India AI Impact Summit Insights
All sessions

From Access to Impact: Enabling SMEs in the AI Economy| AI Impact Summit 2026

Contents

Executive Summary

This panel discussion, held at the AI Impact Summit 2026 and organized by the India Electronics and Semiconductor Association (IESA), examines how AI can be deployed at scale to benefit small and medium enterprises (SMEs) and startups in India. The core argument is that India's AI opportunity lies not in competing with global cloud-centric data centers (already dominated by tech giants) but in building optimized, edge-based, device-centric AI solutions tailored to local manufacturing, surveillance, robotics, and industrial applications—while addressing ethical concerns and data sovereignty.

Key Takeaways

  1. India's AI competitive edge is not in data centers but in building optimized, privacy-preserving, edge-deployable AI solutions for manufacturing, surveillance, and robotics—sectors not yet captured by global giants.

  2. The real blocker for SME AI adoption is not technology or capital but trust, awareness, and affordability. Government schemes exist; the missing link is dissemination and integration support through bodies like IESA.

  3. Ethical AI cannot be an afterthought or a compliance checkbox—it must be embedded in design from day one, account for regional/sector context, and include continuous monitoring throughout a system's lifecycle.

  4. Stop judging deep tech companies by SaaS metrics (1-year sales cycles, rapid pivots). Deep tech needs 9–18 months for R&D and long-term trust-building; payoffs are substantial but timescales differ.

  5. The transition from prototype to deployment at scale is the hardest problem. India can lead by becoming the world's leader in manufacturing + AI integration, turning humans and machines into high-productivity teams.

Key Topics Covered

  • Edge AI vs. Cloud-Centric AI: The distinction between centralized cloud deployments and optimized edge/on-premise solutions for SMEs
  • India's AI Opportunity: Positioning India as a product nation rather than a service nation through domain-specific AI optimization
  • Deep Tech Startup Challenges: Why Indian deep tech companies struggle to scale globally despite talent and capital availability
  • AI Deployment in Manufacturing: Use cases and ROI in inventory management, predictive maintenance, supply chain optimization, and quality monitoring
  • Data Sovereignty & Privacy: Security concerns preventing SME adoption of cloud-based AI
  • Ethical AI & Responsible Deployment: Safety-first design, bias detection, trustworthiness certification, and ongoing monitoring
  • Government Infrastructure & Support: AI grants, shared compute models, data aggregation (AI Kosh), GPU access, and awareness challenges
  • Liability and Accountability: Questions around who bears responsibility when AI systems fail
  • Skills Gap & Awareness: The need for large-scale education about AI use cases in manufacturing and traditional industries

Key Points & Insights

  1. The "Uncomfortable Truth": No Indian deep tech company is currently selling at global scale in semiconductors or AI models, despite abundant talent and venture capital. The issue is not resources but incorrect expectations—deep tech requires 9–18 months for chip development, not the 1-year SaaS paradigm.

  2. India's Untapped Opportunity is Optimized, Edge AI: Cloud data centers are already captured by global players (AWS, Google, Azure). India can differentiate by building cost-effective, privacy-preserving, optimized AI models deployable on edge devices without sending sensitive data to the cloud.

  3. AI Adoption is Limited by Privacy Concerns: SMEs and manufacturing units resist cloud-based AI because they fear data loss, sovereignty violations, and high bandwidth/cloud costs. Edge-centric approaches unlock adoption by processing data locally and sending only metadata.

  4. Manufacturing AI ROI is Proven but Underexploited: Concrete high-ROI use cases include inventory optimization, predictive machine maintenance (preventing downtime), supply chain logistics, and quality monitoring. Manufacturing is expected to grow AI adoption at 40% CAGGR until 2030, yet <5% penetration exists today.

  5. The Cost of Intelligence is Approaching Zero: Foundation models (LLMs, robotics models) have democratized access to world-class algorithms. The new differentiator is not algorithms but trust, integration, and domain-specific optimization—which India can lead.

  6. Ethical AI Requires Concept-Level Design, Not Post-Hoc Patching: Safety, security, and trustworthiness must be built in from the start, not bolted on afterward. Ethics is context-dependent (regional laws, sector regulations, cultural norms differ), requiring governance frameworks beyond one-size-fits-all solutions.

  7. Government Schemes Exist But Lack Awareness: India has GPU grants, the AI Kosh (anonymized shared data platform), and subsidies for SMEs. The barrier is not policy but dissemination—awareness campaigns and integration platforms (like IESA) are critical.

  8. Liability Frameworks Are Still Evolving: Who is liable when AI fails remains unsolved (human-in-the-loop models are emerging as practical interim solutions). Automotive and healthcare cases suggest liability will rest with the human/company deploying the system, but this needs formal legal clarity.

  9. Deep Tech Requires 5–10 Year Vision: Global leaders identified problems 5–10 years ahead and invested accordingly. Indian startups must look beyond exit timelines and identify structural bottlenecks (e.g., photonics companies saw data bottlenecks coming).

