AI Trends 2026: Quantum, Agentic AI & Smarter Automation

What will be the most important trends in AI in 2026? Well, we take a stab at this every year with with some success, I would say. And this time out, I have the knowledgeable assistance of my colleague, Aaron Baughman, to help us out. Well, yeah. You know, after your prediction of infinite memory last year, I thought maybe you could use just a little bit of help. Yeah,

that's that's fair. Well, how about we each take four trends each? That sounds good. How about you first? All right. Okay. So my number one trend of 2026 is multi-agent orchestration. Now last

first? All right. Okay. So my number one trend of 2026 is multi-agent orchestration. Now last

year we said 2025 was the year of the agent. AI agents that can reason and plan and take action on a task and agents I think it's fair to say really delivered. There are new numerous agentic platforms for tasks like coding and basic computer use but no single agent really excels at

everything. So, what if you had a whole team of agents working together? So, maybe we've got an

everything. So, what if you had a whole team of agents working together? So, maybe we've got an agent here that kind of acts as a planner agent that decomposes goals into steps. Maybe we have some worker agents here that do different steps like one specializes in writing code,

others call APIs and so forth. And then perhaps we have a critic agent that evaluates outputs and flags issues. And these agents collaborate under a coordinating layer that is the orchestrator.

flags issues. And these agents collaborate under a coordinating layer that is the orchestrator.

And multi-agent setups like this help introduce cross-checking where one agent checks the other agents work and it can break problems into more discrete verifiable steps. Well, great. So,

how could I really follow that trend? Well, I think I might just have one. So,

the second one is going to be the digital labor workforce. So now these are digital workers that are autonomous agents that can do a couple of items. So the first one is they can parse a task by interpreting multimodal input. So after preparation the worker then executes what's called

a workflow. Now this is where at the end of an action plan you know it would follow a sequence of

a workflow. Now this is where at the end of an action plan you know it would follow a sequence of steps but then it has to be integrated into some sort of system that then in turn can take action.

And these could be downstream components. Now these systems are then further enhanced by what we call human-in-the-loop AI, which then provides a couple of items. The first one would be oversight.

The next one would be correction and then we're looking at these strategic guidance or these rails um to ensure that all of these agents are doing what they're supposed to be doing. Now this

overall trend will create a force multiplying effect to extend human capability. Now trend

number three is physical AI. Now we all know that large language models they generate text like ABC.

And then there are other models as well. So for example there are plenty of diffusion image models and they generate pixels. They generate images. These are all operating in digital space. Now,

physical AI is about models that understand and interact with the world that we live in, the the real 3D world. And this is about models that can perceive their environment, reason about physics, and that can take physical action like robotics. So, previously getting a robot like this to do

something useful meant programming explicit rules. So if you see an obstacle, you should turn left, for example. And it was all done by humans. It was up to yeah, smart guys like this to code these

for example. And it was all done by humans. It was up to yeah, smart guys like this to code these rules. Now, physical AI kind of flips that around. So you train models in simulation that simulate

rules. Now, physical AI kind of flips that around. So you train models in simulation that simulate the real world and it learns to understand how objects behave in the physical world, how gravity works, how to grasp something without crushing it. Now these models are sometimes called

world foundation models. They're generative models that can create and understand 3D environments.

They can predict what happens next in a physical scene. And in 2026, many of these world models are taking things like those humanoid robots that you found there, Aaron, and they're taking them from research to commercial production. Physical AI is scaling. Well, Martin, you just took my trend,

but let's just go ahead and say number four is about social computing. Now, this is a world where many agents and humans operate within the shared AI fabric. So say if I have an agent here and then a human here. So they're going to be connected through this fabric and here if I

have information that flows between the two, they begin to understand each other and then they can gather what the intent is going to be. And then once they have the intent and information, they have actions. They can affect each other or maybe even the environment of which they're in. But all

have actions. They can affect each other or maybe even the environment of which they're in. But all

of this flows seamlessly across this system. It's this shared space that enables collaboration, context exchange as well as event effective understanding. Now the outcome is really an empathetic emergent network of these interactions. It's what we call this collective intelligence

or this real world swarm computing. So teams of agents, digital labor, humanoid robots, and tech that can understand me with effective computing. 2026 could be uh quite the year and we're only

halfway through the trends. So trend number five that is verifiable AI. Now the EU AI act is coming and by mid 2026 it becomes fully applicable. And think of this a little bit like GDPR but for

artificial intelligence. Now, the core idea here is that AI systems, especially high-risk ones,

artificial intelligence. Now, the core idea here is that AI systems, especially high-risk ones, need to be auditable and they also need to be traceable. Now, what does that mean? Well,

it means a few things. It means documentation. So, if you're building high-risk AI, you need technical docs that demonstrate compliance to how you tested the models and the risks that you identified. It means transparency. So, users need to know when they're interacting with the machine.

