India’s AI Ambitions Need Domestic Hardware, Level-Playing Procurement, GPU Expansion: Vantageo CEO Madhusudan Bhor

From GPU gaps to policy bias, Bhor outlines what’s holding back India’s AI infrastructure ambitions

Mumbai: India’s aspiration to become a global AI powerhouse is being constrained by gaps in high-performance computing infrastructure and a procurement ecosystem that continues to favour foreign OEMs, says Madhusudan Bhor, Founder and CEO of Vantageo Private Limited. In an interview with Digital India Times Editor Srinivas G. Roopi, Bhor argues that accelerating domestic manufacturing, enabling fairer RFP frameworks, and investing in distributed, scalable GPU infrastructure are critical for achieving true computational sovereignty and unlocking India’s AI potential.

Edited excerpts:

Is India’s high-performance computing and data centre ecosystem ready to support large-scale AI adoption?

Honestly, when I look at where India stands today, I see a nation that is enormously ambitious but still catching up on the infrastructure side. We have the talent, we have the policy intent, and we certainly have the demand but the physical infrastructure to support large-scale AI adoption is still work in progress.

GPU ready data centres are largely concentrated in Mumbai, Bengaluru, and Hyderabad. The moment you move beyond these metros, the availability drops sharply. And even within these cities, the power reliability, cooling density, and network bandwidth required for serious AI workloads are not uniformly available. We see this firsthand when we work with PSUs and large enterprises trying to set up on premise or hybrid AI infrastructure.

That said, I am genuinely optimistic. The IndiaAI Mission is a meaningful step. The conversation around sovereign compute is happening at the right levels. What we now need is execution speed because the window to establish India as a serious AI infrastructure destination is open, but it will not stay open indefinitely.

How important do you think domestic manufacturing is in reducing India’s dependence on imported compute infrastructure?

This is something I feel very strongly about. At Vantageo, we made a deliberate choice to build as an indigenous OEM and that decision was rooted in a belief that India cannot afford to outsource its computational sovereignty.

Think about it. If your AI systems, your data processing engines, your mission critical compute infrastructure all depend on hardware shipped from overseas, you are exposed in ways that go far beyond cost. Supply chain disruptions, geopolitical uncertainties, export controls. We have seen all of these play out in recent years globally.

For government departments, defence establishments, and PSUs, this is not just a procurement preference. It is a security requirement. These organisations need to know exactly what is inside the systems they are running sensitive workloads on. When the hardware is designed, manufactured, and supported in India, that assurance is possible. When it is imported, it is not.

Domestic manufacturing also means faster support, better customisation, and systems built for Indian operating conditions including our power fluctuations, our ambient temperatures, and our specific compliance requirements. It is not just about nationalism. It is about practical, operational advantage.

But here is the ground reality that nobody talks about openly. A large number of government and PSU RFPs are still written in a way that quietly favours global players. Go and open any public sector bank RFP, any financial institution tender, any oil and gas procurement document and look carefully at the OEM eligibility criteria. You will see specifications and qualification requirements that are quite clearly tweaked to accommodate foreign players and effectively weed out domestic manufacturers. This is something we encounter regularly and it is deeply discouraging for genuine indigenous OEMs.

Now if the concern is trust, quality, or performance doubt on domestic players, I completely understand that. But there are fair and structured ways to address that. The most straightforward one is a Proof of Concept. Let domestic players demonstrate their systems in a controlled test environment. Evaluate them on the same parameters. That is a far more honest approach than writing eligibility criteria that eliminate Indian companies before the evaluation even begins.

And what makes this more frustrating is that even customers know what is really happening. Many of the so called foreign OEM solutions being supplied to government and PSUs today are essentially locally assembled products wearing an international brand badge. The foreign player tweaks their approach, does some local assembly, and walks away with the major share of the contract. The customer is aware of this arrangement and yet the procurement system continues to reward it over a genuinely indigenous manufacturer. This needs to change, and it needs to change at the policy and procurement framework level.

How Vantageo addresses the twin challenges of scalability and cost efficiency in deploying enterprise grade GPU computing systems?

This is the question I get asked most often by our clients, and rightly so. GPU infrastructure is expensive, and if you get the architecture wrong, you end up either under powered or massively over invested. Neither of which is acceptable.

Our philosophy at Vantageo has always been to start with the workload, not the hardware. Before we recommend any configuration, we spend time understanding what the client is actually trying to do. What models they are running, what inference latency they need, how their data volumes are expected to grow. That conversation shapes everything.

