the company is still in the exploratory stage. But pursuing a startup engagement and the Data Park at the same time signals a view that the next generation of AI businesses will be built between infrastructure and innovation, not inside either one.
The real test
The broader economic argument is straightforward. India spent three decades building IT services clusters; the next decade may be about compute clusters. The first wave of Indian technology ran on human capital. The next will be capital-intensive, energy-intensive and talent-intensive at once, and that changes the map. A state now needs power, fibre, compute, electronics supply chains, research and talent, along with policy that lets them work together.
Andhra Pradesh is trying to build all of those layers simultaneously. The obvious risk is fragmentation: infrastructure arriving before startups, startups before talent, talent before customers, and government programmes that exist on paper without adoption. The model works only if each layer reinforces the others.
That is also why the most consequential part of Lokesh’s approach may not be the headline numbers. Gigawatts, billion-dollar commitments and hyperscale campuses make for good announcements. His emphasis on Living Labs, shared compute, skills, startup participation and real government problems reflects a recognition that infrastructure alone does not produce an innovation economy. As a matter of technology policy rather than politics, that deserves credit. If it works, the lesson travels: states competing for AI investment would need to treat data centres less as real-estate projects and more as economic ecosystems.
The true measure will not be visible for five or ten years. Will the state have produced globally competitive startups? Will local engineers be building frontier systems, and will government departments routinely deploy AI? Will universities do serious applied research? Will young people find high-value technology jobs without leaving Andhra Pradesh? And will the value created by the infrastructure travel beyond the facilities themselves?
For now, the state is wagering that infrastructure, startups, skills and AI can be developed together rather than in sequence. Maytree AI is deciding where it fits in that experiment. The immediate story is a startup; the emerging one is a Data Park. The larger story is whether a state can turn compute into capability, and infrastructure into entrepreneurship.
That is a far harder proposition than building data centres. For Maytree AI, the question is whether it can become part of the story while it is still being written.
