Breaking
NEW CEO Alex Wang takes the helm at NobleAI, Dec 2025 FUNDING ~$40M raised - M12, Chevron, Sway, Dorilton RECOGNITION Named a top-10 AI startup to watch, 2024 PLATFORM VIP - Visualizations, Insights & Predictions THESIS Science-Based AI that learns from sparse data PARTNERS Solvay, Microsoft, Chevron Technology Ventures NEW CEO Alex Wang takes the helm at NobleAI, Dec 2025 FUNDING ~$40M raised - M12, Chevron, Sway, Dorilton RECOGNITION Named a top-10 AI startup to watch, 2024 PLATFORM VIP - Visualizations, Insights & Predictions THESIS Science-Based AI that learns from sparse data PARTNERS Solvay, Microsoft, Chevron Technology Ventures
Company Profile // San Francisco, California // Est. 2017

NobleAI

The startup betting that less data is the whole point.

Science-Based AI for chemistry, materials and energy - models that carry the laws of physics with them so they can predict from the sparse, imperfect data that industrial labs actually have.

Science-Based AI Enterprise SaaS R&D Software ~46 people
NobleAI company logo
The wordmark, plain and unglamorous - which is roughly how the company likes it. A physicist's logo for a physicist's product: the AI that reads chemistry before it reads your spreadsheet.
Share this profile
The Feature

A Very Deliberate Argument With the Rest of AI

Here is a thing everyone in artificial intelligence agrees on, which is that artificial intelligence needs data, and ideally a preposterous amount of it. The models that write your emails and misidentify your cat were trained on functionally the entire internet. The working assumption of the field is that if your model is not very good, the answer is more data, and if you do not have more data, you have a problem.

NobleAI, a roughly-46-person company headquartered at 345 California Street in San Francisco, has built an entire business on the position that this assumption, while broadly correct, is useless to the people it wants as customers. Those customers are chemical, materials and energy companies. And chemical, materials and energy companies do not have oceans of data. They have a lab, a handful of expensive experiments, some proprietary spreadsheets nobody wants to email around, and a very pointed question: will this formulation work, and can we sell it?

You cannot solve that with more data, because there is no more data. Running the experiment is the expensive thing you were trying to avoid in the first place. So NobleAI's pitch - which it has branded, with admirable directness, "Science-Based AI" - is to put the missing information somewhere other than the dataset. Specifically, into the model. Its models are built to carry physical laws and chemical properties around with them, so that when they encounter a small, sparse, real-world dataset, they already know a great deal about how chemistry is allowed to behave.

"R&D at the speed of business - from months to minutes." It is a slogan, but the interesting part is that it is aimed at a stopwatch, not a leaderboard.

This is a subtle bit of positioning and it is worth sitting with, because it explains almost everything else about the company. A normal AI vendor sells you accuracy - our model is better than their model on some benchmark. NobleAI is, in effect, selling relief from a constraint. If you have too little data and too much physics to ignore, the ordinary big-data approach does not fail gracefully; it just does not work. NobleAI's whole reason to exist is that gap.

The product that carries this is the VIP Platform, which stands for Visualizations, Insights and Predictions, and which is the kind of acronym that could only have been named by people who genuinely believe those are the three things you want. You upload your messy multi-format data. The platform manages it, builds a Science-Based AI model, and then lets your scientists do the useful things: run parameter sweeps, ask for an optimized design that hits a performance or sustainability or cost target, and - the part that makes chemists cautious people less cautious - see explanations. The predictions come with uncertainty estimates and feature-impact analysis. This matters more than it sounds. The fastest way to lose a scientist is to hand them a black box and ask them to bet a production run on it.

There is also a family of more specific products sitting on top of a SaaS engine called NobleReactor - NobleChemistry for formulation work, NobleBattery aimed at battery-material discovery - plus a solution with the wonderfully bureaucratic name Risk Assessment & Ingredient Replacement, or RAIR, which scans a product portfolio for risky ingredients and proposes safer, compliant alternatives. RAIR is a good example of the company reading the room. PFAS bans, additive-safety reviews, regulators moving faster than reformulation cycles - all of that is a headache for a chemical company and, if you squint, a software feature. NobleAI squinted.

The company sells to labs that don't have big data and can't easily get it. That is not a niche. That is most of industrial R&D.

The founding story fits the product. NobleAI was started in 2017 by Matthew C. Levy, who is not a career software executive but a physicist, with a PhD from Rice, which is roughly what you would guess about a company whose central idea is to smuggle physics into a machine-learning model. For most of its life the interesting question about a startup like this is whether anyone serious will actually use it. NobleAI got an unusually clean answer early: in 2019 it announced funding alongside a working relationship with Solvay, the Belgian chemical giant. When one of the largest chemical companies in the world is both an early partner and an investor, that is about as strong a signal of product-market fit as this corner of software produces.

