Krutrim: India’s First AI Unicorn Building the Nation’s GPU Cloud Infrastructure

From Ola’s Founder to AI Pioneer—The Krutrim Story

When Bhavish Aggarwal, the founder of Ola Cabs, announced in late 2023 that he was diving into artificial intelligence, many dismissed it as another billionaire’s side project. Fast forward to 2026, and Krutrim—India’s first AI unicorn (achieved in January 2024)—has pivoted from ambitious foundation model dreams to becoming the backbone of India’s AI infrastructure, providing GPU cloud services that power Indian startups, enterprises, and researchers at a fraction of global prices.

This is the story of how Krutrim transformed from a full-stack AI dreamer into India’s most pragmatic AI infrastructure play—and why that matters for every Indian developer, startup, and enterprise building with AI.

What is Krutrim?

Krutrim is an AI company founded by Bhavish Aggarwal (Ola founder) in 2023, headquartered in Bengaluru. It became India’s first AI unicorn in January 2024, just months after launch.

Originally, Krutrim aimed to build everything:

• Foundation LLMs (Krutrim-1, trained on 2 trillion tokens across 22 Indian languages)
• Custom AI chips (Bodhi, Sarv, Ojas chip line)
• Consumer AI assistant (Kruti app)
• GPU cloud infrastructure

By 2025-26, Krutrim made a strategic pivot:

• Paused foundation model development (including planned Krutrim-3)
• Paused in-house chip design work
• Discontinued Kruti consumer app (pulled from app stores)
• Focused entirely on AI cloud infrastructure for enterprises

This pivot wasn’t failure—it was pragmatism. Krutrim found its competitive advantage: providing affordable, India-hosted GPU compute for AI training and inference.

Krutrim Cloud: India’s Answer to AWS, Azure, and Google Cloud for AI

💻 What Krutrim Cloud Offers

Krutrim Cloud is a GPU/CPU cloud infrastructure provider specifically designed for AI workloads:

GPU Instances:

• A100 80GB: ₹189/hour (96 GB RAM, 24 vCPU)
• H100: ₹213/hour (200 GB RAM, 24 vCPU)
• Multiple GPU configurations for scaling

CPU Instances:

• Starting from ₹3/hour (1 vCPU, 4 GB RAM)
• Up to 2,048 GB RAM configurations

Services:

• AI Pods (dedicated GPU clusters)
• Storage solutions
• Model hosting services
• Enterprise support

🎯 Who Uses Krutrim Cloud?

• Indian Startups: Building AI products without paying AWS/Azure premiums
• Enterprises: Running AI workloads with data sovereignty (data stays in India)
• Researchers: Training models on India-hosted infrastructure
• Developers: Experimenting with AI at affordable hourly rates

TechCrunch reported (May 2026): 25+ enterprise customers and approximately ₹300 crore FY26 revenue—making Krutrim profitable in FY26, a rare feat for Indian AI startups.

Why Krutrim Matters for India’s AI Future

1. Affordable GPU Access for Indian Builders

Global GPU cloud pricing (AWS, Azure, Google Cloud):

• A100 80GB: $3-4/hour (₹250-330/hour)
• H100: $4-5/hour (₹330-415/hour)

Krutrim’s pricing:

• A100 80GB: ₹189/hour (40-45% cheaper)
• H100: ₹213/hour (35-50% cheaper)

Impact: Indian startups can train models and run inference at nearly half the cost of global providers, extending their runway and making AI experimentation accessible to smaller teams.

2. Data Sovereignty & Compliance

With India’s data localization laws and enterprise concerns about sending sensitive data abroad, Krutrim offers:

• India-hosted infrastructure (data never leaves Indian soil)
• Compliance with RBI, IRDAI, and government data regulations
• Lower latency for Indian users (servers in Indian data centers)

3. Supporting IndiaAI Mission

The Government of India’s ₹10,372 crore IndiaAI Mission aims to provide 100,000 publicly accessible GPUs by December 2026 at ₹115-150 per GPU-hour (42% below market rates). Krutrim’s infrastructure complements this by:

• Providing additional GPU capacity beyond government allocations
• Offering enterprise-grade support and SLAs
• Serving customers who need dedicated, private infrastructure

4. First-Mover Advantage in Indian AI Cloud

While global hyperscalers (AWS, Azure, Google) dominate Indian cloud, Krutrim is the first India-founded, India-focused AI cloud provider. This matters because:

• Local support teams understand Indian business needs
• Pricing in INR without currency fluctuation risks
• Faster onboarding for Indian startups and enterprises
• Alignment with “Make in India” and digital sovereignty goals

Krutrim vs Sarvam AI: Two Paths to Indian AI Sovereignty

Both Krutrim and Sarvam AI are Bengaluru-based AI unicorns, but they’ve taken different paths:

Tip for mobile users: Swipe horizontally across the comparison table to view all columns.

