The $7 Trillion Hardware Renaissance

The Hardware Renaissance is here. Software’s 20-year reign is being challenged. With AI reshaping tech, we have seen the meteoric rise of semiconductors, which is turning conventional wisdom upside down.

Remember when “software is eating the world” was gospel? That era has run its course. Hardware now drives the future.

McKinsey & Company‘s April 2025 report, “The Cost of Compute,” quantifies this shift: a staggering $7 trillion in data center investments will be needed by 2030, with $5.2 trillion dedicated to AI workloads alone.

What’s happening?

• Cloud computing has recreated the mainframe era at a massive scale.
• Intelligence is split between cloud and edge.
• We’ve gone from decentralization back to centralization.

The market confirms this shift. Look at today’s leaders: NVIDIA, Microsoft, Amazon, Google, and Tesla.

In a fireside chat with me, Microsoft CEO Satya Nadella recently described it as “a golden age for systems,” comparing today’s excitement to the early 1990s, when RISC (Reduced Instruction Set Computing) architecture revolutionized computing.

This Hardware Renaissance is fueled by:

• Multiple S-curves of innovation in AI accelerators, GPUs, and other semiconductor areas.
• A fundamental rethinking of the entire compute stack.
• Compounding effects of systems, software, and models.

The leaders are already redesigning the entire stack – from silicon to models to middleware – for the AI era. This is the golden era for deep tech builders. Today’s system challenges are 1000x beyond what mainframes faced.

According to McKinsey, the $5.2 trillion AI investment breaks down:

• 60% ($3.1T) going to technology developers & designers (like NVIDIA, chipmakers).
• 25% ($1.3T) for energizers (utilities, cooling, power infrastructure).
• 15% ($0.8T) for builders (real estate, construction).

The hyperscalers (AWS, GCP, Azure) are systems companies that bundle hardware, software, network, and storage and offer them as a service—infrastructure as a service (IaaS).

Even cooling technology has become a critical innovation area. Who would have thought cooling for AI data centers would become a competitive advantage with $1.3 trillion in opportunity?

The innovation frontier has shifted to:

• Specialized AI accelerators and GPUs.
• Power and cooling solutions (40-50% of data center costs).
• 10x bandwidth connectivity needs for AI data centers, leading to innovation in physical connectivity and a need for scale-up switches for connecting GPUs.
• Demand for AI-related data center capacity could triple by 2030.

This isn’t just a temporary shift. As McKinsey pointed out: “The AI race isn’t just about model quality—it’s a physical race for compute, energy, and capital.” It’s a fundamental return to hardware as the foundation of technological advancement.

I recently explored this shift in more depth with Derek Andersen at the annual Startup Grind Conference. Watch the full conversation below.

Originally published on LinkedIn.