AtlasBase has built a DNA-based storage and computation chip device, Thalia, a molecular computer, with 5.6 billion electro-chemical synthesis sites, a 700x increase in scale over previously available DNA synthesis chips.
This is a hugely impressive technology and, AtlasBase says, Thalia "forms the heart of the AtlasBase molecular computer, providing the scalable engine to drive molecular data storage and molecular compute/processing solutions for the growing compute, data management, and power constraints of the AI era." Well, yes, impressive marketing as well.
CEO Jeff Treuhaft said: “For 2 billion years, DNA has flourished as nature’s data platform, an incredibly capable structure for processing the information that defines every living thing. We’re now harnessing that same architecture for computing at a scale that was never before possible. AtlasBase is delivering a new enterprise data layer where “in memory” computation works efficiently, bringing molecular computing into commercial reality at IT scale.”
Bear in mind that in-memory computing typically is either SW-based, with data loaded into RAM and processed there, or HW-based with processing-in-Memory (PIM) silicon. SW-based examples are GridGain, HazelCast, Redis and SAP HANA. HW examples, all works in progress, include Micron, Samsung and SK Hynix. Both types operate thousands and perhaps millions of times faster than DNA-based in-memory computing which progresses at chemical reaction speed and DNA sequencing read-out times.
Thalia is “a molecular computing platform that uses synthetic DNA for both data storage and in-memory computation, enabling specialized processing on data within the same molecular medium.” So “enterprises [can] archive massive datasets in synthetic DNA while extending molecular computing to solve highly complex computational problems on that same data, without requiring it to be moved, shuffled, or translated between storage and compute.”
Technology
Thalia is called a Bio-Silicon platform. A software codec converts digital data into chains of amino acids – A, C, T, and G – the building blocks of DNA. Once encoding is complete, this four-letter blueprint is sent to the AtlasBase Bio-Silicon platform. A chemical process, Polymerase Chain Reaction (PCR), produces low-cost copies. The DNA is dehydrated and hermetically sealed in stainless steel capsules, stored by AtlasBase or on customer premises. When decoding is needed a commercial sequencer reads the DNA and decodes the A, C, T, G blueprint back into a digital file.
AtlasBase says a petabyte-scale archive fits in a 7mm x 18mm capsule, but a supplied capsule image shows it holding just 10 TB.
DNA Storage
AtlasBase says its storage tech starts at 100 TB per LTO footprint and could scale out to 5 EB in the same space. That’s the same as the compressed capacity of an LTO-10 tape. The LTO roadmap extends out to LTO-14 with a 913 TB compressed capacity. We’re talking an order of magnitude less LTO capacity in the same footprint here
An AtlasBase cartridge appears to have 10 x 10 capsules from its image; 100 capsules, and each would hold 1 TB to make up a 100 TB cartridge.
It would have the usual offline DNA storage capabilities; stable for 1,000+ years, immutable, immune to electro-magnetic radiation, etc. Assume 10 TB capsules and we would have a 1 PB cartridge.
There are no examples supplied of actual robotic DNA cartridge libraries and we understand that competing future technology archive suppliers, such as Cerabyte, are further ahead in this regard.
DNA production
AtlasBase says it can create synthetic molecular data at scale using its 3-layer Bio-Silicon device which controls bio-chemical processes using a CMOS array. It's a DNA writing device with a high throughput.
Level 1 is CMOS control and sensing. Level 2 is a "Bio-Device Array" and level 3 is where the bio-chemistry happens. AtlasBase says this is the densest Bio-Silicon Device ever with " 16.8 Billion Devices Per [and] 250 GB Per Chipset Per 24hrs." It is scalable and mass manufacturable. We don't know if this chip actually exists, apart from being a design concept, and have asked AtlasBase the question. We'll add text to this article when it replies.
AtlasBase does not say how the written DNA fragments are extracted and stored.
DNA Computation
What does “highly complex computational problems” mean? Decoding RSA encryption algorithms? No. It refers to applying “DNA-based computation to specialized optimization and data search problems that benefit from efficient massively parallel processing.”
AtlasBase says this uses million-way parallel processing in microliters of chemistry and you can solve complex, high-dimensional problems at scale while eliminating the massive energy footprint and hardware footprint of traditional supercomputers - and also, we suggest, eliminate their speed. As AtlasBase claims trillions of candidate states can be explored at one, this tradeoff will work when the amount of parallel computation is massive and not feasible.
