Square Robot’s Series B and Marathon collaboration signals AI tank inspection robots are scaling in energy. Here’s what it means for safety, uptime, and ROI.
AI Tank Inspection Robots: Square Robot & Marathon Deal
A refinery storage tank can hold millions of dollars’ worth of product—and a single inspection decision can create weeks of downtime, a safety incident, or a compliance headache. That’s why the recent news that Square Robot raised a new Series B round and entered a collaboration with Marathon Petroleum matters more than another “funding announcement.” It’s a signal that AI-powered inspection robotics is moving from interesting pilots to infrastructure-scale deployment.
Marathon operates the largest refining system in the United States. When a company with that footprint not only invests, but also commits to shaping the next-generation robot platform, it’s a tell: energy operators are done waiting for inspection methods that require draining tanks, issuing confined-space permits, and putting people where they shouldn’t be.
This post sits inside our “AI in Robotics & Automation” series for a reason. The story isn’t just about a robot that swims. It’s about how AI-driven automation is changing the economics of asset integrity—turning inspections into something closer to an always-on data stream instead of a disruptive event.
Why tank inspection robotics is getting funded now
Answer first: Industrial inspection robotics is getting funded because it hits three hard business outcomes at once: fewer safety exposures, less downtime, and better data quality—and those three translate directly into ROI.
Traditional internal tank inspections are expensive for reasons that have nothing to do with “inspection skill.” You’re paying for operational disruption: drain, clean, ventilate, scaffold, permit, and manage entry. Even when everything goes well, the process is slow, and it forces a binary choice—inspect thoroughly or keep the tank available.
Robotic tank inspection changes the premise. Square Robot’s systems are designed to work submerged, enabling inspections while tanks remain online. That single shift—inspect without shutting down—is the difference between a technology that’s “nice” and one that gets a line item in next year’s budget.
The real driver: operational continuity
In late 2025, many energy and chemical operators are under simultaneous pressure to:
- Improve safety performance (especially around confined-space work)
- Prove compliance with tighter, more auditable integrity programs
- Keep throughput steady even as maintenance labor is harder to schedule
Robotic inspection aligns with all three. And it scales well: once you have a validated process and a trained crew, adding more tanks is mostly a logistics problem, not a new engineering project.
What Marathon’s collaboration signals (and why it’s different)
Answer first: Marathon’s participation signals that inspection robotics is becoming a strategic platform decision, not a one-off service.
Plenty of industrial robots get “tested” on one site and stall there. What’s notable in this announcement is the two-way commitment:
- Marathon participates in the Series B round
- Marathon continues deploying Square Robot’s existing fleet
- Marathon helps shape the next-generation platform’s design and development
That last point is the hinge. When an operator co-develops, they’re effectively saying: “We want this capability as a long-term standard, and we care about how it integrates with our workflows.”
Co-development usually means integration becomes the product
In my experience, hardware is rarely the biggest blocker after the first few successful inspections. The blocker is integration:
- How does inspection data land in the asset integrity system?
- How do you compare tank-to-tank and year-to-year?
- How do you triage findings so engineers aren’t buried in footage?
- How do you prove traceability for audits?
A collaboration at Marathon’s scale tends to push vendors toward repeatable deployment packages: standardized reporting, better data schemas, clearer acceptance criteria, and cleaner maintenance processes for the robotic fleet itself.
Where AI actually fits in submersible robotic tank inspection
Answer first: In robotic tank inspection, AI is most valuable for detecting, classifying, and prioritizing anomalies—and for turning raw sensor feeds into decisions that maintenance teams trust.
It’s easy to say “AI-powered robot,” but the practical version looks like this:
1) Perception and navigation in ugly environments
Tank interiors are hostile to neat robotics demos. You’re dealing with low visibility, reflections, sludge, temperature variation, and complex geometry. A submersible robot needs robust perception to maintain coverage patterns and avoid missing critical areas.
AI helps by improving:
- Coverage assurance: confirming the robot scanned the required surfaces
- Localization support: estimating position when GPS is irrelevant and conditions vary
- Adaptive pathing: adjusting scan behavior based on visibility or obstructions
2) Turning inspection into “findings,” not footage
Engineers don’t want more video. They want actionable findings: corrosion patterns, pitting, coating failures, deformation indicators, and “areas to re-check.” AI-assisted analysis can reduce manual review time and make inspection intervals smarter.
A useful mental model is: AI doesn’t replace the inspector; it replaces the endless scrubbing of data.
