COMPANY PROFILE / MATERIALS INFORMATICS
Orbital Industries (formerly Orbital Materials)
It releases its materials-AI model "Orb" openly, and uses it as a tool for building its own cooling fluids and data-centre products
Orbital Industries was founded in the UK in 2022 and was known until 2024 as Orbital Materials. It has released Orb, an AI model that predicts how atoms behave, in open form under Apache-2.0, while the company itself calls itself an "AI Industrial" and centres its business on building its own products (cooling fluids and modular data centres for data centres), using AI across materials, equipment design and manufacturing. In August 2026 it licensed its in-house software CurieOS to BASF's catalyst division (BASF ECMS). This profile concentrates on the materials-AI side and uses official sources only.
- Orbital Industries in 30 seconds, and where it stands in materials informatics
- The logo and how the name is written
- Where the company sits in materials informatics
- Scale
- Resources committed to materials informatics
- What it has done so far
- What comes next
- What the company is aiming for
- Map of partnerships
- Technology, products and services
- The risks this company carries
- Glossary / References / Claim-to-source audit
1. Orbital Industries in 30 seconds, and where it stands in materials informatics
Industrial users (secondary)Our classification; see Fig. 1
This company releases materials-AI models openly and also makes and sells its own products, with the latter at the centre of its business Sourced. In a January 2025 blog post the CEO argued that advanced materials are features, not products, and that owning an AI that discovers materials is hard to turn into profit on its own; value comes from building discovered materials into products. He says the company uses independently benchmarked AI models in-house to discover materials, and builds them into hardware it designs.
So although it too is an "MI company", its character differs from companies whose main product is a model or software. Sales and contracts for its products, however, cannot be confirmed from official sources (Chapter 4).
2. The logo and how the name is written
A note on names: the official website now says Orbital Industries. Announcements up to 2024 used Orbital Materials (the website domain was orbitalmaterials.com, which now redirects to orbitalindustries.com). The UK register, when checked on 22 October 2026, still shows the name ORBITAL MATERIALS LTD. The logo was taken from the SVG embedded in the official website, with its colour fixed to black.
Official site: https://www.orbitalindustries.com/| Name | What it is | Note |
|---|---|---|
| Orbital Industries | The current brand and the name the company uses | Formerly Orbital Materials |
| ORBITAL MATERIALS LTD | The registered legal name in the UK | Company number 14332843. Register checked on 22 October 2026 |
| Orbital IT | The company's own brand for data centres | Launched 27 May 2026 (cooling and modular data centres) |
| Orb / OrbMol | Names of AI models (Orb for crystals and similar, OrbMol for molecules) | Released openly |
| LINUS | The name of the in-house foundation model (Orb is a tuned version of it) | From the 2024 announcements. Not public |
| CurieOS | Agentic-AI platform software for science and engineering | Licensed to BASF ECMS in August 2026 |
3. Where the company sits in materials informatics
In the materials informatics (MI) flow (data, then prediction and design, then narrowing by calculation, then confirmation by experiment), Orbital builds and releases an AI model that predicts atomic behaviour quickly (a prediction tool). What stands out is that it then uses the results in developing its own cooling fluids and equipment (experiment and manufacturing).
The next illustration shows the idea the company describes of putting "materials, equipment design and manufacturing" into one loop, and what supports each part.
4. Scale
Revenue and profit are not published as far as we could confirm. The table shows what the register and company announcements let us confirm Sourced.
| Item | Detail | Source |
|---|---|---|
| Legal entity | ORBITAL MATERIALS LTD (UK private limited company, number 14332843). Status: active. Business class: research and experimental development in natural sciences and engineering | Companies House (UK) |
| Incorporated | 2 September 2022 (the company says it began at the end of 2022) | Companies House / official posts |
| Registered office | 10 John Street, London, WC1N 2EB, UK | Companies House |
| Sites | London and San Francisco. A 2024 announcement describes a materials research facility in Princeton, New Jersey (opened in the first quarter of 2024) | Official announcements |
| Employees | About 50 (May 2026 announcement) | Official press release |
| Revenue and profit | Not disclosed | - |
| Listing | Unlisted | - |
| Funding (official) | Series B of US$50 million (28 May 2026, led by Plural). NVentures (NVIDIA) invested in October 2024 (amount not disclosed) | Official press release |
The May 2026 announcement also gives figures such as the size of the data-centre market, but no source is shown for them, so this article does not use them. We could not find an official announcement of any funding before the Series B.