The Reality: India needs both approaches.

• Krutrim provides the infrastructure (GPUs, compute)
• Sarvam provides the models (LLMs, speech, vision)

Together, they form the foundation of India’s sovereign AI stack—compute + models, both built and hosted in India.

The Pivot That Saved Krutrim

In May 2026, TechCrunch reported that Krutrim:

• Laid off 200+ employees across multiple rounds
• Pulled Kruti app from app stores
• Paused in-house chip design (Bodhi, Sarv, Ojas)
• Paused foundation model development (Krutrim-3)

This looked like retreat—but it was strategic focus.

Why the Pivot Made Sense:

1. Foundation Models Are Capital-Intensive: Training 100B+ parameter models costs $50-100M+. Sarvam raised $234M for this. Krutrim chose a different path.
2. Chip Design Is Hard: Even giants like Google (TPU) and Amazon (Trainium) took years to ship competitive chips. Krutrim’s Bodhi chip (announced for 2026) hasn’t shipped as of September 2026.
3. Consumer AI Assistants Are Crowded: Kruti competed with ChatGPT, Gemini, Copilot, and dozens of Indian alternatives. Differentiation was tough.
4. GPU Cloud Is a Real Business: With AI boom, demand for GPUs exceeds supply. Krutrim’s 2,000 NVIDIA Blackwell GPUs (scaling to 10,000) generate recurring revenue with clear unit economics.

Result: Krutrim became profitable in FY26 on ₹300 crore revenue—a rarity in Indian AI.

Real-World Use Cases: Who’s Building on Krutrim?

🚀 Indian AI Startups

Startups training custom models or running inference on Krutrim:

• Avoid AWS/Azure pricing premiums
• Keep data in India for compliance
• Get local support in Indian time zones

🏢 Enterprises

Indian enterprises using Krutrim for:

• Banking: Running fraud detection models on-premise
• Healthcare: Processing patient data with AI (data never leaves India)
• E-commerce: Product recommendation engines
• Manufacturing: Quality control with computer vision

🎓 Research & Academia

Universities and research labs:

• Training NLP models for Indian languages
• Running large-scale experiments
• Collaborating with industry on AI projects

🛠️ Developers & Hobbyists

Individual developers:

• Experimenting with AI at ₹3/hour CPU or ₹189/hour GPU
• Building side projects without large cloud bills
• Learning AI/ML with affordable access to H100/A100

Challenges Ahead: Can Krutrim Compete with Global Giants?

⚠️ Competition from AWS, Azure, Google Cloud

Global hyperscalers have:

• Deeper pockets for GPU procurement
• Established enterprise relationships
• Global infrastructure and reliability track records
• Broader service portfolios (not just GPU)

⚠️ GPU Supply Constraints

NVIDIA GPUs (A100, H100, Blackwell) are in global short supply. Krutrim’s ability to scale depends on:

• Securing GPU allocations from NVIDIA
• Competing with hyperscalers for limited supply
• Managing capex for GPU purchases

⚠️ Talent Retention

After 200+ layoffs in 2025-26, Krutrim needs to:

• Rebuild engineering teams
• Retain top AI/infrastructure talent
• Compete with global AI labs for researchers

⚠️ Proving Long-Term Viability

Krutrim’s pivot to cloud infrastructure is pragmatic, but investors and customers need to see:

• Sustained revenue growth beyond FY26
• Customer retention and expansion
• Clear path to scaling beyond 10,000 GPUs

Why Krutrim Matters for Every Indian Builder

This isn’t just about one company—it’s about India’s AI infrastructure sovereignty:

• ✅ Affordable Access: Indian startups can experiment with AI without burning cash on AWS bills
• ✅ Data Sovereignty: Sensitive data (healthcare, finance, government) stays in India
• ✅ Local Support: Indian time zones, Indian support teams, INR pricing
• ✅ Ecosystem Growth: Krutrim + Sarvam + IndiaAI Mission = complete sovereign AI stack
• ✅ Global Competitiveness: Indian AI products can be built cost-competitively with global rivals

💡 Did You Know?

