This post was originally published on this site.
Travis Kalanick has never run a mine. In 2026, backed by a $1.7 billion raise led by Andreessen Horowitz, his company Atoms acquired Pronto, an autonomous haulage firm, and built a mining division around it. He is now one of the more consequential new investors in how mines actually get run.
In 2022 I wrote that the energy transition is really a metals transition, and that mining would need its own version of the shale revolution to keep pace with rising demand. The argument rested on the sector’s structural constraints, declining ore grades, remote labor challenges, and environmental pressure, all of which would make technology disproportionately valuable across the physical value chain.
Three years later, a wave of acquisitions across the industry is the clearest evidence yet that the rest of the market has arrived at the same conclusion. As Caterpillar’s chief executive Joe Creed put it at CES earlier this year, the integration of industrial AI is now about powering progress in the physical world. That sentiment, coming from the largest equipment maker in the industry, is itself a signal of how far the argument has traveled since 2022.
Three distinct groups of buyers illustrate how this is unfolding. The first is traditional equipment manufacturers, assembling a software layer to protect their installed base and secure higher-margin, recurring revenue across their hardware footprints. Caterpillar completed a roughly $733 million acquisition of RPMGlobal in February 2026, then followed it with the purchase of Skycatch.
Sandvik announced the acquisition of ThoroughTec Simulation, a company generating around $17.5 million in high-margin revenue. The second group is consumables and civil engineering incumbents converging on the same territory. Orica’s recent acquisitions, alongside Bentley Systems’ purchase of Seequent, suggest that chemical and geospatial leaders intend to own the integrated mine-to-mill digital workflow as well.
The third group, Atoms among them, is venture capital moving into physical automation directly, treating heavy industrial automation as a route to building real-world AI models rather than another software category to license into.
The money is clearly arriving. The more interesting question is whether it is working.
Despite hundreds of millions spent on acquisitions, the industry is running into an operational bottleneck. Miners have accumulated dozens of isolated software applications for geology, fleet management, and processing that operate in silos, and many cannot extract the return on investment they expected because they lack a single source of truth. Buying point-solution software has turned out to be the easy part. Making legacy systems communicate in real time is where value creation stalls. There are rare, early indicators of what happens when a buyer gets this right—Orica, for instance, recently reported a significant earnings uplift in its Digital Solutions business.
By successfully integrating major acquisitions like Terra Insights, they drove enough recurring digital revenue to push segment earnings growth to 29 percent. But across the broader sector, Orica’s momentum remains the exception, and a meaningful share of the capital deployed by others over the past year is still waiting to pay off.
That bottleneck is pushing acquirers toward two kinds of companies next. The first generates high-fidelity, real-time data at the point of operation, since AI models are of limited use without real-time operational data. Exum Instruments and Minpraxis are both working on this problem, bypassing weeks-long laboratory delays with trace-level physical and chemical data generated on site.
The second is the unified spatial model, bringing geological, fleet, and environmental data into a single platform so operators can run predictive simulations instead of working off stale, disconnected records. Strayos and AiMinr are both building toward this.
This friction is frequently cited in industry conversations. Senior people at tier-one mining majors, global equipment manufacturers, and major chemical and service providers all describe the same bottleneck once the cameras are off. As one senior technical director at a major equipment manufacturer put it, everyone in the industry knows digital tools and AI are necessary to drive real gains in performance. Knowing that hasn’t solved the problem.
That same firm’s own experience makes the point. It has spent heavily on point solutions, machine-vision platforms, AI process optimizers, and specialized mine-planning software, each one good at its specific job. But none of them were built to talk to each other, and stitching them into a single system has proven far harder than buying the pieces. Mining majors increasingly agree that point solutions are necessary but not sufficient. What’s missing is the integration layer to connect them from the start, not bolted on after the fact.
The first real test of whether any of this actually works is happening on a desert copper mine in Utah. Mariana Minerals, backed by Andreessen Horowitz, Breakthrough Energy Ventures, and Khosla Ventures, restarted the previously idled Lisbon Valley copper operation in April 2026 with autonomy built into production from the first day, running Pronto’s autonomous haul trucks and Sandvik’s autonomous drilling under its own software layer.
The bet is that a mine designed around integration from the outset can outrun incumbents still stitching legacy systems together after the fact. It is far too early to call this proof of anything. The environment is unforgiving, the equipment takes real wear in the field, and scaling one site to the volumes the industry needs is a different challenge than getting a single mine running. It remains, even so, the sharpest live test available for the idea that owning the integration layer, not just the tools that feed it, is what actually closes the productivity gap.
For institutional investors, this multi-directional consolidation creates a more legible exit market than the sector has had in some time. The collision between heavy industry incumbents, adjacent consumables and engineering players, and well-capitalized new entrants ensures a persistent bid for whoever controls core data assets and orchestration software. A year ago, an investor backing an unproven mining technology company had a narrow set of buyers to point to. Today that list is considerably longer.
The sector has spent a decade talking about declining ore grades and slow permitting. The more urgent constraint may be simpler than either. It comes down to whether anyone can actually make all this new technology work together.
Tem Tumurbat is co-founder of Nomadic Venture Partners, an investor in critical minerals and mining technology.