AtasBase has an airline scheduling example: “Airline scheduling is a combinatorial problem: every route, aircraft, and time slot multiplies against every other. Conventional systems test the options in sequence. AtlasBase tests them all at once.
- Input - Developers input the airline's routing variables and constraints through the Molecular Expression Layer (MXL) software layer, which compiles those logic rules in standard code into an exact biological sequence blueprint.
- Translation - The AtlasBase Bio-Silicon Platform reads that MXL blueprint and synthesizes short, single-stranded DNA sequences representing the individual airport hubs, flight numbers, and time slots.
- DNA Printing - The AtlasBase Bio-Silicon Platform reads that MXL blueprint and synthesizes short, single-stranded DNA sequences representing the individual airport hubs, flight numbers, and time slots.
- Mixing - An automated fluidics system rehydrates the archived database powder and introduces these printed "query strands" into the same tube. Once mixed, the liquid itself executes the math through a proven molecular mechanism called DNA strand displacement — genuine in-memory computation.
- Chemical Logic Gates - Driven by chemical energy, the query strands move through the liquid. When a strand encounters its matching flight option, it binds at an exposed toehold and displaces the existing data strand through branch migration. That binding and displacing is the logic gate executing.
- Massively Parallel Scale - Because a milliliter of fluid contains a vast number of DNA strands reacting at the same time, the solution doesn't test one route at a time — it evaluates thousands to millions of candidate routes in parallel.
- Output - A selection step isolates the strands that encode valid, complete routes meeting the airline's constraints — the optimized flight schedules. A microfluidic sensor reads them out and returns the result as digital data.
The total processing time here would be similar to getting data off a tape library system and then processing it;
-
Locate and retrieve the DNA capsule from cartridge carrier
-
Input computation data through MXL SW
-
Synthesize DNA fragments
-
Rehydrate stored DNA from the capsule and add in new DNA fragments
-
Wait for a period of time.
-
Sequence the mixture and return the digital data result.
-
Return the cartridge carrier back to its store.
This is going to take several minutes, perhaps 30 or more overall. A similar computation could be done on a rack-scale GPU system but, AtlasBase would say, that involves data center-scale power, cooling, staffing, and space. Thalia molecular computing could be less expensive overall.
The company has graphics showing a molecular computer on its own and in racks.
Bill Baggelaar, former CTO of Sony Pictures Entertainment and co-author of the MovieLabs 2030 Vision, said: “Molecular storage can fundamentally change that by preserving our most valuable creative assets for generations, without the constant migration and maintenance that plagues today’s solutions. What’s especially exciting about AtlasBase is that it goes beyond archiving: it turns DNA into a molecular platform capable of both storing data and computing on it in entirely new ways.”
The technology is impressive and we will see what transpires. There’s no commitment here by a movie archive business to use AtlasBase technology. Which organization is going to take the first bite of the Thalia DNA molecular computing cherry?
Availability
AtlasBase is offering Early Access Programs for Molecular Compute and Molecular Data Storage with select partners and customers across its molecular computing platform. Initial applications include digital preservation, enterprise archives, and complex optimization challenges across industries such as media, financial services, logistics, manufacturing, scientific research, and resource exploration.
Bootnote
AtlasBase was founded in May 2025 in San Francisco when Twist Bioscience spun off its DNA storage business as Atlas Data Storage, a commercializing startup led by Varun Mehta, co-founder and CEO of HPE-acquired Nimble Storage. Bill Banyai, Twist co-founder and general manager of DNA data storage, became Atlas’s CTO.
Atlas Data Storage closed a $155 million seed financing round and licensed DNA storage assets from Twist, using them in its aim to develop end-to-end DNA storage. The core technology combines new “semiconductor chips and enzyme engineering, ushering in a new era of high-throughput and massively parallel chemistry performed on a chip.”
It renamed itself AtlasBase and appointed ex-NetApp Spot PC virtual desktop product GM Jeff Treuhaft as its CEO in March this year. NetApp sold Spot PC to HP in 2024. Initial CEO Varun Mehta left seven months in, with no fanfare, to join Premji Invest as a CXO board member in January 2026