Today, we are announcing our $310 million Series B financing, led by Khosla Ventures, with continued support from Andreessen Horowitz (a16z) and Breakthrough Energy Ventures, and new participation from Greenoaks, Halo Fund, Pax Ventures, StepStone Group, BHP Ventures, Washington Harbour Partners, Greycroft, General Innovation Capital Partners, Mitsubishi Corporation, In-Q-Tel (IQT), and Earthshot Ventures, with a number of additional strategic capital partners joining the round. This round brings the company’s total parent and project capital raised to date to over $400 million.
Demand for critical minerals continues to accelerate across every major industry – AI infrastructure, electrification, grid expansion, and aerospace. Meeting that demand requires new mining and refining capacity, complemented by increased metal recycling capacity. At the same time, lead time and cost to expand minerals mining and refining infrastructure continue to increase: ore grades are declining, ore bodies are deeper, mineralogies are more complex, and the industry has been slow to adopt new technologies that would offset these challenges (see our blog post announcing Copper One for more background).
Building a vertically-integrated company focused on end-to-end autonomy, speed to market, and lowering costs is the only answer – that is what we’re doing at Mariana Minerals. Our software stack, MarianaOS, has three key core operating systems that allow us to build better, mine better, and refine better.
CapitalProjectOS – Build Better
CapitalProjectOS is an integrated project lifecycle management tool that breaks down data silos and deploys a suite of AI agents to eliminate bulky, manual workflows. Globally, more that 90% of mega projects experience budget and schedule overruns. The problem is systemic, spanning execution, technology, labor, and quality. CapitalProjectOS consolidates engineering deliverables, equipment data, procurement, scheduling, and cost control into a unified, data driven environment, cutting deployment time and cost in half. It ensures the right data is delivered to the right person at the right time – automating data pipelines and aggregating project context into a single data frame, enabling agents to do useful work across engineering, procurement, and construction management. CapitalProjectOS makes an already strong team even stronger.
CapitalProjectOS – Build Better
CapitalProjectOS is an integrated project lifecycle management tool that breaks down data silos and deploys a suite of AI agents to eliminate bulky, manual workflows. Globally, more that 90% of mega projects experience budget and schedule overruns. The problem is systemic, spanning execution, technology, labor, and quality. CapitalProjectOS consolidates engineering deliverables, equipment data, procurement, scheduling, and cost control into a unified, data driven environment, cutting deployment time and cost in half. It ensures the right data is delivered to the right person at the right time – automating data pipelines and aggregating project context into a single data frame, enabling agents to do useful work across engineering, procurement, and construction management. CapitalProjectOS makes an already strong team even stronger.
MineOS – Mine Better
A modern mine runs on thousands of interdependent decisions a day across geology, drill-and-blast, fleet, and processing – and today those decisions are siloed, reconciled by humans in endless coordination meetings. The result is shift-to-shift variability and a widening bottleneck: there aren't enough experienced operators to run and optimize these systems as the workforce retires. MineOS pairs physics-informed world models with reinforcement learning agents that run mining fleets in a closed loop. It connects block models, mine plans, equipment and telemetry into a single orchestration brain, rather than running separate systems that require manual intervention to connect the dots across different parts of the operation. With MineOS, decisions are driven by data, not coordination meetings.
PlantOS – Refine Better
Today's mineral processing plants and refineries often iterate their process conditions a few times per week based on time-intensive manual sampling and human review. PlantOS replaces that loop with inline sensing and reinforcement learning (RL) control. Lab-scale experiments and kinetic parameter fitting are used to build a "world model", a digital simulation of a refinery. Reinforcement learning policies are trained against this simulation, learning through trial and error by taking actions and receiving feedback on how well they performed. The world model sits inside an environment wrapper that handles the back-and-forth of states, rewards, and action signals during training. Once trained, the system is connected to the real plant's control system, exchanging control signals and sensor feedback to help optimize the refinery. We are now using RL on one shared digital architecture across seven different unit operations at Copper One and Lithium One: brine pretreatment, direct lithium extraction, osmotically assisted reverse osmosis, heap (bio)leaching, solvent extraction, electrowinning, and oxidative leaching.
MineOS – Mine Better
A modern mine runs on thousands of interdependent decisions a day across geology, drill-and-blast, fleet, and processing – and today those decisions are siloed, reconciled by humans in endless coordination meetings. The result is shift-to-shift variability and a widening bottleneck: there aren't enough experienced operators to run and optimize these systems as the workforce retires. MineOS pairs physics-informed world models with reinforcement learning agents that run mining fleets in a closed loop. It connects block models, mine plans, equipment and telemetry into a single orchestration brain, rather than running separate systems that require manual intervention to connect the dots across different parts of the operation. With MineOS, decisions are driven by data, not coordination meetings.
PlantOS – Refine Better
Today's mineral processing plants and refineries often iterate their process conditions a few times per week based on time-intensive manual sampling and human review. PlantOS replaces that loop with inline sensing and reinforcement learning (RL) control. Lab-scale experiments and kinetic parameter fitting are used to build a "world model", a digital simulation of a refinery. Reinforcement learning policies are trained against this simulation, learning through trial and error by taking actions and receiving feedback on how well they performed. The world model sits inside an environment wrapper that handles the back-and-forth of states, rewards, and action signals during training. Once trained, the system is connected to the real plant's control system, exchanging control signals and sensor feedback to help optimize the refinery. We are now using RL on one shared digital architecture across seven different unit operations at Copper One and Lithium One: brine pretreatment, direct lithium extraction, osmotically assisted reverse osmosis, heap (bio)leaching, solvent extraction, electrowinning, and oxidative leaching.
CapitalProjectOS, MineOS, and PlantOS are the core pillars that makeup MarianaOS – one platform that connects the pit to the plant – and it’s already in use today across our Copper One and Lithium One sites. At Copper One, we compressed a multi-year acquisition, permitting, and restart timeline into months. At Lithium One, we’re on track to bring the first designed-for-autonomy greenfield refinery online less than 3 years after founding Mariana.
CapitalProjectOS, MineOS, and PlantOS are the core pillars that makeup MarianaOS – one platform that connects the pit to the plant – and it’s already in use today across our Copper One and Lithium One sites. At Copper One, we compressed a multi-year acquisition, permitting, and restart timeline into months. At Lithium One, we’re on track to bring the first designed-for-autonomy greenfield refinery online less than 3 years after founding Mariana.
We are only scratching the surface. We are already seeing meaningful uplift across our operations today – the bigger prize comes as we extend end-to-end autonomous orchestration across commercial mining and refining operations and expand our portfolio of projects.
We’re not waiting to start – we are building the mining company of the future, today. And that is what our Series B accelerates.
We are only scratching the surface. We are already seeing meaningful uplift across our operations today – the bigger prize comes as we extend end-to-end autonomous orchestration across commercial mining and refining operations and expand our portfolio of projects.
We’re not waiting to start – we are building the mining company of the future, today. And that is what our Series B accelerates.
