The State of Physical AI

2026-03-18 · Source: The Business Engineer · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Intermediate, quick

Summary

The software industry's foundational principle of near-zero marginal cost, which enabled immense wealth creation through frictionless digital distribution, is being fundamentally challenged by artificial intelligence at the frontier of capability. Unlike software, AI's intelligence is not weightless, its production is not frictionless, and its underlying infrastructure scales more like steel than traditional software. This expense is structural, driven by the physics of AI systems, which impose constraints that capital alone cannot overcome. The digital layer of AI is increasingly dependent on a physical layer that is more concentrated, fragile, and geopolitically exposed than previous computing infrastructures.

Key takeaway

For VPs of Engineering and Data evaluating AI investments, recognize that AI's structural costs mean traditional software scaling models do not apply. Your strategic planning must account for significant, ongoing physical infrastructure dependencies and their associated geopolitical risks, rather than expecting costs to approach zero over time. Prioritize robust supply chain considerations for hardware components.

Key insights

AI's structural costs challenge software's near-zero marginal cost principle due to physical infrastructure dependencies.

Principles

• AI's intelligence is not weightless.
• AI infrastructure scales like steel, not software.

Topics

• AI Economics
• Software Economics
• AI Infrastructure
• Marginal Cost
• Physical Layer Constraints

Best for: Investor, VP of Engineering/Data, Director of AI/ML, CTO, AI Architect, AI Product Manager

Related on AIssential

• How Do We Fix This? — Focus AI investments on use cases where returns outpace costs to achieve economic viability and avoid negative unit economics.
• Why Scaling Products Without Architecture Slows You Down — Scaling product models without rethinking enterprise architecture creates hidden costs and slows delivery.
• The internet was ‘too expensive’ too — New infrastructure platforms initially seem uneconomic due to bespoke systems, immature supply chains, and lack of scale.
• AI starting to look economically impossible outside hyperscalers? — AI's economic landscape splits between hyperscaler-monopolized frontier model training and accessible application-layer development.
• Why Your AI Costs Skyrocket As Your Product Gets Better — AI optimization paradoxically increases costs as product performance improves, challenging traditional SaaS economics.
• The AI Free Ride Is Ending: A Reckoning Is Coming for Software Engineering — The current AI-assisted development ecosystem relies on unsustainable investor subsidies, masking true costs and fostering dependency, leading to an inevitable pricing correction.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Business Engineer.