5. Resources committed to materials informatics
| Angle | What we could confirm | Type |
|---|---|---|
| People | About 50 employees (May 2026). The co-founders are CEO Jonathan Godwin (said to have worked on AI for science, engineering and materials at DeepMind for five years), CTO James Gin-Pollock and COO Daniel Miodovnik. The number dedicated to MI is not disclosed | Sourced |
| Money (R&D) | R&D spend and capital expenditure for MI are not disclosed | Not disclosed |
| Money (funding) | Series B of US$50 million (led by Plural; NVentures, Radical Ventures, Compound and Fly Ventures took part). The amount of the October 2024 NVentures investment is not disclosed | Sourced |
| Investments and acquisitions | No investment in or acquisition of another company by Orbital Industries could be confirmed | Not confirmed |
| Universities, institutes, public funding | The lead author of OrbMol did the work as an intern at the company (a note in the official post). It used the public OMol25 dataset for training. Orb can also be used through the outside interactive tool Rowan. No public funding such as DOE or NSF could be confirmed | Sourced |
| Cloud and compute | A multi-year partnership with AWS (SageMaker HyperPod for pre-training; evaluating Trainium). It uses NVIDIA ALCHEMI tools to speed up inference (OrbMol-v2 post) | Sourced |
Status of the main models, products and facilities
| Name | Kind | Main specifications (company-stated) | Form of release and licence | Stage and date |
|---|---|---|---|---|
| Orb v1 and v2 | Interatomic potential for crystals | v1: said to be more accurate than the then Google and Microsoft models (September 2024; details of the comparison are not in the announcement text). v2 has a paper (arXiv) | The September 2024 blog described a licence "free for non-commercial uses and startups". The current GitHub says Apache-2.0 | Released (September and October 2024) |
| Orb-v3 | For crystals (several models with different accuracy and speed) | The sparse direct model (orb-v3-direct-20) is 10 times faster than MACE-MPA-0 with one fifth of the memory (company statement) | Code and weights released under Apache-2.0 | Released (paper April 2025, blog 4 September 2025) |
| OrbMol | For molecules (charge and spin can be given as input) | Trained on OMol25 (over 100 million DFT calculations) | Apache-2.0. A note says it has not been carefully tested on periodic systems (crystals) | Released (27 August 2025) |
| OrbMol-v2 | For molecules (adds long-range electrostatics) | Normalised error ratio (NER) of 1.62 on GSCDB; said to be 21% more accurate than the next-best MLIP (conditions in Chapter 10) | Released under Apache-2.0 | Released (May to 2 June 2026) |
| LINUS | In-house foundation model | The company says Orb is a tuned version of it | Not public | Described in 2024 only |
| CurieOS | Agentic-AI platform for science and engineering | Provides a shared data layer (context graph), workflows and apps together. Says the AI provider can be chosen | Commercial licence. BASF ECMS has deployed it | Announced 5 and 11 August 2026. Open to "select partners" |
| Own facilities | The company's own lab and manufacturing | A 2024 announcement mentions a materials research facility in Princeton (opened in Q1 2024). Its size and current operation could not be confirmed | - | Not confirmed |
The Orb code and weights are public, so you can run them on your own machines and try them on your own systems Sourced. From the official notes, here is what a materials engineer should know.
1. Which model. Orb-v3 comes in several variants, such as "conservative" models that take forces from the derivative of energy and "direct" models that predict forces directly, with different accuracy, speed and memory. The official README suggests starting initial tests with the most accurate setup (conservative, 120 neighbours, highest precision). 2. The range of OrbMol. OrbMol's training data contain no periodic systems and it is noted as not carefully tested on them; it is meant for molecules. 3. The claim of "100,000 atoms on one GPU". The company says it runs at scales where others fail, but the announcement does not give the details of the comparison, and we could not confirm them. For large systems, measure speed and memory on your own system before deciding.
