AI Infrastructure · Bharat

कल्पना Building India's AI compute layer.

Purpose-built GPU infrastructure and data centres for production AI in Bharat. Owned on Indian soil. Billed in rupees. Built for the workloads India is already producing.

Partner with us → See the problem ↓
Without Kalpana. India's data flows out. Dollars follow.
Tap the dot on India
The Problem

India generates 20% of the world's data.
It owns 3% of the compute.

India
20%
1.4 billion people generate it.
WhatsAppUPIedtechfintech
Your data
Billed in $ $ $ $
Foreign compute
97%
The world's GPUs sit elsewhere. We rent.
USAChinaEU
Today
We rent our own intelligence.

Every prompt, every model, every product Indians ship is metered, throttled, and priced in foreign currency.

If we don't build
Indian founders queue while the West iterates.

Models train abroad on data extracted from us, then come back as products billed in dollars. Iteration speed dies.

The widening gap
We have everything but the machines.

1.4 B users · 20% of the world's data · 3% of the compute. The asymmetry compounds every quarter the West builds and we wait.

The future,
if we build →
Bharat builds its own LLMs on its own GPUs. Founders iterate in hours, not weeks. Health, defence, fintech, and language data stays on Indian soil under Indian law. Compute is billed in rupees, not rationed in dollars. The intelligence India produces, India owns.

America already runs more AI power
than India plans to have by 2030.

Running today · EoY 2025 Planned addition by 2030
USA
45 GW
+55 GW
China
32 GW
+28 GW
India
1
+4
India · today & tomorrow 5 GW total · the US runs ~9× that today
Installed data-centre power capacity · gigawatts
India's 2030 plan is less than 1/9 of where the US runs today.
5 GW by 2030 is less than a twelfth of China's running stock today (60 GW). We are not behind by a quarter — we are behind by a decade. The gap widens every quarter the West builds and we wait.
A collective vision

Ek Kalpana.

एक कल्पना

One vision, held by many. India needs dozens of companies building compute on Indian soil. Kalpana is one of them — and we won't be the last. This belongs to all of us.

→ Without it
India keeps renting its own intelligence.
  • Models trained abroad on data extracted from us, then sold back in dollars.
  • Founders queue weeks for GPU time while Silicon Valley iterates daily.
  • National data — health, finance, language — sits on foreign servers.
  • The next decade of AI is built somewhere else, by someone else.
→ With it
Bharat builds AI on its own machines.
  • Production-grade compute, owned on Indian soil, billed in rupees.
  • Founders ship in hours, not weeks. Iteration speed becomes a moat.
  • Data stays here. Sovereignty, privacy, and law align under one flag.
  • India is no longer a data colony. It is a compute power.
The way we build this

India needs many builders. We are one.

Kalpana operates the integrated stack — land, building, power, cooling, GPUs, cloud — but the country needs more of us. More land secured, more buildings raised, more racks owned by Indian hands. We're inviting partners, operators, capital, and engineers to build alongside us.

↓ How we build it
The Stack

Most AI clouds rent four of these five layers.
We own all five.

A single facility — land to GPU to API. Click any layer to step inside.

Layer 01 · Building

Land & Building

We secure the land and design the building for AI from the slab up.

The Industry View

In the West, compute is the bet of the decade.
In India, it's rationed.

The companies leading AI are spending whatever it takes to build more of it. The founders building in India are queueing for what little there is.

Spending into it
Google
The risk of underinvesting is dramatically greater than the risk of over-investing.
Sundar PichaiCEO · Google · Q2 2024 earnings call
~ 1,000,000 GPUs · $75B+ annual capex run-rate
Spending into it
Meta
The very worst case would be that we have just prebuilt for a couple of years.
Mark ZuckerbergCEO · Meta · Q3 2025 earnings call
~ 600,000 GPUs on order by end of 2024
GPUs · same scale · same dot
Each dot ≈ 5,000 GPUs
Google
Hyperscaler · USA
~ 1,000,000
Meta
Hyperscaler · USA
~ 600,000
India
All of it · 1.4 B people
~ 100,000

Two American companies operate over 16× more AI compute than is available across all of India. Indian founders are not behind on talent — they are behind on access to the machines.

The companies leading AI are spending whatever it takes.
The founders building in India are still waiting for their turn.

The Team

People who have already built things
that cannot be allowed to fail.

Founder · Photo

Ishita Srivastava

Founder & CEO

Electronics Engineer, BITS Pilani. Raised ₹650 Cr for deep-tech ventures. Former partner at Elixir Capital, Abu Dhabi.

She started Kalpana because India's intelligence cannot keep running on machines it does not own — and because no single company can solve this alone.

Engineering · Defence-tech

Israeli Defence-Tech Veteran

Name & bio to follow
Engineering · Hyperscale

USA Data-Centre VP of Engineering

Name & bio to follow
Legal · Corporate

UC Berkeley-Educated Legal Counsel

Name & bio to follow
Where this team has built before. Logos to follow.
Join our Kalpana

India will build this compute.
The question is who builds it.

partner@kalpanacloud.com →
Building compute for a brighter, smarter Bharat.
एक उज्ज्वल, चतुर भारत के लिए कंप्यूट।