identified. It means transparency. So, users need to know when they're interacting with the machine.

So things like synthetic text, they need to be clearly labeled and it means data lineage. You

need to be able to summarize where your training data came from and prove you respected copyright optouts. And just like how GDPR has shaped global privacy, not just folks in the EU, the EU AI act

optouts. And just like how GDPR has shaped global privacy, not just folks in the EU, the EU AI act will probably set the template for AI governance worldwide. Wow, that's great. And you know, trend number six, right? It really changes everything, but it also changes nothing at the same time.

And now this is where we put in quantum utility everywhere. So 2026 is where we start to see this quantum computing to reliably start solving real world problems better, faster, or more efficiently than classical computing methods. Now, at this point, we have this quantum utility scale.

is these systems that begin working alongside and together with classical infrastructure to deliver these practical value in everyday workflows. Now, this is going to help with optimization and then we'll also look at simulation and decision-making. Now, all three of these tasks were previously out of reach within the classical realm. But this hybrid quantum classical error, it will begin to

transform quantum computing into this mainstream paradigm as it's going to be woven into our everyday business operations. Now my trend number seven is reasoning at the edge. Now last year, we talked about very small models, models with just a few billion parameters that don't need huge

data centers to run. They work on your laptop or well maybe even your phone. Well, in 2026, those small models are learning to think. So, if we think about the best models that we have today, the frontier models, well, pretty much all of them now use something called inference time compute.

They spend extra time thinking before giving you an answer, working through problems step by step.

Now, the trade-off for that is they need more compute. But here's what's changing. Essentially,

teams have figured out how they can distill all of this reasoning information into smaller models.

So now these smaller models can perform thinking as well. You're taking massive reasoning models that generate tons of step-by-step solutions and we're using that data to train the smaller models to reason the same way. And that's resulting in reasoning models with only a few billion parameters. They work offline. Your data never leaves your device. And there's no roundtrip

parameters. They work offline. Your data never leaves your device. And there's no roundtrip latency to a data center. So for anything that's real time or mission critical, having a model that can actually reason through a problem locally is a pretty big deal. Yeah. So that's all very true,

Martin. But now our last and final trend is number eight. So this is what we're calling amorphous

Martin. But now our last and final trend is number eight. So this is what we're calling amorphous hybrid computing. So this is a future where both AI model topologies and the cloud infrastructure,

hybrid computing. So this is a future where both AI model topologies and the cloud infrastructure, they blend into what's called a fluid computing backbone. So AI models, they're shifting beyond just this pure transformer design, right? They're beginning to evolve into these other architectures

that integrate transformers and we call them these state space models. And then in 2026, you're also going to see different emerging algorithms that are combine both the state space and transformers and other elements together, right? And that's going to be really fun to watch, very artful. And then at the same time, we have this cloud computing piece that's becoming fully

very artful. And then at the same time, we have this cloud computing piece that's becoming fully differentiated by combining many different chip types. So we're going to have CPUs, GPUs, TPUs as well. And finally, what we just talked about in trend six, quantum, we're going

to have QPUs. I did also want to mention and note that you'll see these neuromorphic chips that are coming out and those emulate the brain. But all of these are going to be put together right into this unified compute environment where parts of each of these types of models, they're going to be automatically mapped to the optimal compute substrate. And this is

really going to help to deliver this maximum performance and efficiency. And you know what?

Who knows? But at this pace, probably not in 2026, but I think further out, you might see DNA computing entering into the mix. Well, those are some lofty goals. And look,

these are what we think are some of the biggest AI trends in 2026. But what are we missing?

Which AI trend do you expect to be a big deal in 2026? Yeah, let us know in the comments below.