AI Is Deflationary But Its Energy Demand Could Fuel Inflation

Artificial intelligence is being heralded as one of the most transformative technologies of all time, with the potential to radically improve productivity, reduce costs, and drive economic growth. But as the AI revolution unfolds, a paradox is emerging: AI has deep deflationary potential, yet its explosive demand for electricity and infrastructure introduces inflationary pressures. Understanding these contradictory forces is essential to understanding how Artificial Intelligence could shape macroeconomic trends in the coming decade.

The Deflationary Power of AI

AI’s primary deflationary force lies in its ability to automate tasks across every sector of the economy — from legal document review and radiology scans to coding, customer service, and logistics optimization. By augmenting or outright replacing human labor, AI can significantly reduce business operating costs.

1. Labor Efficiency: AI systems like large language models (LLMs) and machine vision tools can perform tasks that once required skilled professionals, at scale, and with minimal incremental cost. For example, a customer support center powered by AI chatbots might handle 70–90% of inquiries without human involvement, reducing the number of customer support personnel and thus lowering overhead.
2. Manufacturing and Logistics: AI-driven supply chain models are enhancing inventory management, forecasting, and transportation routing — minimizing waste and lowering prices for consumers.
3. Software and Design: In industries ranging from architecture to pharmaceuticals, AI is accelerating R&D by enabling rapid prototyping and simulation. For example, in drug discovery, AI models can analyze massive datasets of molecular structures and predict how new compounds will interact with the human body — all before any physical testing occurs.

This process is known as working “in silico”, which means performed via computer simulation rather than in a lab (in vitro) or in a living organism (in vivo). By running experiments in silico, researchers can test thousands of hypothetical drugs in minutes, identify the most promising candidates, and dramatically reduce the need for expensive lab trials. This not only speeds up innovation but also cuts costs at nearly every stage of the R&D pipeline making AI a powerful deflationary tool in high-cost, high-risk industries.

For instance, by enabling rapid prototyping and simulation in drug discovery, AI models can analyze massive datasets of molecular structures and predict how new compounds will interact with the human body before any physical testing even occurs.

A perfect example of a drug discovered in silico is Exscientia’s DSP-1181, a drug candidate developed for obsessive-compulsive disorder (OCD).
Instead of relying solely on lab-based (in vitro) or animal testing (in vivo), Exscientia used AI algorithms to screen billions of molecular combinations via AI and then predict how compounds would behave in the human body, and then optimize the molecular structure for safety and efficacy.
This created massive time savings, reducing the preclinical development phase from the typical 4–5 years down to just 12 months.
These efficiency gains — especially when compounded across industries — are deflationary. They increase supply while reducing the labor and capital inputs required, pushing prices down over time.

In addition, as AI becomes more embedded in software platforms and embedded devices, the marginal cost of intelligence falls toward zero. This dynamic is similar to Moore’s Law, where computing power increased exponentially as prices dropped. In macroeconomic terms, AI boosts total factor productivity — a key driver of long-term deflationary growth.

But Here’s the Catch: AI’s Inflationary Side

The same AI tools driving efficiency and cost savings are enormously energy-intensive. This creates short- to medium-term inflationary pressure, especially in sectors tied to energy and infrastructure.

1. Data Centers and Energy Consumption: AI models like OpenAI’s GPT or Google’s Gemini require massive computational resources, both during training and inference. Training a single large model can consume millions of kilowatt-hours of electricity. As AI use scales, so does demand for power-hungry data centers filled with GPUs and cooling systems.
2. Grid Strain and Utility Pricing: The surge in electricity demand is outpacing grid upgrades in some regions. In the U.S., data centers already consume about 3% of total electricity, but some estimates predict that AI-related demand could double or triple that by the end of the decade. This rising demand pushes utilities to raise rates, especially in areas where infrastructure is aging or constrained.
3. Hardware and Commodity Inputs: The AI boom has triggered skyrocketing demand for high-end chips (like Nvidia’s H100), rare earth metals, and specialty cooling equipment. Supply constraints in any of these inputs can cause cost-push inflation — particularly in the tech and industrial sectors.
4. Construction and Real Estate Costs: Building data centers requires land, skilled labor, and materials — all of which are in high demand. In some regions, like Northern Virginia and parts of Texas, the AI data center boom is raising commercial real estate prices and straining local infrastructure, feeding localized inflation.

Balancing the Two Forces

The net economic effect of AI will likely be deflationary in the long run, as productivity gains offset short-term bottlenecks. But in the near term, especially during this infrastructure build-out phase, AI may contribute to sector-specific inflation, particularly in:

• Electricity and utilities
• Semiconductors and chipmaking equipment
• Construction and real estate
• Data center-related services

Economists and central banks are watching this closely. For example, Federal Reserve officials have acknowledged that while AI boosts potential output, the energy and capital costs associated with rapid deployment could pressure certain price indexes. Unlike the digital revolution of the 2000s, which piggybacked on existing infrastructure, the AI era demands entirely new physical and energy systems to scale.

Conclusion

AI is a deflationary force at its core, driving productivity, lowering costs, and enabling more output with fewer inputs. But the infrastructure needed to support this revolution — particularly the energy and hardware-intensive backbone of AI computing — introduces inflationary risks that cannot be ignored.

In the short term, policymakers and investors should expect a two-speed economy: one where prices fall in digital and service sectors enhanced by AI, while rising in power, construction, and materials. Over time, as energy grids modernize and AI hardware becomes more efficient, the deflationary effects are likely to dominate.

But for now, AI is both a promise of cheap intelligence — and a catalyst for expensive power.

Notes:

The large pharmaceutical companies are already using in silico tools. For instance,  Pfizer is collaborating with IBM Watson and XtalPi for molecule modeling. Novartis runs the Novartis AI Innovation Lab, collaborating with Microsoft. And AstraZeneca partnered with BenevolentAI and other platforms for drug repurposing.