Manufacturing produces roughly 30% of global emissions, and gets a fraction of the VC attention paid to energy or mobility. That gap is structural.
The part that surprises people is where the emissions actually sit. It’s not just the energy source. A large share lives in process inefficiency: scrap, rework, unplanned downtime, over-processing.
Cement alone accounts for roughly 8% of global CO2 emissions, more than three times aviation’s footprint (WEF, 2024). We backed Alcemy early because those emissions are baked into the chemistry of the production process, not just the energy input. Their AI optimizes clinker ratios at the recipe level (clinker being the carbon-intensive intermediate that makes up most of a ton of cement). That’s exactly the kind of process-layer intervention that gets missed when people hear “manufacturing decarbonization” and picture renewables.
The sector stays overlooked because it’s structurally hard to invest in. Sales cycles are long. Customers are conservative. Most industrial AI pilots never reach production. McKinsey calls it “pilot purgatory” (and most VCs can smell that risk from the first customer slide). This isn’t a market that rewards impatience, and most early-stage investors don’t have the patience or the domain access to play here.
Two forces are converging now. On the technology side, AI can finally reach into previously siloed operational data and surface the anomalies, inefficiencies, and failure signals that used to be buried across disconnected systems. Work that once took months of custom engineering now ships in weeks, with models that generalize across plants rather than being rebuilt from scratch each time.
On the demand side, European manufacturers are under pressure from multiple directions at once. Energy costs are volatile. Supply chains remain fragile. The EU Carbon Border Adjustment Mechanism entered its definitive phase in January 2026. Certificates are priced at €75 per metric ton of CO2 for steel, cement, aluminum, and fertilizer producers. That cost scales roughly 40x by 2034. Carbon inefficiency now shows up on the P&L. Separately, experienced operators are retiring faster than knowledge can be transferred. 2.1 million US manufacturing jobs are projected to go unfilled by 2030 (Deloitte, 2022). A workforce crisis and a regulatory squeeze, hitting the same shop floors at the same time.
Our investment in Aris Machina fits exactly this window. Co-founders Siddharth Khullar and Peter Carlsson come at the same problem from opposite ends. Siddharth is the ex-Apple AI scientist who became Northvolt’s VP of Software Engineering. Peter built Tesla’s global supply chain during the Model S launch, then raised $13B to launch Northvolt as Europe’s flagship gigafactory. Between them, they saw firsthand where data and intelligence failures compound at scale. The FactoryOS they’re now building stitches fragmented IT/OT stacks into a single data and AI layer across complex manufacturing environments.
At AENU we see both system-level platforms and AI co-pilots as real opportunities. The two play very differently.
System-level plays have higher ceilings. They pull heterogeneous data into one place and can become the rails that co-pilots and other industrial applications run on. By unifying energy, quality, and equipment telemetry, they enable plant-wide process optimisation and predictive maintenance, catching consumption anomalies, flagging quality failures before they occur, and scheduling interventions before breakdowns cascade. They also demand deeper integration and founders who can navigate enterprise procurement. Aris Machina sits here, working across automotive OEMs and battery manufacturers on energy and material waste. Alcemy is a vertical-specific version of the same thesis, applied to cement and concrete.
Co-pilots are more accessible. Task-specific tools that sit on top of existing infrastructure, help operators troubleshoot faster, and pay back inside a quarter. They capture fragile expert knowledge into software-supported workflows, guiding technicians through predictive maintenance checks and process adjustments that would otherwise rely on a shrinking pool of senior experts.
We lean in when co-pilots show a credible path to becoming an infrastructure layer, or when platforms have strong potential to automate over time. We pass on solutions that are useful but capped, with no expansion path beyond the initial use case.
Distribution is the real bottleneck in this pipeline. We keep seeing strong product with weak go-to-market.
The ideal pitch on our desk: founders with real unfair access (industry relationships, shop floor credibility, or both), at least one paying customer with measurable results, and value that’s easy to audit. Strong go-to-market means knowing exactly who you’re selling to, why they buy now, and what the expansion path looks like beyond the first pilot.
Alcemy reinforced this for us. The AI mattered, but what made the investment compelling was a team that could credibly sell into conservative cement plant operators and deliver measurable CO2 reductions they could point to.
If you’re building in smart manufacturing, or if we’re missing a name we should know, get in touch at aenu.com/contact.
This research and article was supported by Visiting Analyst Hélène Dupré. research and article was supported by Visiting Analyst Hélène Dupré.
