Next-Gen Aerospace Materials Developed With AI Go Into Mass Production

In a historic first, US company Aether Biomachines is scaling up production of novel materials developed using AI to industrial quantities. Their first product is RapidPrint™, a polymer filament which enables printing four times the speed of existing alternatives. This will be followed shortly by Ultra Print filament, a composite with the strength of aviation-grade aluminum at half the weight.

These products are the tip of the iceberg. Because what Aether’s technology offers is not just new materials, but the ability to identify useful new materials with AI and then construct them, atom by atom.

“One of the things we have got very good at is targeting a novel molecular species and making it very quickly,” Pavle Jeremić, Aether Biomachines CEO and founder told me.

The company has been in stealth startup mode for seven years. Now it is sharing the fruits of its research –- and ramping up production to an industrial scale.

Molecular Engineering

Researchers have talked about of nanomachines, tiny robots able to break down matter and reassemble it in any desired form ever since Eric Drexler’s landmark 1990 book Engines of Creation. Jeremić’s background is biomolecular engineering at UC Davis, researching how natural processes work at this scale. The book was a big influence on Jeremić , but rather than Drexler’s mechanical nanobots, he saw a future based in biotechnology, hence ‘Biomachines’ in the company name.

Nature is incredibly efficient compared to human manufacturing when it comes to producing high-performance materials. Aramid fiber for example, the synthetic material used in Kevlar bullet-proof vests, requires high temperatures and pressures, very specific ingredients – “aromatic diamines and aromatic diacid halides” — and a large industrial facility. Spider silk, which is stronger, is made on demand at room temperature and pressure from an organ the size of a grain of sand. And the only raw material a spider needs to make silk is a supply of dead flies.

Aether Biomachines’ approach is more like a spider than a factory, usings biological processes specifically engineered to produce materials with the exact properties needed, and to it economically.

An AI Treasure Map To Find Valuable New Materials

In the case of RapidPrint, the first step was to identify a suitable new material.

“What we were trying to figure out was what kind of molecule, which in its initial state would flow, then at a specific temperature it would rapidly polymerize and harden,” says Jeremić.

The required material was a novel pre-polymer, the magic ingredient which makes it harden seconds after printing.

Researchers have longed used biological enzymes for industrial processes and more – it is enzymes for breaking down stains which make washing powder ‘biological’ – but one of Aether Biomachines’ particular skills is designing and manufacturing new enzymes for specific tasks.

An enzyme can be thought of as a molecular machine for modifying other molecules. Having identified the task that needed doing, the challenge finding an enzyme to do it. Such synthetic enzymes which do not occur in nature are easy to make, but there are billions of possible proteins which can be made by assembling amino acids in different ways. How do you know if there is one to do what you need, and how do you identify it?

Jeremić claims Aether’s means of solving this challenge makes them unique.

“This is what we call Indexing,” he explains. “It is a way of summarizing the problem mathematically, asking what protein can carry out a particular molecular function. Every protein is a data point, and we label each one with its many functions by empirically testing them.”

In practice this means a laboratory full of hundreds of plastic plates, each with hundreds of tiny wells, each well containing a different protein, going through an automated assay process to determine how the protein reacts to thousands of other molecules. Over time, Aether Biomachines has mapped out the function of millions of different proteins.

This is very different to other biotechnology researchers, who are interested first in the structure of every protein, and in particular how it is folded which affects how it reacts. Rather than just checking for one particular function, Aether’s analysis process looks at a wide range of reactions,

“People used to think each protein just had one function, now we know they are promiscuous, they can have many different functions,” says Jeremić. “If you only test for one or two functions, you do not get the full scope of what each protein can do.”

All these data points and labels are all fed into a machine learning process with carries out the indexing, turning the individual dots into a big picture so it is possible to see what lies in the blank spaces in between. In effect it provides a treasure map, guiding researchers to rapidly home in on the exact molecule they need. Indexing is the sort of task that is virtually impossible for humans but easy for AI with its ability to handle vast quantities of data.