We design our systems to be modular and expandable. An organisation can start with a configuration that matches their current needs and scale incrementally as workloads grow, without having to rip and replace. That modular approach dramatically improves the financial case for AI infrastructure investment.

And because we are an indigenous OEM, we can offer pricing and support structures that imported alternatives simply cannot match. Our clients are not paying for a global supply chain markup. They are paying for Indian engineering and Indian service which, I would argue, delivers far better value in the long run.

The irony is that even when we offer better value, better support, and comparable or superior performance, the procurement process in many large institutions is structured in a way that makes it very difficult for us to even participate on equal footing. That is a structural problem that needs urgent attention.

How high stakes deployments like Chandrayaan-3 shape Vantageo’s engineering standards and enterprise credibility?

Chandrayaan-3 holds a very special place for our team. When you are contributing to a mission of that magnitude where the margin for error is essentially zero, it permanently changes how you think about engineering.

Everything we learned from that experience including the redundancy planning, the thermal resilience testing, the absolute insistence on reliability under sustained high load conditions, all of that became embedded in how we build every single system, whether it is for a space mission or an enterprise data centre.

I often tell our team: if it would not pass muster for a mission critical government deployment, it should not leave our facility. That standard has served us well. When a large corporate or a PSU chooses Vantageo, they are not just buying hardware. They are buying a certain assurance of quality that has been stress tested in environments where failure is not an option.

The credibility that comes from such deployments also opens doors. Decision makers in government and large enterprises are sophisticated buyers. When they see that our infrastructure has supported India’s space programme, the conversation changes entirely.

And this brings me to a point I feel strongly about. If there is any residual doubt about the capability or reliability of domestic players like us, the answer is not to write us out of the RFP. The answer is to put us to the test. Give us a Proof of Concept opportunity. Let us demonstrate performance in a real test environment against any benchmark you set. We welcome that scrutiny because we are confident in what we have built. What we do not welcome is being disqualified on paper before we even get the chance to prove ourselves.

What are the strategic shifts necessary for India to evolve from an AI talent hub to a global AI infrastructure powerhouse?

India’s AI talent is genuinely world class. There is no debate on that. But talent running on borrowed or imported infrastructure is not the same as a sovereign AI powerhouse. We need to close that gap deliberately and urgently.

On the policy front, I would love to see procurement frameworks evolve faster to prioritise domestic compute infrastructure, not just in principle but in practice. The GFR provisions are a good foundation, but implementation needs to be more consistent. We also need clearer mandates around data localisation and sovereign AI compute for critical sectors.

But let me be very direct here because I think this conversation demands honesty. The Make in India vision and the reality on the ground in government and PSU procurement are still quite far apart. Pick up any RFP from a public sector bank, a financial institution, or an oil and gas company and read the OEM eligibility section carefully. You will find qualifications that are structured, sometimes very subtly, to favour foreign players and exclude domestic manufacturers. This is not speculation. This is something indigenous OEMs like Vantageo encounter repeatedly in the market.

What is even more telling is that a significant portion of what gets supplied to government and PSUs under a foreign brand name today is essentially local assembly. The foreign OEM tweaks their model slightly, sets up a local assembly arrangement, and qualifies under the procurement criteria while a genuinely indigenous manufacturer who builds from the ground up in India gets screened out. Customers know this. Procurement officers know this. And yet the system continues to reward it.

If there is a genuine concern about the quality or performance of domestic players, I respect that. But the solution is simple and fair: conduct a Proof of Concept. Test the systems. Evaluate performance in a controlled environment. That is a far more credible and constructive way to build trust than engineering eligibility criteria that eliminate Indian companies before evaluation begins.

On the investment side, we need distributed, accessible GPU infrastructure in Tier 2 cities, in universities, and in research institutions. Public private partnerships can play a big role here, and the financing structures need to reflect the long term infrastructure nature of these assets.

On innovation, India needs to invest in the full stack. Not just AI applications, not just software, but hardware design, energy efficient compute architecture, and advanced cooling systems. We have the engineering capability to do this. Organisations like ISRO and DRDO have demonstrated that. The challenge is channelling that capability into commercial and industrial infrastructure at scale.

My belief is that within this decade, India has a genuine opportunity to not just consume AI infrastructure but to be a meaningful exporter of it. That requires treating compute infrastructure with the same strategic seriousness that we give to roads, ports, and power grids. And it requires the courage to genuinely level the playing field for Indian manufacturers, not just in policy documents but in every single RFP that goes out. The moment that happens at a national level, everything else will follow.