What the AI Industry Gets Wrong About Energy
EMPHOS Group · April 16, 2026 · 6 min read
Meta: The AI industry has an energy problem that cannot be solved by better chips, greener data centers, or more efficient training runs. It is structural. Here is what the numbers actually say — and why Heinrich is the only architecture that addresses the root cause.
Excerpt: GPT-4o uses 0.34 watt-hours per query. Heinrich uses 0.00003. That is not a 10% improvement. That is not even a 10x improvement. That is 11,000 times less energy per query — and the gap does not close at scale. It widens. Here is why.
GPT-4o uses 0.34 watt-hours per query. Heinrich uses 0.00003. That is not a 10% improvement. That is not even a 10x improvement. That is approximately 11,000 times less energy per query — and the gap does not close at scale. It widens.
This is not a marketing claim. The Heinrich number is measured directly on a production system running on a standard Windows laptop, no GPU, no optimization applied. The GPT-4o number is from OpenAI's own public disclosure. Both numbers are cited in EMPHOS Group's Environmental and Resource Efficiency Report, published April 2026, with full methodology and source documentation.
The question is not whether the gap is real. It is. The question is why it exists — and whether anything the AI industry is currently doing will close it.
The answer is no. And understanding why requires understanding what the energy problem actually is.
What the numbers actually say
GPT-3 required 1,287 megawatt-hours to train. That is the energy consumption of approximately 120 average US homes for a full year, spent once to produce a single model. GPT-4 required an estimated 16,200 megawatt-hours — the equivalent of 1,500 homes for a year. These are not operational costs. They are the cost of creating the system before a single user query is processed.
BLOOM 176B — one of the most carefully measured large language models — consumes 3.9 watt-hours per query, measured directly on 16 Nvidia A100 GPUs by Luccioni et al. in 2022. At 1 billion queries per day that is 3,900,000 kilowatt-hours of electricity consumed daily by a single model. At the IEA's 2023 global average grid intensity of 0.4 kilograms of CO₂ per kilowatt-hour, that is 1,560 tonnes of CO₂ per day from one model running at scale.
Heinrich at 1 billion queries per day consumes approximately 30,000 kilowatt-hours. That is 12 tonnes of CO₂ per day at the same grid intensity. The difference is approximately 50,000 tonnes of CO₂ per year — equivalent to removing 10,000 cars from the road — from a single deployment decision.
These are not projections built on assumptions. They are extrapolations from a directly measured baseline, using published competitor data and standard grid intensity figures. Every number is documented and sourced.
Why the industry cannot fix this with better hardware
The response from the AI industry to energy concerns has followed a predictable pattern: better chips, more efficient data centers, renewable energy procurement, and claims of improved performance per watt on each new hardware generation.
None of this addresses the structural problem.
Large language models store knowledge as numerical parameters — billions of floating-point weights in a matrix. Retrieving knowledge from those parameters requires matrix multiplication: multiplying the input by the weight matrix, applying activation functions, passing the result through dozens of layers. This operation requires GPU-accelerated hardware because CPUs are too slow to do it at the scale required. The compute cost is proportional to the number of parameters. The energy cost is proportional to the compute cost. There is no version of this architecture that runs at milliwatt power levels, because the arithmetic required to retrieve knowledge from a parameter matrix cannot be done in microseconds on a general-purpose CPU.
Better chips reduce the energy per floating-point operation. They do not change the number of floating-point operations required. The efficiency gains from hardware improvements are real but bounded. The structural cost of the architecture remains.
Renewable energy procurement changes where the energy comes from. It does not change how much is consumed. A data center running on solar power still consumes the same number of kilowatt-hours as one running on coal. The carbon intensity changes. The energy demand does not.
Why Heinrich is structurally different
Heinrich does not store knowledge as parameters. It stores knowledge as frequency coordinates — sinusoidal components in a layered signal field. Retrieving knowledge is Goertzel correlation: a single-frequency signal processing operation that determines whether a specific frequency is present in a signal. This operation runs in microseconds on any CPU. It requires no GPU. It requires no matrix multiplication. It requires no dedicated AI silicon.