6. What it has done so far
| When | What happened | Meaning |
|---|---|---|
| September 2022 | ORBITAL MATERIALS LTD incorporated in the UK (register) | The company's starting point |
| Q1 2024 | Opened a materials research facility in Princeton (company statement) | A set-up to make and test materials itself |
| September 2024 | Released Orb. The company says it is more accurate than Google and Microsoft models and 5 times faster for large-scale calculations (details of the comparison are not in the announcement text) | An open materials-AI model released |
| October 2024 | Investment from NVentures (NVIDIA; amount not disclosed). Orb-v2 paper | More compute and team |
| December 2024 | Multi-year strategic partnership with AWS. Orb to be generally available on AWS SageMaker JumpStart and Marketplace. A plan to deploy and test carbon-removal technology by the end of 2025 is stated | Cloud availability. A later official report on the plan could not be found |
| January 2025 | The CEO set out the business approach in a blog ("advanced materials are features, not products") | Business approach explained |
| April to September 2025 | Released Orb-v3 (Apache-2.0). Released OrbMol in August | Model family widened |
| May 2026 | Launched Orbital IT (27 May). Series B of US$50 million (28 May) | Toward its own data-centre products |
| June 2026 | Released OrbMol-v2. Announced collaboration with NVIDIA on DSX AI Factory (8 June) | Model update and a product-side collaboration |
| August 2026 | BASF ECMS deployed CurieOS (5 August). Published an article introducing CurieOS (11 August) | The first case of in-house software offered outside |
Note: older posts on the official site (such as the December 2024 AWS post) are written under the current name "Orbital Industries". We could not confirm from official sources when the name actually changed. The posts of September and October 2024 still use the then name "Orbital Materials".
7. What comes next
All of these are the company's plans and targets; their execution and results could not be confirmed Sourced Not yet confirmed.
None of the following is stated in official sources Not yet confirmed.
Revenue and contract amounts; the result of deploying and testing the carbon-removal technology set out in 2024; the date of the name change; details of in-house models such as LINUS; and measured data for materials CurieOS found (a PFAS-free cooling dielectric is said to have been found, but no measurements are published).
8. What the company is aiming for in materials informatics
This part includes things the company has not said directly. It sets out inferences we draw from primary sources, with the evidence.
| What we read from it | Primary-source basis | Type |
|---|---|---|
| It treats materials AI as "a tool for building its own products" more than "a product to sell" | The CEO's January 2025 blog (a view that licensing a materials-discovery AI or selling software seats is hard to turn into a business). Its own cooling fluids and data centres are the way in | Sourced (company view) |
| It keeps releasing Orb to gather users, talent and credibility | Release under Apache-2.0, outside routes such as AWS and Rowan, and statements that a small team can compete | Inference |
| CurieOS is the first attempt to extend an in-house tool outward | In August 2026 it licensed it to BASF ECMS and said it would open it to other partners. No outside offering had been mentioned before | Inference |
| It leans heavily on the NVIDIA ecosystem | NVentures investment (2024), use of ALCHEMI tools (OrbMol-v2), collaboration on DSX AI Factory (2026) | Inference |
| It is widening the targets of materials search to industrial catalysts | BASF ECMS is the automotive emissions-catalyst division. The contract is to use CurieOS to explore catalysts | Sourced |
The CEO's blog describes how hard an "AI materials-discovery platform" is as a business and concludes that value comes only when a discovered material is built into a product Sourced. This fits how the company has widened its scope step by step: cooling fluids (materials), then cooling equipment (hardware), then modular data centres (manufacturing).
But the August 2026 contract with BASF ECMS licenses CurieOS to an outside party. We read this as a move to offer outside a tool built for its own product work Inference. The contract value and scope are not disclosed, so we cannot tell what weight it carries in the business as a whole.
9. Map of partnerships in materials informatics
| Counterparty | Relationship | Type |
|---|---|---|
| AWS | Multi-year strategic partnership in December 2024. Orb to be offered on SageMaker JumpStart and Marketplace (as announced). Joint evaluation of data-centre carbon-removal and cooling technologies | Sourced (we did not check that it went live) |
| NVIDIA | NVentures invested in October 2024. Uses ALCHEMI Toolkit-ops in OrbMol-v2. Announced collaboration on DSX AI Factory in June 2026 | Sourced |
| BASF ECMS | On 5 August 2026 a CurieOS licence was announced, to be used in research and development of automotive emissions catalysts | Sourced |
| Rowan | "Launch partner" for a service that lets Orb-v3 be used interactively | Sourced |
| Plural and others | Led the Series B (Plural). Radical Ventures, Compound and Fly Ventures also took part | Sourced |
10. Technology, products and services
What is public is the models and code of the Orb series, and CurieOS, which the company offers to partners it chooses. Its own products such as cooling fluids are sold separately from MI, as products for data centres.
