For the past few years, the artificial intelligence industry has been obsessed with building bigger models.
More parameters. More computing power. More data centres. More sophisticated reasoning. More powerful AI agents. More Next AI Revolution.
And understandably so.
We have witnessed extraordinary progress in what AI systems can accomplish.
But recently, I have become interested in a different question.
What if the next important AI breakthrough is not about making intelligence bigger, but about bringing intelligence closer to where it is actually needed?
On October 7, 2026, Microsoft and NVIDIA shared a vision for a new generation of personal computers designed to run increasingly powerful AI models locally.
Microsoft’s new Surface Laptop Ultra, for example, supports configurations with up to 128 GB of unified memory and is designed to run AI models exceeding 120 billion parameters locally.
That caught my attention.
Not because I need another powerful laptop.
But because I think it represents something much bigger than a hardware announcement.
We may be entering a period when the location of artificial intelligence becomes almost as important as its capability.
And I believe that could change several industries.
We Have Built AI Around a Particular Assumption
Think about how most people experience generative AI today.
They open an application, write a prompt, and receive an answer.
Behind that seemingly simple interaction, the request is often sent to powerful computing infrastructure located somewhere else.
The cloud processes the request and sends the response back.
This architecture has made sophisticated AI accessible to millions of people without requiring them to own expensive hardware.
I consider that an extraordinary achievement.
But somewhere along the way, we began treating cloud-based intelligence as the default architecture for everything.
And that is where I think we need to reconsider our assumptions.
Does every intelligent decision need to travel to a distant server?
Does every device need an internet connection to understand its surroundings?
Should every organization send sensitive information outside its own environment to benefit from AI?
And what happens when connectivity becomes unreliable?
These are not questions about which AI company has the best model.
They are questions about how intelligent systems should actually be engineered.
My Interest in This Is Personal
Before becoming an entrepreneur and working extensively in artificial intelligence, I studied electronics and communication engineering.
One of my early engineering projects involved controlling a robot through a mobile device.
That project belonged to a very different technological era.
We were thinking about control, communication, distance, and how machines could respond to instructions.
Today, the possibilities are dramatically different.
We are no longer limited to telling a machine exactly what to do.
We can develop systems capable of interpreting images, recognizing patterns, evaluating their environments, and recommending or taking actions.
Yet one fundamental engineering question remains unchanged:
What should happen locally, and what should depend on communication with another system?
I think this question will become increasingly important as AI moves beyond computer screens and enters factories, agricultural fields, hospitals, vehicles, and robots.
Because intelligence in the physical world operates under very different conditions from intelligence inside a chatbot.
A Robot Cannot Always Wait for The Cloud
Imagine an autonomous machine operating in a factory.
It detects an unexpected obstacle.
Should it capture the information, send it to a distant server, wait for an AI model to process the situation, receive instructions, and then decide how to respond?
For some tasks, that architecture might be perfectly acceptable.
For time-critical tasks, it may introduce unnecessary delay and dependence on connectivity.
A safety-critical machine needs appropriately designed local control and reliable fail-safe mechanisms.
Not every decision should depend on a network connection.
The same principle applies to many other environments.
Consider agricultural technology.
A smart agricultural system could analyze sensor readings, soil conditions, and images to identify possible problems.
Some analysis can happen locally, while more complex forecasting and historical comparisons can happen in the cloud.
Or consider intelligent crowd management.
An event-monitoring system may need to identify unusual congestion patterns as quickly as possible.
Local processing can reduce the need to continuously transmit every video frame elsewhere, although reliable detection, communication, and human oversight remain essential.
At ReThynk AI, as we explore opportunities in areas such as smart agriculture and crowd safety, this distinction interests me more and more.
The intelligence that works best in a data centre may not be the intelligence that works best at the point of action.
And I think some of the most interesting AI innovations will come from understanding that difference.
I personally believe AI is making Coding Easier and Learning Harder
The AI Industry is Already Moving In This Direction
This is not merely a theoretical idea.
Microsoft is now promoting what it calls hybrid intelligence, where AI workloads can be routed between local hardware and cloud computing.
Google provides on-device AI capabilities through its AI Edge ecosystem, including tools for running language models and other machine-learning applications on devices.
NVIDIA is also investing heavily in edge computing for robotics through platforms such as Jetson Thor.
These developments point toward an important architectural shift.
We are moving from a world where intelligence is primarily accessed through remote computing infrastructure toward one where intelligence can also reside inside the devices and machines we use.
I don’t believe this means cloud AI will disappear.
In fact, I expect the cloud to remain essential for large-scale training, demanding inference tasks, coordination, and services that require considerable computing power.
But perhaps the future won’t be cloud versus device.
Perhaps it will be about making better decisions regarding which intelligence belongs where.
There is Another Advantage We Often Overlook: Privacy
When discussing on-device AI, people usually emphasize speed.
And speed matters.
But I think privacy and control may be equally significant.
Imagine a professional working with confidential business documents.
