Qualcomm Inc.

01/17/2026 | News release | Distributed by Public on 01/16/2026 16:53

Edge intelligence: 10 trends driving startup success worldwide



What you should know:

  • In 2025, Qualcomm enabled over 60 startups across the Americas, Africa, Middle East, India and Asia-Pacific to deploy AI solutions at the edge. Collectively, Qualcomm incubated startups have filed 1,350+ patents and 25,000+ inventors have received training in IP rights - demonstrating the massive scale of innovation and edge intelligence adoption worldwide.
  • Startups are complementing cloud dependent AI with edge deployment for ultra low latency inferencing, on-device processing and data sovereignty, enabling applications in robotics, healthcare and industrial automation while maintaining regulatory compliance and user privacy.
  • The future belongs to context-aware AI systems that orchestrate predictions and actions autonomously, while no-code platforms are empowering SMEs and non-experts to rapidly deploy sophisticated edge AI solutions without deep technical expertise.


In 2025 Qualcomm Government Affairs' ecosystem development team enabled over 60 startups across the Americas, Africa, the Middle East, India and Asia-Pacific, to bring wireless connectivity, IoT and edge AI-based products to market, scale business, and secure IP rights. Ten key technological trends emerged from their edge AI implementations. Let's look at what's shaping the future.

1. Connecting bits and atoms

Edge AI transcends the realm of digital assistants, by embedding intelligence adjacent to the physical world. By orchestrating seamless interactions among machines, sensors and humans, it transforms business workflows into systems that are not only precise and auditable but resilient against failure, making it a necessity in robotics, industrial IoT and transportation.

Industrial IoT

Transport

2. Reimagined workflows

Reimagined workflows in edge AI are defined by their relentless generation and assimilation of real-time data, demanding not only technical acumen but deep domain expertise to manage complex inter-dependencies and regulatory constraints. The true innovation lies in their capacity to unlock capabilities previously out of reach, through combining perception (sensing) with on-device cognition and agency, whether in clinical diagnostics, industrial automation or adaptive learning environments.

Industrial IoT

Healthcare

Pharmacology

Healthcare and Pharmacology

Education and Training

3. Real-time intelligence

Edge AI enables ultra-low latency, high-volume inferencing and dynamic actions on-device. Reliance on cloud-AI introduces round-trip delays that are incompatible with real-time needs in industries such as video analytics, industrial automation and autonomous systems.

Media

Retail and Media

Industrial IoT

Agriculture

4. Agentic AI systems

Agentic AI systems orchestrate predictions and generative outputs to drive context-aware actions, with each step governed by operational constraints and checkpoints. This architecture enables flexibility in adapting to variability in inputs and operational conditions, while maintaining auditability and reliability. Agentic orchestration is now central to edge AI applications where every action must be traceable and robust, especially in environments demanding both adaptability and operational control. While several of our startups (mentioned elsewhere) have implemented Agentic AI systems, two that stand out are:

Industrial IoT

5. Enabling tech for AI

Foundational innovation underpins edge AI. Startups are developing custom silicon and integration tools, each addressing distinct challenges in on-device AI deployment. These offerings complement Qualcomm's Edge Impulse and AI Hub suite of services to augment and automate workflows through rapid data collection, AI-enabled analysis and enhanced decision-making.

Industrial IoT

  • Manovega (India) were advised on custom ASIC for custom RISC-V SoC purpose-built for edge AI processing.
  • Netrasemi (India) were also advised on cusom ASIC to enable power-efficient Edge AI SoCs for IoT solutions.

6. Democratizing AI access through no-code AI

No-code platforms are lowering barriers for small and medium enterprises and non-experts to deploy AI solutions. By enabling rapid prototyping and domain-specific automation without deep technical expertise, these tools accelerate adoption of edge-AI across industries, making advanced capabilities accessible to a broader range of users.

Healthcare

Enterprise

  • MoBagel (Taiwan) used the Dragonwing AI On-Prem Appliance for no-code AI agent platform with generative BI and predictive analytics.
  • Tricuss (Taiwan) were advised on multi-device innovation to enable a no-code AI agent builder with a proprietary data asset platform.

Retail and Media

Legal and Compliance

  • iGotAI (Vietnam) were advised on multi-device innovation to enable no-code audit automation with secure local deployment and full control.

Education and Training

Industrial IoT

  • Orangecat (India) used the Snapdragon X Elite Platform to enable an agentic AI coding platform for developers and enterprises with voice-activated website building to support Indian languages.

7. Privacy and data sovereignty

Edge AI startups are embedding federated learning, on-device inference and secure workflows to keep sensitive data local. This approach enables personalization and regulatory compliance while minimizing exposure to external risks, making privacy and data sovereignty foundational for deployment in regulated and sensitive domains.

Customer Operations

Legal and Compliance

Industrial IoT

8. Use of country-specific and sovereign AI models

Edge deployments increasingly rely on sovereign or locally trained AI models to address linguistic, cultural and regulatory requirements. By tailoring solutions to local contexts, startups ensure compliance and relevance in sensitive domains such as healthcare, legal and education, strengthening trust and adoption.

Legal and Compliance

Education and Training

Similarly, aforementioned startups Mobisense, PixConvey and Agile Loop are using Saudi Arabia's Allam model, Raxa supports several Indian languages, while SqueezeBits has also used South Korea's ExaOne from LG.

9. AI for environmental resilience

Edge AI is advancing sustainability by enabling real-time monitoring of ecosystems, optimizing resource use and mitigating climate risks without reliance on cloud connectivity. Startups are deploying solutions for agriculture, climate prediction and environmental management, supporting resilience and efficiency in diverse settings.

Agriculture

Climate and Environment

Education and Training

10. Building for AI safety and trust

Edge AI startups are prioritizing safety and trust by embedding explainability, ethical safeguards and reliability checks into their solutions. These measures are essential for responsible deployment in sensitive contexts, ensuring that AI systems operate transparently and meet high standards for accountability.

Legal and Compliance

Customer Operations

IP generation

In 2025, we achieved two major intellectual property milestones: over 25,000 inventors worldwide completed training in IP rights through free, localized online courses and our equity-free startup incubation programs enabled supported startups to collectively file more than 1,350 domestic and international patents. This marks a substantial share of deep-tech patent activity in their respective countries.

Particularly in the U.S., The Inventor's Patent Academy (TIPA) reached 3,800 learners across a dozen states, embedding IP education into entrepreneurship and workforce curricula at major institutions (including SDSU, UCSD, CSU San Marcos, Houston Community College and Georgia Tech) and national conferences, establishing itself as a trusted resource for building patent skills essential to U.S. innovation and advanced manufacturing.

Looking ahead to 2026

Designing edge AI systems is a discipline apart - requiring precise engineering under tight memory and processor bandwidth, across heterogeneous hardware like CPUs, GPUs, DSPs and NPUs. Qualcomm and Arduino platforms, and associated developer tools are crucial to practicing this genre of engineering design. Success depends on balancing model compression, token throughput and accuracy, while minimizing hallucinations and "mispredictions" through robust checkpoints. Integrating new sensor and operational data into model updates, and using workflow feedback for continuous improvement, is essential. The next wave of innovation will be shaped by those who master this convergence of physical and digital intelligence, building resilient systems where real-world constraints are not obstacles, but vectors for differentiation and progress.

Learn More


Qualcomm Inc. published this content on January 17, 2026, and is solely responsible for the information contained herein. Distributed via Public Technologies (PUBT), unedited and unaltered, on January 16, 2026 at 22:53 UTC. If you believe the information included in the content is inaccurate or outdated and requires editing or removal, please contact us at [email protected]