  10. SMEs Face a "Capability Trap": Many family-owned businesses on legacy IT infrastructure don't know what AI can do for them. Overcoming this requires cluster-by-cluster deployment, skilling programs, and affordable infrastructure upgrades (not enterprise SaaS pricing).

Notable Quotes or Statements

  • Jotis Sundrai (Net Semi): "AI is right now being utilized primarily as a good search engine and an analytics tool, but it can solve a lot of problems like trash management, surveillance without losing data."

  • Anand (Applied Ventures): "The fundamental unit that is stopping us is trust—not algorithms, not GPUs. The cost of intelligence is going to go to zero. The question is how do you scale and build world-class companies on top of it."

  • Sep Gupta (Qualcomm/Alpha Wave): "The manufacturing sector is going to grow AI deployment at 40% CAGR until 2030. That's tremendous opportunity. But how do we get small and medium scale industries to adopt it?"

  • Pri Yadav (NXP Semiconductors): "Trustworthiness has to be there from concept to end of life. What is true today might not be true tomorrow. There should be a check and balance throughout the lifecycle of the AI deployment."

  • Naven (Marvel Semiconductor, Vice Chair IESA): "We've moved from talking about models, data, and compute to the real question: How do we move from prototype and benchmark into actual deployment and use cases? Especially in India, how do we leverage for startups and SMEs?"

Speakers & Organizations Mentioned

Panelists:

  • Naven — Marvel Semiconductor, Vice Chairperson IESA (moderator); focus on fabless chip design, data centers, AI networking
  • Anand — Applied Ventures (corporate venture arm of Applied Materials); invests in deep tech globally
  • Sep Gupta — Vice President Central R&D, Qualcomm (formerly Alpha Wave); semiconductor industry veteran (27 years)
  • Jotis Sundrai — CEO & Co-founder, Net Semi Private Limited; edge AI chips for surveillance, robotics, drones
  • Pri Yadav — Head of Innovation Ecosystem, NXP Semiconductors (India); automotive semiconductor focus

Organizations & Government Bodies:

  • IESA (India Electronics and Semiconductor Association) — Industry voice representing government, academia, industry
  • Applied Materials — Semiconductor equipment maker (~$250–300B market cap)
  • Qualcomm — Global semiconductor and IP leader
  • NXP Semiconductors — Automotive and IoT semiconductor leader
  • Net Semi — Indian edge AI chip startup
  • Marvel Semiconductor — Indian fabless design company
  • VVDN — Indian EMS and design services company (cited as a success scaling globally)
  • Government of India — AI Kosh (shared data aggregation platform), GPU grants, subsidies for SMEs
  • Times of India — Source of AI adoption statistics by profession

Technical Concepts & Resources

  • Edge AI / On-Device AI: Localized processing to avoid cloud latency, data sovereignty loss, and bandwidth costs
  • Optimized AI / Domain-Specific AI: Models tailored to specific industries (manufacturing, surveillance) rather than general-purpose LLMs
  • Foundation Models: Large pretrained models (LLMs like Claude, GPT; robotics models) reducing algorithm R&D barriers
  • AI Kosh: Government of India's anonymized, shared data platform for SME training
  • Hybrid Compute Model: Sensitive data processed on-premise; metadata sent to cloud
  • Predictive Maintenance: Using energy spikes, vibration data, and machine learning to predict equipment failure before downtime
  • Functional Safety: Engineering discipline ensuring systems fail predictably and safely (critical in automotive, healthcare)
  • Data Poisoning / Biased Data: Training data contamination causing AI systems to perpetuate or amplify societal biases
  • Physical Simulation / Physics-Informed AI: Chip design, plasma simulations, materials discovery accelerated 1000x via AI
  • Contract Review / Legal AI: AI-assisted document analysis reducing turnaround time from weeks to 15 minutes
  • EMS (Electronics Manufacturing Services): Companies providing manufacturing and design support
  • RTL (Register Transfer Level): Hardware design abstraction layer
  • DV (Design Verification): Chip validation process

Implicit Opportunities & Challenges

Opportunities:

  • Cluster-by-cluster manufacturing adoption model (thousands of Indian factories as addressable market)
  • Job creation in AI integration, domain expertise, not just software development
  • Manufacturing leadership through human + AI teams (untapped globally)

Challenges:

  • Legacy IT infrastructure in SMEs (fragmented data, outdated systems)
  • Affordability barriers (AWS/GCP costs prohibitive for small factories)
  • Lack of accessible standards and governance frameworks for ethical AI
  • Liability ambiguity (who is responsible when AI fails?)
  • Brain drain and skill gaps in Indian deep tech compared to SaaS

Note: This transcript reflects a live Q&A panel with overlapping dialogue and some audio transcription artifacts. Quotes and attributions have been preserved as they appear but should be cross-referenced with the original video for full accuracy.