| Item | What the company shows | Conditions and cautions |
|---|---|---|
| Speed and scale of Orb | "The only model that can simulate 100,000 atoms on a single GPU, where every competitor crashes" and "ten times faster than the nearest alternative" (BASF ECMS and Series B announcements) | Cites the Orb-v3 paper (arXiv:2504.06231). The comparison targets and conditions are not given in the announcement text and could not be confirmed. A company claim |
| "Predictions don't drift" | Says independent benchmarks show predictions do not drift or err over long runs | Cites LAMBench (npj Computational Materials 2026). We have not checked its content |
| Orb-v3: lightweight models | The sparse direct model is 10 times faster than MACE-MPA-0 with one fifth of the memory, and 40 times faster than SevenNet with one fortieth of the memory (company technical blog) | The company's measurements under the conditions in that blog |
| OrbMol-v2 | On GSCDB (a broad benchmark for molecular DFT), the normalised error ratio (NER) over 5,152 reactions without single atoms is 1.62. It beats hybrid functionals (PBE0, B3LYP and others) and is 21% more accurate than the next-best MLIP; about 5 times faster than the comparison models at 10,000 atoms (one H100) | NER 1.0 means matching the best DFT functional. Tasks involving single atoms are excluded. The company's own evaluation |
| Cooling dielectric | A PFAS-free cooling liquid found with AI (part of the Orbital IT product) | Measured values and certification are not published. Described as "in development" |
Note: every figure is a company statement. We could not find reproduction by a third party or confirmation on real equipment.
1. The public models can be tried yourself. Orb is released so that it can be plugged in as a calculator for ASE (the standard Python environment for atomistic simulation), as the README says. You can use it on your own structures to compare against DFT results or existing models.
2. The company's main uses are catalysts and cooling fluids. The BASF ECMS contract is for catalyst search, a use that fits atomic-level calculations of surfaces and adsorption. But what materials were found has not yet been published.
3. How to read "AI discovers a new material". The cooling dielectric is a company claim, and computational prediction and experimental confirmation are not shown separately. Until product performance and certification are published, it is safest to treat it as a company claim.
11. The risks this company carries
| Risk | Detail | Type |
|---|---|---|
| The centre of the business lies outside MI | By the company's own description, the core is hardware for data centres. Release of materials-AI models continues, but how much it invests in MI is not disclosed | Sourced |
| Conditions behind performance claims are hard to see | The targets and conditions of comparisons behind "only" and "ten times" are not shown in the announcement text | Sourced |
| Following up earlier plans | The result of deploying and testing the carbon-removal technology set out in 2024 could not be found in later official announcements | Sourced |
| Confusion over names and entities | The old name, the current name and the registered name differ, and the date of the change is unknown. Easy to misread when searching for sources | Sourced |
| Outside offering is only starting | CurieOS has one published outside case, BASF ECMS, and the contract is not disclosed | Not yet confirmed |
| Reliance on a small team | About 50 employees, working on a wide range at once (materials, equipment, manufacturing and software) | Inference |
| Many open competitors | Others also release similar models. If the difference lies in speed or memory, the comparison conditions matter | Inference |
12. Glossary
- Machine-learning interatomic potential (MLIP)
- An AI model that predicts the force on each atom and the energy from how atoms are arranged. Used instead of DFT to gain speed.
- Conservative / direct model
- A difference in whether forces come from the derivative of energy (conservative) or are predicted directly (direct). The latter is faster but can differ in physical consistency.
- DFT
- Density functional theory. The standard way to calculate the properties of matter from quantum mechanics: precise but slow.
- Matbench Discovery
- A public benchmark that compares MLIPs on predicting the stability of crystals.
- AI Industrial
- A category the company gives itself: a firm that builds AI into materials, design and manufacturing and makes its own physical products.
- CurieOS
- Agentic-AI platform software for science and engineering developed by the company.
- Dielectric coolant
- A cooling liquid that does not conduct electricity. Used to cool server chips directly.
- PFAS
- A group of persistent man-made organofluorine compounds. Regulation is tightening.
