Why Writing a Software or AI Check can be a Materials Innovation Bet
Matt Saunders, July 2026
As a VC investing in packaging and materials innovation, I get a version of the same question a lot: why are you looking at software?
The short answer: it’s wrong to put an ‘or’ in software or materials. Framing software and materials as separate investment categories misses what’s actually happening in the space.
At New Earth Ventures (NEV), we’re singularly focused on accelerating the adoption of new materials — across packaging, supply chain, and the infrastructure that connects the two. When I talk to partners and collaborators in our ecosystem, they naturally picture the physical stuff — biopolymers, fiber-based substrates, next-generation barrier coatings. And yes, that’s part of it. But a significant share of what we back is software and AI. Those aren’t separate bets.
They’re the same bet, expressed differently.
Here’s what I mean.
The problem with materials has never been ideas. It’s time.
The traditional materials R&D pipeline is brutal. A researcher has a promising hypothesis. She runs experiments. She waits. She tweaks. She waits some more. Getting from initial discovery to something commercially viable has historically taken 20 years or more and hundreds of millions of dollars. Some estimates put the timeline for truly transformative new materials at closer to 50 years from lab to widespread adoption — and that pattern, according to researchers who study it, is “shockingly consistent.”
That timeline is why so many promising materials never make it out of the lab. It’s not that graphene, perovskite solar cells, or metal-organic frameworks aren’t interesting — they are. It’s that a 20-year payoff horizon, with hundreds of millions of dollars in R&D spend before you see a return, simply doesn’t work for private capital. Every year shaved off that timeline isn’t just a scientific win — it’s a financial one. Compress discovery from 20 years to 2, and you’ve fundamentally changed the investment math. The funding dries up not because the science fails, but because the clock runs out.
The bottleneck isn’t imagination. It’s velocity.
AI is attacking that bottleneck directly.
A few data points that frame the shift:
- AI-driven discovery methods are compressing that 10–20 year commercialization timeline to 1–2 years in some cases — through computational prediction, inverse design, and automated experimentation.
- AI models can simulate the properties of 1,000 two-dimensional materials in about 10 seconds, compared to over 10 million seconds with classical methods — roughly a 10,000x speedup.
- DeepMind’s GNoME has already predicted stable structures for over 2.2 million crystals (think battery materials, semiconductors, superconductors) — a tenfold increase in the pool of known stable compounds — and some of those candidates are already being synthesized in robotic labs.
- The generative AI in materials science market is growing from $1.1B in 2024 to a projected $11.7B by 2034 — a 26%+ CAGR.
What’s happening isn’t incremental. AI is acting as an embedded research partner across the entire pipeline — predicting which candidates are worth synthesizing before anyone touches a lab bench, designing experiments autonomously, and compressing the feedback loop between hypothesis and validated result from years to weeks.
MIT Technology Review recently visited Lila Sciences in Cambridge, where an AI agent — trained on vast scientific literature and datasets — is running physical lab experiments entirely autonomously, varying elemental combinations and learning from each result in real time. This is what the new materials R&D stack looks like.
So when I invest in software, here’s what I’m actually buying:
- New material procurement tools that make switching from plastic easier for large CPGs. The single biggest barrier to new material adoption at scale isn’t performance. It’s friction. Procurement teams at large consumer goods companies are navigating a thicket of supplier options, compliance requirements, and digital specs that don’t talk to each other. Software that eliminates that friction is software that accelerates material proliferation — full stop.
- Supply chain tools that help companies reduce material usage or substitute lighter-weight alternatives. A meaningful share of our supply chain software investments sit here. If a tool can show a packaging engineer that a 12% lighter substrate meets the same performance requirements, that tool is a materials innovation. It just doesn’t look like one.
- AI and simulation platforms that shrink the distance between a material scientist’s hypothesis and a validated result. This is the most direct expression of the thesis. When discovery timelines collapse, more materials make it to market. When more materials make it to market, the ecosystem we’re building into gets better inputs. Every lab-automation platform, every generative model for synthesis route prediction, every AI-powered characterization tool — these are all upstream bets on material proliferation.
Imagine a different kind of online marketplace
Here’s the model I keep coming back to — and I want to be specific about what it looks like, because I think the vision gets undersold when it’s described abstractly.
You open an app or visit an online marketplace. You browse not a catalog of physical goods sitting in warehouses, but an online catalog of validated, printable designs — some created by independent designers, some generated on the spot by AI from your own description. You find what you want, or describe what you want, and place an order. Within hours, a print node ten miles away receives the file and produces your item. A courier picks it up and delivers it to your door, packaged in minimal, right-sized material designed for a five-mile journey, not a transoceanic one.
The item didn’t exist before you ordered it. No factory retooled for it. No container ship carried it. No warehouse stored it for eight months while someone hoped it would sell.
That’s not science fiction. Every component of that system exists today. What doesn’t exist yet is the orchestration layer — the software, standards, and network design that ties print capacity, design validation, demand signals, and last-mile logistics into a seamless experience. And critically, this isn’t a solution for everything. Complex electronics, precision components, textiles — many categories will remain in traditional supply chains for a long time. But for a meaningful slice of the everyday goods that move across oceans today, the case for a local, on-demand alternative is becoming hard to argue with.
The sustainability math is compelling on every axis — less waste because nothing is made that isn’t sold, less capital tied up in inventory that may never move, less packaging because a locally printed item needs protection for a five-mile trip not a 8,000-mile one, and dramatically less carbon because a print node in New Jersey serving the Northeast eliminates most of the emissions associated with ocean freight, port handling, and long-haul trucking.
And the design surface expands dramatically. AI-generated design means the catalog isn’t a static list of SKUs — it’s a generative space. Want a phone case in a specific color, with a specific texture, sized for a model that launched last week? That’s a prompt, not a product line. The universe of purchasable objects grows by orders of magnitude, and none of it requires a factory in Guangzhou to retool.
The reframe I keep coming back to
In the data center piece I wrote last year, I made the argument that packaging failure is a mission-critical reliability problem disguised as a wooden crate. There’s an analogous reframe here:
A software investment in materials R&D is a materials bet — with venture-scale return potential to match.
The companies that will define what packaging looks like in 2035 — what substrates are available, what performance profiles are achievable, what the cost curves look like — are being built right now. Some of them are polymer chemists. Some of them are writing Python.
I’m interested in both.
If you’re building software, AI, or automation that touches the materials discovery or supply chain space, we’d love to hear what you’re seeing on the ground.