Or a doctor reviewing sensitive patient information.
Or a company analyzing proprietary manufacturing data.
In certain situations, local AI processing could allow information to be analyzed without sending the underlying data to an external AI service.
That does not automatically make the system secure. Devices can still be compromised, local software can contain vulnerabilities, and data can still be exposed through poor design.
But it creates another option.
Organizations can make more deliberate choices about where information is processed.
That matters because the future of AI will not be determined only by how accurately a system answers questions.
It will also depend on how comfortably people and organizations can trust the way that system handles their information.
But Local AI Has Its Own Problem
I also don’t want to romanticize this shift.
Running powerful AI locally creates engineering challenges.
Hardware is expensive.
Memory is limited.
Thermal management matters.
Battery consumption matters.
Model updates become more complicated.
And smaller or compressed models may not perform as well as larger cloud-based alternatives on every task.
A device capable of running a model is not necessarily capable of running it economically or at the speed a particular application requires.
There is also a broader environmental question.
The International Energy Agency has projected that global data-centre electricity consumption could roughly double by 2030, reaching around 945 TWh.
That makes energy efficiency an important part of the AI conversation.
But moving workloads from data centres to millions of devices does not automatically reduce total energy consumption.
The outcome depends on hardware efficiency, utilization, workload design, manufacturing, and the wider computing architecture.
This is where my background in process excellence influences my thinking.
I have learned that moving a problem from one part of a system to another is not the same as solving it.
We should not celebrate local AI simply because it reduces cloud dependence.
We should ask whether the complete system becomes more effective, reliable, affordable, and sustainable.
The Next AI Competition Could Be About Architecture
For a long time, public conversations about AI have largely revolved around model rankings.
Which model is more intelligent?
Which model has the largest context window?
Which model performs better on benchmarks?
These questions are useful.
But I suspect that businesses will increasingly care about a different set of outcomes.
Can the system operate reliably when connectivity fails?
Can it respond within the required time?
Can it protect sensitive information?
Can it function within a realistic hardware budget?
Can it be maintained and updated safely?
And perhaps most importantly:
Does it actually solve the problem for which it was designed?
I believe we are approaching a stage where the quality of AI architecture will become a major competitive advantage.
Two companies may have access to the same AI models.
Yet one may build a far superior product because it understands how to combine local processing, cloud intelligence, sensors, software, and human decision-making.
This is something I find particularly exciting.
It creates opportunities not just for companies building frontier models, but also for engineers, startups, researchers, and organizations capable of designing better systems.
At ReThynk AI, we believe that AI Eldercare is the next revolution, and that’s why we covered AI ElderCare Project as DIY in the September Magazine.
What This Could Mean For The Next Generation of Innovators
I often think about the opportunities that emerge when two disciplines begin working together.
AI and electronics.
AI and robotics.
AI and mechanical engineering.
AI and agriculture.
AI and industrial automation.
These combinations can produce innovations that are much more meaningful than simply adding a chatbot to an existing application.
For innovators in India and around the world, I believe this is particularly important.
Many valuable applications must work under real-world constraints: limited connectivity, restricted budgets, demanding operating conditions, and users who care more about reliability than technological sophistication.
A brilliant AI system that fails when the internet goes down may be less useful than a simpler system designed around those constraints.
The future will need people who understand not only artificial intelligence, but also the environments in which intelligence must operate.
That is one reason edge computing and robotics are areas I want ReThynk AI to explore more deeply in the coming years.
Final Thoughts: Intelligence Needs to Move Beyond the Screen
I am excited about increasingly capable AI models.
I am fascinated by advances in reasoning, autonomous agents, and scientific discovery.
But I don’t think the AI revolution will be complete merely because we build a system that can answer almost any question.
The bigger transformation may happen when intelligence becomes an everyday part of the physical systems around us.
When machines can interpret their environments.
When agricultural systems can respond intelligently to changing conditions.
When industrial equipment can recognize problems before they become serious.
When devices can provide useful intelligence without constantly depending on distant computing infrastructure.
None of this eliminates the need for careful engineering, human responsibility, or appropriate safeguards.
But it changes what we can build.
For years, we have focused on bringing people to artificial intelligence through websites, applications, and chatbots.
Perhaps the next chapter will be about bringing artificial intelligence to the places where people, machines, and real-world problems already exist.
And if that happens, the next great AI innovation may not be another chatbot.
It may be a machine, a device, or an intelligent system quietly solving an important problem right where it occurs.
That is the future of AI I find especially interesting.
— Jaideep Parashar
About the Author
Jaideep Parashar is the Founder & Director of ReThynk AI Innovation and Research Pvt. Ltd., Six Sigma Black Belt, Lean Expert, AI Strategist, researcher, author, and keynote speaker. His work focuses on combining Artificial Intelligence, Lean Six Sigma, systems thinking, and continuous improvement to help organizations build reliable, scalable, and continuously improving AI-powered operations. To explore new ideas, join ReThynk AI community: LinkedIn Profile!