13. References
- Orbital Industries Official website (home). The "AI Industrial" positioning and links to CurieOS and Orbital IT. https://www.orbitalindustries.com/
- Orbital Industries About. Description of the business (materials discovery, hardware design and manufacturing in one loop). https://www.orbitalindustries.com/about
- Orbital Industries AI Research. Official posts on the Orb series, OrbMol and OrbMol-v2. https://www.orbitalindustries.com/ai-research
- Orbital Industries Press release, "Open-sourcing Orb" (3 September 2024, under the then name Orbital Materials). https://www.orbitalindustries.com/news/announcements/press-release-orb
- Orbital Industries CEO blog, "Open-sourcing Orb" (2 September 2024). The licence as described at the time. https://www.orbitalindustries.com/news/blog/introducing-orb
- Orbital Industries Press release, "Investment from NVentures (NVIDIA)" (22 October 2024, under the then name Orbital Materials). https://www.orbitalindustries.com/news/announcements/press-release-orbital-materials-to-advance-materials-and-climate-technologies-with-ai-using-investment-from-nventures
- Orbital Industries Press release, "Multi-year strategic partnership with AWS" (posted 5 December 2024). General availability of Orb and use of SageMaker. https://www.orbitalindustries.com/news/announcements/orbital-materials-and-aws-enter-strategic-partnership-to-develop-technologies-for-data-center-decarbonization-and-efficiency
- Orbital Industries Technical blog, "Orb-v3" (4 September 2025). Place of each model in the family and release under Apache-2.0. https://www.orbitalindustries.com/news/blog/orb-v3-atomistic-simulation-at-scale
- arXiv Rhodes et al., "Orb-v3: atomistic simulation at scale", arXiv:2504.06231 (April 2025). https://arxiv.org/abs/2504.06231
- Orbital Industries Announcement, "OrbMol: Extending Orb to Molecular Systems" (27 August 2025). Trained on OMol25, Apache-2.0. https://www.orbitalindustries.com/news/announcements/orbmol-extending-orb-to-molecular-systems
- Orbital Industries Technical blog, "OrbMol-v2" (2 June 2026). Evaluation conditions on GSCDB. https://www.orbitalindustries.com/news/blog/introducing-orbmol-v2
- Orbital Industries Announcement, "Introducing Orbital IT" (27 May 2026). The data-centre brand and its products. https://www.orbitalindustries.com/news/announcements/introducing-orbital-it
- Orbital Industries Press release, "Series B of US$50 million" (28 May 2026). Investors, headcount and description of the business. https://www.orbitalindustries.com/news/press/orbital-industries-series-b-50-million
- Orbital Industries Press release, "Teaming with NVIDIA on DSX AI Factory" (8 June 2026). https://www.orbitalindustries.com/news/press/orbital-industries-partners-nvidia-dsx-ai-factory-infrastructure
- Orbital Industries Press release, "BASF ECMS deploys CurieOS" (5 August 2026). https://www.orbitalindustries.com/news/press/basf-ecms-ai-materials-discovery-platform
- Orbital Industries Announcement, "Introducing CurieOS" (11 August 2026, article by the CTO). Design and use of CurieOS. https://www.orbitalindustries.com/news/announcements/introducing-curieos
- GitHub README, MODELS and LICENSE of orbital-materials/orb-models (Apache-2.0). Model list and cautions. https://github.com/orbital-materials/orb-models
- Companies House (UK) Public record for ORBITAL MATERIALS LTD (company number 14332843). https://find-and-update.company-information.service.gov.uk/company/14332843
- arXiv Neumann et al., "Orb: A Fast, Scalable Neural Network Potential", arXiv:2410.22570 (October 2024). https://arxiv.org/abs/2410.22570
- npj Computational Materials Han et al., "LAMBench: a benchmark for large atomistic models", 12 (2026). Cited by the company in its Series B release as an "independent benchmark". https://doi.org/10.1038/s41524-025-01929-3
- Orbital Industries CEO blog, "Advanced Materials Are Features Not Products" (11 January 2025). The company's statement of its business approach. https://www.orbitalindustries.com/news/blog/advanced-materials-are-features-not-product
14. Claim-to-source audit
| Claim in the text | Type | Source |
|---|---|---|
| Registered name (ORBITAL MATERIALS LTD), incorporation date, address and status | Sourced | Reference 17 https://github.com/orbital-materials/orb-models |
| The current name Orbital Industries (formerly Orbital Materials), and that the change date is unconfirmed | Sourced | The change date is not in official sources. Reference 1, 2, 4, 7 https://www.orbitalindustries.com/ https://www.orbitalindustries.com/about https://www.orbitalindustries.com/news/announcements/press-release-orb https://www.orbitalindustries.com/news/announcements/orbital-materials-and-aws-enter-strategic-partnership-to-develop-technologies-for-data-center-decarbonization-and-efficiency |