3) Trend intelligence across a fleet of tanks
The under-discussed win is longitudinal data. If your inspections are consistent and frequent enough, AI can help identify:
- Which tank designs degrade faster in specific service conditions
- Which maintenance interventions actually slow corrosion
- Which locations should get higher inspection frequency
That’s when inspection shifts from compliance to predictive maintenance.
A strong inspection program doesn’t just find defects—it reduces the number of defects that reach “urgent.”
The ROI math operators should use (and the traps to avoid)
Answer first: The cleanest ROI model for robotic tank inspection is: downtime avoided + labor risk reduced + rework reduced + better planning.
Here’s a practical checklist of what to quantify before buying or expanding robotic inspection:
What to measure (and what finance will accept)
- Days of tank downtime avoided (and the operational value of that availability)
- Confined-space entries eliminated (trackable safety exposure reduction)
- Cleaning and disposal volume reduced (environmental handling costs)
- Inspection cycle time (from planning to report acceptance)
- Finding-to-work-order time (how quickly defects translate into action)
Traps that make good tech look “too expensive”
- Comparing service price to labor alone. The big cost is downtime, preparation, and disruption.
- Ignoring data workflow costs. If outputs don’t fit the integrity program, you’ll pay for manual translation.
- Treating every tank like a snowflake. Standardize acceptance criteria and reporting templates early.
- Skipping operational readiness. Battery handling, decon procedures, staging areas, and permits still matter.
If you want robotic inspection to expand beyond one enthusiastic site champion, build the business case around repeatability and auditability, not just “cool robotics.”
What “next-generation” should mean in 2026 for inspection robots
Answer first: Next-generation inspection robots should reduce human effort in three places: setup, data interpretation, and system integration.
Marathon helping to shape Square Robot’s next platform suggests the roadmap will be influenced by real operational pain points. If you’re evaluating solutions in 2026, ask vendors (and yourself) whether improvements are focused on outcomes rather than specs.
Setup: faster mobilization, fewer constraints
Look for improvements like:
- Faster deployment and retrieval procedures
- Better robustness across product types and temperature ranges
- Simplified decontamination workflows
Interpretation: AI that supports decisions, not dashboards
Ask how the system handles:
- Automated anomaly flagging with clear confidence and rationale
- Human review workflows (approval, escalation, annotation)
- Repeatable scoring so you can compare inspections over time
Integration: inspection data that “lands” cleanly
The highest-value feature is boring: structured outputs that connect to your CMMS/EAM and integrity records. If the robot produces beautiful media but no clean work order triggers, you’ll feel the friction in month two.
Practical buying guidance: how to evaluate robotic tank inspection
Answer first: Choose a robotic tank inspection approach based on risk profile, inspection frequency goals, and how much you care about trending over time.
Here’s a straightforward evaluation framework you can run in a procurement cycle.
Questions that separate pilots from scalable programs
- Can the tank remain in service during inspection? If not, your downtime savings shrink.
- What’s the minimum data package you’ll accept? Define this before the demo.
- How is coverage verified? “We think we scanned it” isn’t acceptable.
- How are findings prioritized? Your engineers need triage, not raw output.
- Who owns the data and how is it stored? Plan retention, access, and audit needs.
- What does the crew model look like? Vendor-operated, hybrid, or fully in-house?
A sensible first deployment pattern
If you’re starting in 2026, a strong rollout pattern is:
- Pick one high-impact asset class (e.g., large storage tanks with painful downtime)
- Run 3–5 inspections with the same reporting format
- Compare results to your baseline method on: cycle time, findings quality, and readiness impact
- Lock the process into a standard operating procedure
- Scale site-to-site with a consistent KPI set
This is how you avoid “demo theater” and build a program that survives leadership changes.
Where this fits in the bigger AI in robotics & automation story
Answer first: Inspection robotics is one of the clearest examples of AI delivering real industrial value because it improves safety and uptime without requiring a full plant redesign.
Humanoid robots get headlines, but specialized industrial robots are quietly becoming the workhorses of automation—especially when they operate in hazardous, hard-to-access environments. Energy infrastructure is full of those environments: tanks, pipes, terminals, refineries, and remote sites.
Square Robot and Marathon’s collaboration reinforces a trend I expect to accelerate through 2026: operators will fund robotics that reduces exposure, keeps assets running, and produces better integrity data. Anything else will struggle to get out of pilot mode.
If you’re responsible for reliability, inspection, or automation strategy, now is the time to decide: do you want inspections to be periodic disruptions, or a controlled, data-driven operation that’s easier to audit and easier to schedule?
The next question is a good one to end on: If you could inspect more often without downtime, how would that change your maintenance strategy—and what failures would you prevent before they become “urgent”?