The Matter42 vision
Everything we want next is made of something we can’t make yet.
Cooler computers. Sensors that read a single cell. Electronics that survive deep space, speak in light, and bend like paper. Every one of them is waiting on a material.
What becomes possible
01 / 06Computing, cooled
Intelligence without the heat.
Devices that switch with a fraction of today’s energy mean data centers that draw less power, shed less heat, and drink less water. AI keeps scaling; the grid doesn’t break.1
Data center cooling · Rsparks3, CC0
02 / 06Space
Electronics built for the dark.
A material a few atoms thick gives radiation almost nothing to hit. Lighter satellites, longer missions, and compute in orbit and beyond.2
Hubble after release from Atlantis, 2009 · NASA
03 / 06Medicine
Sensors that see a single cell.
Place one defect on purpose and a crystal becomes a quantum sensor. Only atoms thick, it can sit nanometers from living tissue: toward brain imaging without cryogenics, and magnetic resonance at the scale of a molecule.3
Magnetoencephalography scanner · NIMH
04 / 06Light
Computers that talk in light.
Single-layer semiconductors emit light efficiently, opening a path to on-chip optics that link tomorrow’s AI clusters at a fraction of the energy.
Photonics team configuring fibers · NASA Goddard
05 / 06Everywhere
Electronics on anything.
Grown cool enough for glass and plastic, the next materials can bend, flex, and disappear into implants, wearables, and everyday surfaces.4
Flexible display · U.S. Army RDECOM, CC BY 2.0
06 / 06Resilience
Know-how that travels.
Recipes that transfer across tools and sites let advanced manufacturing happen in more places. Critical knowledge stops living inside a handful of companies.
Aurora and city lights from the ISS, Expedition 43 · NASA
The catch
Every one of these is a materials problem.
Each future needs a material controlled at the scale of single atoms, over every square centimeter, run after run. We know what most of these materials are. We can’t yet make them on purpose.
Growth simulated atom by atom with Matter42’s kinetic Monte Carlo engine, drawn in the style of an electron micrograph (brightness tracks atomic number). Circled sites are the sulfur vacancies the run produced.
Discovery was never the bottleneck.
Making is.
The missing map 01
Nobody has the map.
Making a material means choosing temperature, pressure, chemistry, substrate, and timing: dozens of coupled knobs. Every lab walks one path through that space by trial and error.
The missing piece
The data to draw the map already exists. Labs pour out spectra, maps, and diffraction every day, then wait days for an expert to read them.
By the time the answer arrives, the next run has already started blind.
How we’ll make it
We’re mapping how matter is made.
Matter42 connects measurement, physics, and AI into one loop that learns the whole landscape instead of a single point. Where others find a recipe, we find the family.
01Measure
Read every measurement.
Atlas reads AFM, Raman, photoluminescence, cathodoluminescence, XRD and more with one engine: tracing islands, finding defects, mapping domains, and pulling out the number that matters.
Atlas › segment-afm-surface · 82 islands6 techniques · one engine- AFMWS₂ islands, MOCVD0.57 nmmonolayer step · 82 islands traced
- RamanWS₂, ion-beam patterned5.5 µmdamage halo found around every milled line
- CathodoluminescenceInGaN/GaN quantum well, 10 K1.4×10⁹ cm⁻²nonradiative defects · 85 located
- XRDGa₂O₃ on Si, annealing series19 nmβ-Ga₂O₃ crystallites, film crystallizes by 850 °C
- PhotoluminescenceMoSe₂–WSe₂ lateral heterostructure (CVD)82 meVWSe₂ shell around a MoSe₂ core, mapped from emission
- Atlas Physics AIWS₂ defects, simulated vs. measured5 defect typessimulated, then checked against 553 measured pixels
Data from five instruments and our physics models, each read by Atlas.
02Understand
Ground it in physics.
Atomistic-to-continuum models predict how a material will behave and explain why, then get checked against what was measured. Here, six simulated defect types meet 553 measured Raman linewidths.
Atomistic prediction × Raman measurement · WS₂6 defect models · 553 pxPredicted: E₂g linewidth vs defect concentration (MLIP phonons)Dashed line: median measured linewidth, 7.4 cm⁻¹ after removing instrument broadening and the resonant 2LA(M) band (13.5 → 7.7).