| Release of Orb (September 2024), the description at the time and the licence statement | Sourced | Company statement. Reference 4, 5 https://www.orbitalindustries.com/news/announcements/press-release-orb https://www.orbitalindustries.com/news/blog/introducing-orb |
| Investment from NVentures (October 2024, amount not disclosed) | Sourced | Reference 6 https://www.orbitalindustries.com/news/announcements/press-release-orbital-materials-to-advance-materials-and-climate-technologies-with-ai-using-investment-from-nventures |
| The multi-year partnership with AWS (December 2024) and the carbon-removal plan | Sourced | Company plan. Reference 7 https://www.orbitalindustries.com/news/announcements/orbital-materials-and-aws-enter-strategic-partnership-to-develop-technologies-for-data-center-decarbonization-and-efficiency |
| Release of Orb-v3 (Apache-2.0) and the speed and memory comparison of lightweight models | Sourced | Company measurement conditions. Reference 8, 9, 16 https://www.orbitalindustries.com/news/blog/orb-v3-atomistic-simulation-at-scale https://arxiv.org/abs/2504.06231 https://www.orbitalindustries.com/news/announcements/introducing-curieos |
| The Orb v2 paper | Sourced | Reference 18 https://find-and-update.company-information.service.gov.uk/company/14332843 |
| OrbMol (August 2025): training data, licence and the note on periodic systems | Sourced | Reference 10, 16 https://www.orbitalindustries.com/news/announcements/orbmol-extending-orb-to-molecular-systems https://www.orbitalindustries.com/news/announcements/introducing-curieos |
| OrbMol-v2 evaluation (GSCDB, NER 1.62, 5,152 reactions, speed on an H100) | Sourced | Company evaluation. Reference 11 https://www.orbitalindustries.com/news/blog/introducing-orbmol-v2 |
| Launch of Orbital IT (27 May 2026) and its products | Sourced | Reference 12 https://www.orbitalindustries.com/news/announcements/introducing-orbital-it |
| Series B of US$50 million, investors, about 50 people and sites | Sourced | Company announcement. Reference 13 https://www.orbitalindustries.com/news/press/orbital-industries-series-b-50-million |
| Collaboration with NVIDIA on DSX (June 2026) | Sourced | Reference 14 https://www.orbitalindustries.com/news/press/orbital-industries-partners-nvidia-dsx-ai-factory-infrastructure |
| The CurieOS licence to BASF ECMS (5 August 2026) | Sourced | Reference 15 https://www.orbitalindustries.com/news/press/basf-ecms-ai-materials-discovery-platform |
| The design of CurieOS and the explanation that requirement changes update within days | Sourced | Company description. Reference 16 https://www.orbitalindustries.com/news/announcements/introducing-curieos |
| The claims of "100,000 atoms on one GPU", "ten times faster" and "predictions don't drift" | Sourced | Company claims. Conditions not given in the announcement text. Reference 9, 13, 19 https://arxiv.org/abs/2504.06231 https://www.orbitalindustries.com/news/press/orbital-industries-series-b-50-million https://arxiv.org/abs/2410.22570 |
| The CEO's business approach (advanced materials are features, not products) | Sourced | Company view. Reference 20 https://doi.org/10.1038/s41524-025-01929-3 |
| Revenue, contract amounts, date of the name change, result of carbon-removal tests, size of own facilities, measurements of the cooling fluid | Not yet confirmed | Could not be confirmed from official sources. Not estimated in this article Reference 1, 13 https://www.orbitalindustries.com/ https://www.orbitalindustries.com/news/press/orbital-industries-series-b-50-million |
| Orbital IT roll-out, two-phase cooling, other industries and more CurieOS partners | Not yet confirmed | Company plan. Reference 12, 13, 16 https://www.orbitalindustries.com/news/announcements/introducing-orbital-it https://www.orbitalindustries.com/news/press/orbital-industries-series-b-50-million https://www.orbitalindustries.com/news/announcements/introducing-curieos |
| The four types of MI company, where Orbital Industries sits, and the "one loop" illustration | Our calculation | Our own organisation |
| "Materials AI as a tool for own products", "CurieOS as the first outward move", "leaning on NVIDIA" and similar readings | Inference | Our own reading. Evidence is in Chapter 8 |
| The risks of the business centre, opaque comparison conditions and name confusion | Inference | Our own reading. Evidence is in Chapter 11 |
Last updated: 22 October 2026 / Troy Technical
The facts in this article rest on the official website and posts of Orbital Industries, GitHub, published papers and the UK Companies House record. We have not used market estimates from research firms or market-size figures with no stated source. Performance and comparison figures are company statements, not independent third-party verification. Passages marked "Inference" are our own reading of primary sources. The article does not touch share prices, valuations or investment judgements.