When software moves fast but hardware can’t: Why modularity matters more than ever

At our recent Innovation Breakfast, one theme cut across almost every industry represented in the room: the widening gap between how fast software can evolve and how slowly hardware is allowed to change.

Whether people worked in engineering, research, manufacturing, infrastructure or advanced tech, the same tension surfaced again and again. It became one of the most insightful threads of the morning, and it’s a topic that deserves its own spotlight.

The uncomfortable truth: software can sprint, hardware moves in decades

Modern software teams can prototype, iterate and release faster than ever. But when the product relies on physical systems, reality hits:

• Materials are procured years before designs are final.

• Power, heat, weight and cabling impose hard physical limits.

• Regulatory and safety requirements slow iteration to a crawl.

• Long-life assets (vehicles, devices, industrial systems) stay in service for 10–30 years.

During the roundtable, one contributor described a real scenario:

“As the software develops, we’re having to turn off capability because the hardware can’t physically run it.”

This is a painful but common reality. Innovation is happening at the software layer, but the hardware underneath wasn’t designed to keep up.

The shift to a software-defined world… meets physical constraints

The room broadly agreed that the future is software-defined. Put as much intelligence, adaptability and differentiation in the software layer as possible.

But the group also identified a major blocker: You can’t have software-defined products sitting on top of hardware-defined limits.

And those limits show up everywhere:

• Not enough compute to run modern models

• Thermals and heat dissipation barriers

• Power constraints, especially in remote or off-grid environments

• Space limitations inside physical housings

• Cabling that can’t be replaced without dismantling half a machine

Even organisations pushing hard into AI and automation face the same physical bottlenecks.

What emerged from the room: modularity is the only way forward

Across different sectors, leaders echoed a shared view:

If hardware can’t evolve at the pace of software, the only viable strategy is to make the hardware modular.

Not modular for the sake of it, but intentionally designed so that:

• Components can be upgraded in situ

• Redundant capability can be activated when needed

• New intelligence can be added without redesigning the whole system

• Physical architectures support unknown future demands

One example raised in the room was consumer automotive: companies building excess, compute into vehicles today so they can “switch on” future features tomorrow. Not because the customer needs it now, but because hardware cannot be retrofitted easily later.

A similar pattern appeared in marine, aerospace and industrial systems - anywhere hardware has a long lifecycle.

Why modularity isn’t just technical… it’s strategic

The conversation revealed something deeper: a modular hardware foundation doesn't just solve engineering headaches. It creates new business opportunities.

With the right architecture:

• New features can be sold as upgrades

• Intelligence can be rolled out progressively

• Business models can evolve without redesigning the product

• Organisations can respond to shifting markets without waiting for the next hardware cycle

One guest framed it perfectly: “The innovation in the hardware is making it modular. The innovation in the software is everything that comes after.”

This is where AI and hardware design collide

AI is accelerating the pressure on physical systems. Models are getting larger. Edge computers are becoming more capable. Expectations around autonomy, decision-making and real-time insight are rising.

But unless the hardware can support those demands, the software can only go so far.

This led to a clear takeaway from the roundtable: AI maturity is increasingly limited by hardware maturity. And hardware maturity is increasingly dependent on modularity.

Where organisations are still guessing

When the group reflected on this topic, several open questions emerged:

• How much redundancy is “enough” when future demands are unknowable?

• How do you design modular systems without inflating cost?

• What parts of the system should be modular vs fixed?

• How do you communicate the value of modularity to stakeholders who prioritise short-term budgets?

• Who owns the long-term roadmap when software and hardware evolve at different speeds?

These questions weren’t solved in the room, but elevating them is the first step.

Why this matters for innovation strategy

Every organisation exploring AI, automation or digital transformation eventually hits the same barrier: your physical systems define your ceiling.

The conversation at the breakfast showed that modularity isn’t just an engineering preference. It is a strategic foundation for innovation over the next 10–15 years.

Teams that build modular hardware now will unlock:

• Faster adoption of new intelligence

• Lower long-term costs

• Greater resilience in volatile markets

• More flexibility to pivot as AI capabilities grow

• New recurring revenue opportunities

Teams that don’t will find themselves limited by the decisions they made years earlier.