“We’ve discovered seven classes of novel chemistry that are useful, and hundreds of others that are not so useful,” says Jeremić. “We have serendipitously discovered many interesting things along the way.”

Ultimately, the enzymes should be able to carry out pretty much any transformation that you can think of.

“We are producing tunable molecular factories that can take a given input and transform it into a particular output,” says Jeremić.

It sounds great in theory. But can it make real products in commercially useful quantities?

Turning Theory Into Products

What investors and customers want is some useful output – and perfecting that side is partly why Aether are only now emerging after seven years of quiet development.

The challenge was to identify a material to produce big returns quickly without the need for massive investment. The company fixed on RapidPrint, which can greatly boost the efficiency of 3D printers widely used in the aerospace industry.

The exact benefits of RapidPrint depend on the specific setup, but Jeremić says that describing it as speeding printing by a factor of four is “conservative.”

“For example, typical Nylon-CF filament can print anywhere from 30-80 mm/s, strongly depending on the material and the printer. In Aether’s case, our customers typically start printing at 300-450 mm/s without compromise to mechanical performance, with some customers pushing to 900 mm/s,” says Jeremić.

Jeremić says that potential customers tend to be skeptical of the speed increase until they have tried test spools. Then they want more. The potential benefits of faster production are obvious.

Drone makers like Ukraine’s Wild Hornets already have 3D printer farms turning out drone components; the outfit runs over 350 printers nonstop. This type of operation might be interested in a new material that could multiply production overnight.

In the process of developing RapidPrint, Jeremić’s team made a surprising discovery – one of those ‘serendipitous’ finds. Carbon fiber is added to composites to increase strength, but this only works up to a point, usually around 30% fiber. With the new formulation, they found they material could absorb 40-50% or more carbon fiber, giving unprecedented strength – the new Ultra Print range. This is far stronger than existing 3D print material.

“Suddenly we can create a polymer which is stronger than metal, but only half the density,” says Jeremić. “That’s a huge deal.”

The industry has known about such material for years, but due to high viscosity they have always been unusable materials. Aether Biomachine’s new formulation changes that. The new material offers the potential to create aircraft, missiles and drones with the precise, intricate engineering made possible by 3D printing, and at high speed.

Novel manufacturing techniques are a feature of the new drone age. Anduril’s giant ‘hyperscale’ Arsenal-1 facility, intended to build missiles in vast numbers, relies heavily on 3D printing. Ukraine’s FP-5 Flamingo cruise missile features a body made of carbon fiber on a winding machine, taking six hours. New materials could make traditional metal working techniques look antiquated.

Ramping Up Production

The first challenge is scaling up, which Aether are doing with the aid of $15m of recently announced capital investment.

The total amount of RapidPrint produced to date amounts to 1.5 tons. Jeremić says the company is now on course to make 10 tons per month by mid-2026, scaling to hundreds of tons by the end of the year. In 2027 Aether could be making thousands of tons per month, depending on demand.

“The demand for RapidPrint already exceeds the supply, but as we scale up we are being careful to learn to walk before we can run,” says Jeremić.

In the longer term, Jeremić talks about re-imagining industry, and a world where units the size of shipping containers can produce almost any material required using basic raw ingredients. The company already has a number of promising products lined up for processes like economically extracting lithium from seawater.

There are likely to be more lucrative possibilities revealed by the indexing process, and Aether Biomachines are the only ones with the treasure map to find them.

“Our core concept of a protein function model is unique,” says Jeremić. “We’re the only people on the planet generating this data, so only we can predict what enzymes will do.”

The protein space is vast, and indexing it is an ongoing process. Improved AI software and hardware will accelerate the process, opening up more possibilities not yet dreamed of,

If RapidPrint performs as claimed, it will certainly attract considerable interest, and the Ultra series could shake up aerospace production. But if Jeremić is right, they are just the start of something much bigger.