The compute cost per query is proportional to the number of concepts that activate in response to the query — the subfield that resonates around the relevant frequency coordinates. It is not proportional to the total number of concepts in the field. Heinrich at 50 million nodes costs the same to query as Heinrich at 1.75 million nodes, because the subfield that activates for any given query is the same size regardless of how large the surrounding field is.
This is why the 0.2% CPU and 78 megabyte RAM measurements taken at 128 concepts in April 2026 have not changed as the field has grown to 1.75 million concepts. The architecture does not work any other way. The efficiency advantage is not something that will erode at scale. It is a structural property of how knowledge is stored and retrieved.
You cannot get to this efficiency by optimizing a large language model. The parameter matrix is the bottleneck. The only way to remove the bottleneck is to not have a parameter matrix. That is what Heinrich is.
The training problem
The energy numbers above cover inference — running a model after it has been trained. The training numbers are worse.
Every large language model requires a full training run before it can answer a single question. GPT-4's estimated training cost of 16,200 megawatt-hours is spent once. But it is not spent once in the way a factory is built once and then runs indefinitely. It is spent once per version. When the model needs to be updated with new knowledge, the options are full retraining — spend the energy again — or fine-tuning, which is partial retraining and still requires significant compute.
Heinrich has no training run. Knowledge is added by writing a value to a frequency coordinate. The cost of adding one concept to the field is the cost of computing its harmonic address and writing it to a database. On the current ingestion pipeline that cost is measured in fractions of a millisecond per concept. The field has grown from 128 concepts to 1.75 million in three days, continuously, at near-zero marginal energy cost per concept added.
The total training energy cost of Heinrich AI to date is effectively zero. Not low. Not efficiently managed. Zero in any meaningful comparison to the systems it is being measured against.
What this means for the industry
The AI industry's energy problem is real, growing, and structurally unsolvable within the current architectural paradigm. Data center electricity demand from AI workloads is projected to double or triple within five years. The chips are getting more efficient. The models are getting larger faster than the chips are getting more efficient. The net energy consumption is going up.
The environmental commitments being made by large technology companies — net zero by 2030, 100% renewable procurement, carbon offset programs — are responses to a problem they do not believe they can solve structurally. They are managing the consequences of an architecture they are committed to, because the architecture works and they have invested too much in it to change course.
Heinrich is not a response to that problem. It was not designed to be more efficient than a large language model. It was designed as a fundamentally different approach to storing and retrieving knowledge — one that happens to be 11,000 times more efficient as a consequence of its physics, not as a design goal.
The efficiency is real. It is measured. It is documented. It does not erode at scale. And it is available to anyone who builds on the architecture — including the hearing aid sitting on your bedside table, and the assistant platform running on your laptop, and every deployment that comes after.
What this is not
This is not an argument that large language models should not exist. They are remarkable systems that have demonstrated genuine capability across a wide range of tasks. The argument is narrower: the energy cost of those systems is structural, not incidental, and the approaches being taken to manage it do not address the structural cause.
Heinrich is not a replacement for every AI application. It is a fundamentally different architecture for storing and retrieving structured knowledge — one that is honest about what it knows, deterministic in how it retrieves it, and structurally efficient in a way that no parameter-based system can match.
For the applications where that matters — personal intelligence, on-device inference, accessible AI, the hearing aid, the assistant that runs on your laptop without a subscription — the architecture is not one option among several. It is the only architecture that works.
What comes next
The Environmental and Resource Efficiency Report documenting Heinrich's measured resource profile is available to investors and grant reviewers on request. EMPHOS Group is pursuing environmental innovation funding through Innovate BC, the NRC Industrial Research Assistance Program, and the federal Strategic Innovation Fund.
The field is at 1.75 million nodes and growing. The efficiency numbers have not moved. They will not move at 50 million. That is the experiment currently running, on a laptop in Chilliwack BC, at 0.2% CPU, right now.
Engineered for Presence.
Stay in the loop
EMPHOS publishes twice a week — product updates, research, and the thinking behind the build.
Explore Haven · HEINRICH Intelligence · The EMPHOS Vision · All Posts
EMPHOS Group · Chilliwack, BC, Canada · info@emphosgroup.com