Interesting Facts About Krutrim

• ✓ India’s First AI Unicorn: Krutrim achieved unicorn status in January 2024, just months after founding—making it the fastest Indian AI startup to reach $1 billion valuation at the time.
• ✓ ₹300 Crore Revenue, Profitable in FY26: Unlike most AI startups burning cash, Krutrim reported profitability in FY26 on approximately ₹300 crore revenue—a rare feat in capital-intensive AI infrastructure.
• ✓ 2,000 NVIDIA Blackwell GPUs Deployed: Krutrim has already deployed around 2,000 of NVIDIA’s latest Blackwell GPUs, with plans to scale to 10,000 GPUs—making it one of the largest AI compute clusters in India.
• ✓ 40-50% Cheaper Than AWS/Azure: Krutrim’s A100 pricing at ₹189/hour vs AWS’s ₹330/hour and H100 at ₹213/hour vs AWS’s ₹415/hour means Indian startups can save nearly half their cloud bills.
• ✓ 25+ Enterprise Customers: Despite pivoting away from consumer products, Krutrim serves 25+ enterprise customers—including banks, healthcare providers, e-commerce platforms, and government agencies.
• ✓ Pivoted from 4 Verticals to 1: Krutrim originally pursued foundation models, custom chips, consumer app (Kruti), AND cloud infrastructure. By 2026, it focused entirely on cloud infrastructure—a pragmatic move that led to profitability.
• ✓ Named After Sanskrit Word: “Krutrim” (कृत्रिम) means “artificial” in Sanskrit—a nod to both AI (artificial intelligence) and India’s linguistic heritage.
• ✓ Ola Connection: While Krutrim is a separate company, it’s backed by Ola’s ecosystem and Bhavish Aggarwal’s vision. Ola Electric vehicles will reportedly use Krutrim’s AI for autonomous driving features.
• ✓ Bodhi Chip Ambitions: Krutrim announced three custom chips—Bodhi (AI accelerator), Sarv (cloud CPU), and Ojas (edge AI for Ola vehicles)—though chip development has been paused as of 2026 to focus on cloud infrastructure.
• ✓ Krutri AI Assistant: The Kruti app (discontinued in 2026) was India’s first homegrown AI assistant, supporting 20 Indian languages natively. When asked to name India’s best electric scooter, Kruti controversially picked Ather 450X over Ola’s own S1 Pro—showing unbiased AI responses.

The Road Ahead: Krutrim’s Next Chapter

As of September 2026, Krutrim’s roadmap includes:

Short-Term (2026-27)

• Scale GPU fleet to 10,000 GPUs
• Expand enterprise customer base beyond 25
• Improve cloud platform features (orchestration, monitoring, DevOps tools)
• Achieve sustained profitability beyond FY26

Medium-Term (2027-28)

• Potentially restart chip development if market conditions improve
• Explore partnerships with IndiaAI Mission for government deployments
• Expand to additional Indian data centers for redundancy

Long-Term (2028+)

• Become India’s default AI infrastructure provider
• Compete with AWS/Azure for large enterprise AI workloads
• Potentially restart foundation model work if compute costs decrease

Final Thoughts: Infrastructure Is the Unsung Hero of AI

While Sarvam AI grabs headlines with open-source LLMs and voice agents, Krutrim’s quieter pivot to GPU cloud infrastructure may be equally important for India’s AI future.

Why?

Because every AI model—whether Sarvam’s 105B LLM, a startup’s custom NLP model, or an enterprise’s fraud detection system—needs compute to run. Without affordable, India-hosted GPU infrastructure, Indian AI builders remain dependent on foreign cloud providers, paying premiums and sending data abroad.

Krutrim’s bet is simple: India will generate massive AI compute demand, and it’s better served by an India-founded, India-focused provider than by AWS, Azure, or Google Cloud.

Whether that bet pays off depends on execution, GPU supply, and competition. But one thing is clear: India needs companies like Krutrim to build the infrastructure layer of its sovereign AI stack.

The AI revolution needs compute. And Krutrim is building India’s compute.

Want to Try Krutrim Cloud?

• Visit Krutrim’s website for GPU/CPU pricing
• Explore enterprise plans for dedicated infrastructure
• Check out documentation for developers
• Follow Krutrim’s updates for new features and data center expansions

Made in India. Built for India. Powered by India. 🇮🇳