Zero Cost Compute. Zero Cost Energy.

For most of history, the progress of civilization has been defined by scarcity. Scarcity of power. Scarcity of processing. Scarcity of bandwidth. Every major breakthrough, whether it was the steam engine, the electric grid, or the silicon chip,

has been about reducing the cost of getting things done.

Today, for the first time, we are approaching a threshold where two of the most fundamental inputs in the modern economy, compute and energy, are both trending toward zero marginal cost.

This is not a hypothetical future. It is already happening.

Compute at Zero Cost

Cloud providers and semiconductor companies are in an arms race to drive down the cost of computation. NVIDIA’s newest accelerators, custom AI chips for inference, and open-source model optimizations are compressing the price of running models to fractions of a cent. What used to require an entire room of GPUs will soon run on a single smartphone chip.

When ChatGPT launched, the cost of inference was measured in pennies per query. Now it is a fraction of a penny, and the next generation of hardware will cut it further. Over the next decade, compute will be so cheap it will feel free for most applications.

Every form of knowledge work, coding, design, analysis, writing, becomes infinitely scalable. Once you have the model, creating the next copy costs almost nothing.

Energy at Zero Cost

At the same time, the marginal cost of producing energy is collapsing.

Solar and wind continue to follow learning curves that consistently reduce the cost per kilowatt-hour. In many parts of the world, the cheapest source of new electricity is already solar, often backed by grid-scale storage or distributed batteries. Small modular reactors are beginning to gain traction as emissions-free baseload power.

Combine renewables’ deflationary economics with improvements in storage and grid distribution, and the incremental cost of running another server, charging another vehicle, or powering another device starts approaching zero.

This is why the largest technology companies are investing heavily to secure their own energy supplies. Compute and energy are converging into a single, integrated cost curve.

Who Will Benefit

When compute and energy costs both collapse, the structure of economic value shifts. The companies positioned to benefit most are the ones that control infrastructure, models, and distribution.

Hyperscalers and Cloud Providers

NVIDIA will remain the foundational supplier of GPUs and the CUDA software ecosystem. Even if per-query costs keep dropping, total demand for compute will grow exponentially.

Amazon AWS, Microsoft Azure, and Google Cloud will be the default platforms for running AI workloads at scale.

Tesla is investing in its own Dojo supercomputer to handle enormous volumes of autonomous training.

Renewable and Next-Generation Energy Producers

NextEra Energy and Brookfield Renewable are leading operators of utility-scale renewables.

Tesla Energy is integrating solar and storage to push costs down even further.

Emerging nuclear companies, including Nuscale, are working to bring clean baseload generation online.

AI Software and Foundation Model Leaders

OpenAI and Anthropic will continue to license large language models that underpin the next wave of automation.

Palantir has positioned itself as the operating system that integrates AI into defense, manufacturing, and enterprise workflows.

Companies with Proprietary Data and Distribution

Apple controls the devices that will run these models locally.

Adobe owns the workflows where compute costs will fall, but creative control remains essential.

Intuit and Bloomberg sit on proprietary financial and business data that becomes even more valuable in a world of abundant compute.

Robotics and Automation Leaders

Boston Dynamics, ABB, Fanuc, and Tesla are all building platforms to turn cheap energy and cheap inference into widespread automation.

What Happens Next

As compute and energy approach zero marginal cost, the constraints move elsewhere. The new bottlenecks will be access to high-quality training data, rights to the most capable models, and the systems that govern their use.

Scarcity will not disappear. It will shift toward trust, relationships, intellectual property, and the physical materials that cannot be digitized.

This transformation is happening faster than most people realize. Just as the internet vaporized the cost of distributing information, the convergence of abundant compute and abundant energy will vaporize the cost of creating many types of value.

The upside is nearly limitless abundance. The challenge is ensuring that this abundance benefits more than a small number of companies and institutions.

Because when compute costs nothing, and energy costs nothing, the real question becomes: What do you build with it? And who gets to decide?