The investor list kept saying the same thing. The 2023 Series A, more than $17 million, was led by M12, which is Microsoft's venture arm; Chevron Technology Ventures also put money in. In 2024 the company added a Series A extension of more than $10 million led by Sway Ventures, with Dorilton, bringing total funding to roughly $40 million. The pattern worth noticing is that NobleAI's investors and its customers keep overlapping - Microsoft, an energy major, a chemical major. When the people writing checks are also the people who would use the thing, you are usually looking at genuine pull rather than a good pitch deck. That same year, NobleAI was named one of the ten hottest AI startups to watch, which is the sort of list that is easy to be cynical about and still nice to be on.

In December 2025 the company did the thing that growing startups eventually do, which is change who runs it. Alex Wang, a former Senior Vice President of Strategy and Corporate Development at VMware with a couple of decades of experience scaling technology companies, was named CEO, succeeding Sunil Sanghavi, who moved to the board. This is a founder-and-early-leadership-to-operator transition, and it tends to signal that a company thinks its next problem is less "can we build the thing" and more "can we sell a lot of it." Given that NobleAI's models are pitched as deployment-ready in about a month - time-to-first-value being the currency that actually moves conservative industrial buyers - that is a reasonable bet about what comes next.

What can you actually do with NobleAI, then, if you are a person and not a chemical conglomerate? Not much - and that is the honest and slightly charming thing about it. This is not a consumer product. It is software for the people who figure out what goes into your battery, your coating, your packaging, your fabric, the additives in your food. Its whole appeal lives in an unglamorous corner of AI where there are no viral demos, just the quiet and lucrative work of helping a scientist decide, faster and with a stated degree of confidence, which experiment is worth running next. There is a lot of value hiding in that corner. NobleAI decided to go stand in it.

By the Numbers

NobleAI, Tallied

2017
Founded
~$40M
Total Funding
~46
Employees
1 mo.
To Deployment
What They Build

The Product Shelf

Platform

VIP Platform

Visualizations, Insights & Predictions. Data and model management, parameter sweeps, optimized design generation and rich visualizations for R&D teams.

Core Tech

Science-Based AI (SBAI)

Models that embed physical laws and chemical properties so they can learn from small, sparse datasets - private by design, deployment-ready in about a month.

Engine

NobleReactor

The SaaS engine that powers NobleAI's chemistry and materials informatics solutions.

Application

NobleChemistry

Chemical formulation and product-development workflows built on the NobleReactor platform.

Application

NobleBattery

Focused on accelerating battery-material discovery and development.

Compliance

RAIR

Risk Assessment & Ingredient Replacement - flags risky ingredients and proposes safer, compliant, more sustainable alternatives (e.g. PFAS replacement).

Follow the Money

Funding, Round by Round

Seed · 2019
~$8.6M
Series A · 2023
$17M+
A Extension · 2024
$10M+
Total Raised
~$40M

Bars scaled to total raised. Investors include M12 (Microsoft), Chevron Technology Ventures, Sway Ventures, Dorilton and Solvay. Figures are public and approximate.

In Their Words

The House Style

"Science-Based AI for real-world predictions."

"R&D at the speed of business - from months to minutes."

The Record

A Short History

2017

NobleAI founded

Physicist Matthew C. Levy starts the company in San Francisco to apply AI to chemistry, materials and energy R&D.

2019

Seed round & Solvay partnership

Announces ~$8.6M in venture funding and a working relationship with chemical giant Solvay.

2023

Series A & Azure launch

Closes $17M+ led by Microsoft's M12 and launches informatics solutions on the Azure Marketplace.

2024

Extension & recognition

Raises a $10M+ Series A extension led by Sway Ventures and is named a top-10 AI startup to watch.

2025

New CEO

Appoints Alex Wang as CEO to lead the next phase of growth; Sunil Sanghavi moves to the board.

Reader Questions

FAQ

What does NobleAI do?

It builds "Science-Based AI" software that helps chemical, materials and energy companies predict product performance, optimize formulations and speed up R&D using small, real-world datasets.

Who founded NobleAI and who runs it now?

It was founded in 2017 by physicist Matthew C. Levy. In December 2025, Alex Wang was appointed CEO, succeeding Sunil Sanghavi, who moved to the board.

How much funding has NobleAI raised?

Roughly $40M in total, including a $17M+ Series A led by Microsoft's M12 (2023) and a $10M+ Series A extension led by Sway Ventures (2024).

What makes NobleAI's approach different?

Its models embed physical laws and chemical properties so they can make useful predictions from sparse datasets, rather than requiring the massive data typical AI depends on.

Who are NobleAI's customers?

R&D and product-development teams at chemical, materials, energy, consumer-goods, food & beverage, semiconductor and building-materials companies; early partners include Solvay.