If every defect were: S→O substitution0.59 % – 1.21 %lowhighEvery curve states its limits0 K harmonic phonons, MLIP on WS₂ defect supercells, validated against DFT. A linewidth beyond a curve’s computed range is reported as a lower bound, not extrapolated.
Disagreement is reported, not hiddenAn independent LA(M)/A₁g intensity ratio estimate gives 3.7 ± 1.2 % (calibration transferred from MoS₂; order of magnitude only). Atlas shows both.
03Decide
Choose the next run.
Agents weigh every result against everything learned so far and recommend the next adjustment, so experts spend their time deciding, not deciphering.
How the work changes with Matter42 Today With Matter42 Days of manual analysis per result Interpretation in seconds Expert-dependent judgment Physics-grounded, repeatable recommendations Insight after the run Insight that shapes the next run Measurement as overhead Measurement as the engine 04Make
Run it. Learn. Repeat.
Every run sharpens the map, and every lab on the platform makes the models better for the next one. Our own closed-loop lab pushes the loop further than any single lab could alone.
60–100×5
faster iteration than manual, one-knob-at-a-time experimentation.
Others discover materials.
We make them manufacturable.
Why now
Three curves just crossed.
945 TWh1
Projected data center demand by 2030
The world needs new materials.
Compute, energy, and communications are scaling faster than the materials beneath them. Demand is outrunning the expertise to make what comes next.
60–100×5
Faster iteration than manual experimentation
The models finally work.
Physics-informed AI now reads spectroscopy and in-situ data the way an expert does, and closes the loop to the next run.
~100
Atomically thin semiconductors, most unmapped
The map is still blank.
The process know-how for atomic-scale materials is mostly unwritten. Whoever maps it first sets the rules for the next era.
What we believe
- 01
Measurement is the foundation of every breakthrough.
- 02
AI is strongest when it’s grounded in physics, not replacing it.
- 03
Experts should spend their time deciding, not deciphering.
- 04
New materials should reach the world in months, not decades.
The plan
Four steps from the lab bench to the next age.
- 1
Give every materials lab Atlas.
Every sample, measurement, and recipe, understood in one place.
Available today: explore Atlas - 2
Draw the map.
Connect process to property across thousands of experiments and our own closed-loop lab.
- 3
Make what no one else can.
Atomically thin materials, and the process know-how that makes them manufacturable.
- 4
Extend to all atomic-scale matter.
Computing, sensing, photonics, quantum. Every material a few atoms thick, made on purpose.
Atlas is how we learn. The map is what we own.
The materials are what change the world.
The people
Built by people who’ve measured matter at every scale.
Our founders’ work spans attosecond physics to reactor-scale transport, with more than 500 publications and about 20,000 citations between them. They’re joined by materials scientists, growth experts, and AI researchers, and advised by pioneers in 2D materials.
Meet the team
The moonshot
No material should wait a decade to exist.
Every lab and every line measures, understands, and adjusts in real time. New materials go from idea to the world in months. Yield is designed in, not discovered. That’s the world Matter42 is building, with scientists, engineers, partners, and investors who want to make matter do what it has never done.
Earthrise, Apollo 8, 1968 · NASA/Bill Anders
Notes
- 1.International Energy Agency, Energy and AI (2025): data center electricity demand projected near 945 TWh by 2030.
- 2.Roughly two orders of magnitude greater radiation tolerance than silicon; Matter42 whitepaper, “Beyond Moore” (2026).
- 3.Spin defects in atomically thin hexagonal boron nitride have been demonstrated as quantum sensors of magnetic field and temperature. Medical applications are a research direction, not a product.
- 4.Monolayer growth on polymer and glass substrates near 150 °C has been demonstrated; see Matter42 whitepaper, “Beyond Moore” (2026).
- 5.Closed-loop iteration speed versus manual, one-knob-at-a-time experimentation; Matter42 whitepaper, “Beyond Moore” (2026).
Photographs via Wikimedia Commons: NASA, NIMH, and Rsparks3 (public domain / CC0); flexible display by U.S. Army RDECOM, CC BY 2.0. The growth run is a kinetic Monte Carlo simulation; the process landscape is illustrative. Every other figure is measured data.