Fractal Turns Data Centers Into Dinosaurs
There is no application that needs a data center today - we can prove it
The False Narrative Driving America’s Data Center Explosion
America needs a sustainable path to A.I. leadership - without the constraints of data centers consuming the water of entire towns.
The world is wasting a trillion dollars to build more data centers because of a myth.
The myth is that the only way to make software, or AI agents, fast and scalable is to put everything in the cloud or a data center.
There is a sustainable solution — at 1/10th the cost, consuming a fraction of the energy – without the environmental devastation.
Fractal Computing reimplements application software, or AI agents, so each is much smaller -- placing copies of the smaller software everywhere data resides.
Fractal Computing places the computing where data resides, not moving all the data to a central, power-hungry data center.
Fractal Computing coordinates the copies, so they act as a single system.
No cloud, no data center, required.
Fractal Computing provides Locality of Reference optimization ensuring each copy has the data it needs to "do useful work" accessing the required data thousands of times faster than in a centralized data center.
Thousands of times faster means less hardware, less energy, less cost and no cloud or data center.
Fractal Computing provides sustainable computing, today, delivering A.I. applications faster and without the negative environmental impact.
The False Narrative Is Driven By An Obsolete Computing Model
The false narrative that America needs trillions of dollars in new data centers, consuming electric power that does not exist is driven by legacy forces.
1980s technology vendors, with 40-year-old technology stacks, indeed need massive data centers to operate.
American A.I. dominance should not be dictated by the constraints of legacy software vendors.
In A.I., demand for building new data centers is often driven by the need to train new Large Language Models (LLMs).
Placing an application, AI agent or critical data into an LLM, in a cloud data center and bringing all one’s data directly into any LLM or cloud is expensive and inefficient – and there is an alternative.
Leading innovative firms like Power Analytics have proven LLMs can be bypassed through intelligently abstracting to databases outside the LLMs – delivering A.I. systems at 1/16th the cost of conventional technology.
Data center requirements are eliminated, delivering cost savings of 10 to 1.
Adopting a 1980s compute model for A.I. imperils U.S. security.
It is dangerous for America to commit blindly to building data centers as the default path for future A.I. dominance.
A typical data center can consume over 400,000 gallons of water a day and use the electric power equivalent to a town of 25,000.
Data centers do not produce significant numbers of jobs, and they destroy rural farmland.
Even if the massive data center model were optimal, there is not the electric power available to support its needs.
It is estimated that data centers today use about 2%-3% of available electric power.
Current forecasts – if America were to blindly default to the conventional wisdom that data centers are needed for A.I. – would consume up to 7% or more of electric power.
In January 2025, DeepSeek showed the need for massive data centers for A.I. is a fallacy.
Fractal and its customers have taken that argument to the next level – showing there is no need, for any government or commercial application, or for A.I. applications, for a data center at all.
Currently billion-dollar corporations use Fractal technology to move energy-consuming applications out of data centers to tiny computers residing with the data.
In a recent demonstration project, the Fractal team took the Federal Election Commission database – 680,000,000 records, which requires a massive U.S. Government data center, plus a cloud, scores of employees, a 4 to 5 figure-a-month electric bill and moved the entire database to a computer 4-inches by 4-inches in size.
The Fractal equivalent of the FEC database did not require a 5-figure monthly electric bill – it ran on less power than a kitchen microwave oven.
No data center needed – just a computer one can hold in one’s hand.
No expensive staff – the system is managed by one engineer, a few hours a month.
America is at an inflection point.
The false narrative that energy-hungry data centers are needed for A.I. dominance has proven to be false.
Now is the time to commit to projects demonstrating any application can be built or migrated without a data center or cloud.
Sustainable computing is here – proven today.
America has it. America’s adversaries do not.
This is a generational opportunity to embrace sustainable computing and the advantages it brings to A.I. development.
FractalComputing Substack is a newsletter about the journey of taking a massively disruptive technology to market. We envision a book about our journey so each post is a way to capture some events.
Subscribe at FractalComputing.Substack.com
Fractal Website: Fractal-Computing.com
Fractal Utility Site: TheFractalUtility.com
Fractal Government Site: TheFractalGovernment.com
Portions of our revenue are given to animal rescue charities.
Just follow us here. We are having huge success commercially and the Fractal future is assured. We are doing weekly updates so I hope you like them.
I thought I would respond by having a current customer respond:
Jay, clearly this "president of a tech company" is part of the obsolete, data center centric guys we meet every day and he misses the point entirely.
We brought in Fractal for our billing system. We were told by IBM the cost of a new system, which we needed was around $4 million to $7 million and would take 18 - 24 months.
Using Fractal, we had it fully operational in 23 days and we eliminated the need to expand the data center for it.
Your comment from the dinosaur gets into disaster recovery too.
Again, he has a 1980s mind set - and he is clearly not a critical thinker.
Restart recovery, disaster recovery - in a 100% distributed system like Fractal is a different concept.
We run Fractal in 3 concurrent systems - that means we run 3 copies of our billing system, 24/7 with less than 30 seconds of downtime a year.
If any Fractal system has a problem, the others immediately replace it. There is nothing we did with current tech in disaster recovery that comes close to what we do with Fractal.
More interesting is the "costs are minimal" comment.
Clearly this person is a low tech CEO - because in one Fractal system - just our billing system, we saved over $4 million on development, and we were able to run it without using the data center, and when we retire the current billing system, the savings will exceed $8 million.
The yearly savings are over $3 million for DevOps - which we no longer need, and maintenance, and we eliminated Oracle and VMware and a lot of other software.
So do not spend time with dinosaurs, focus on people who have deep technology backgrounds and can use Fractal as it is intended.