Tesla says Optimus will cost less than a car. Figure just raised $675 million at a $2.6 billion valuation. Boston Dynamics is licensing Atlas for commercial use. NVIDIA is building the simulation layer that will train all of them.
This is not a CES demo anymore. This is capital allocation.
And the financial services industry is not paying attention.
The factories are already talking about it
I spend my weeks talking to banks, wealth managers, and asset managers about AI. The conversation is always about software — models, compliance, back-office automation. Nobody in these rooms is talking about hardware.
Meanwhile, outside those rooms, something very concrete is happening. Amazon has over 750,000 robots in its warehouses. BMW is testing Figure 02 humanoids on its assembly lines in Spartanburg, South Carolina. Hyundai — which owns Boston Dynamics — is integrating robots into its own manufacturing plants. Toyota is doing the same in Japan.
These aren't research projects. These are line items on a balance sheet. And every single one of them has financial implications that nobody in traditional finance is modeling yet.
The economics of a robot that works 22 hours a day
A warehouse worker in the US costs roughly $35,000–$45,000 a year. A humanoid robot, based on Tesla's projections, might cost $20,000–$30,000 to buy. But it works 22 hours a day. It doesn't take sick leave. It doesn't quit after six months because the warehouse down the road pays $2 more per hour.
Do the math. A single robot replaces the equivalent of 2.5 to 3 full-time workers on a shift-adjusted basis. Even with maintenance, power, and depreciation, the unit economics are brutal for human labor in repetitive physical tasks.
Now multiply that by the 4 million warehouse workers in the US alone. Or the 12 million manufacturing workers. Or the estimated 300 million jobs globally that McKinsey flagged as automatable.
This is not a labor story. This is a capital formation story. And capital formation is finance.
Who finances a fleet of warehouse bots?
Here is where it gets interesting for anyone in financial services.
A logistics company that wants to replace 500 warehouse workers with 200 robots needs capital. Not the kind of capital you raise on a Series B. Industrial capital. Equipment financing. Leasing structures. The same financial products that exist for trucks, forklifts, and CNC machines.
But robots are different from forklifts in a few critical ways.
First, they depreciate differently. A forklift is a forklift for 15 years. A robot gets software updates. Its capabilities change. Its residual value is tied to the quality of its AI, not just its mechanical parts. How do you underwrite a loan for something whose value might go up with a firmware update?
Second, they generate data. A robot on a factory floor produces a continuous stream of operational data — uptime, error rates, throughput, energy consumption. That data is an asset. It can be securitized, analyzed, and used for pricing. But no bank has the infrastructure to ingest it.
Third, they create new liability questions. If a robot drops a pallet on someone's foot, who pays? The operator? The manufacturer? The company that wrote the navigation software? The insurer who underwrote the policy without understanding any of this?
Every one of these questions is a financial product waiting to be built.
Robot-as-a-Service is the real play
If you want to understand where this is heading, stop thinking about robot purchases. Think about subscriptions.
The most likely deployment model for humanoid robots isn't buying them outright. It's Robot-as-a-Service. RaaS. Same model that made Salesforce, AWS, and every other infrastructure company rich.
Figure is already moving in this direction. Their pitch to BMW isn't "buy 500 robots." It's "pay per task completed." Agility Robotics, which makes the Digit humanoid, is doing the same with Amazon.
This matters for finance because RaaS turns a capital expenditure into an operating expense. It changes how companies budget. It changes how they report. It changes what their balance sheets look like. And it creates a massive new category of recurring revenue businesses that need financing, insurance, and risk management.
The company operating a fleet of 10,000 warehouse robots on a RaaS model looks a lot like a fleet management company. It needs working capital. It needs liability coverage. It needs contracts with terms that don't exist yet. It needs a bank that understands all of this.
Right now, that bank doesn't exist.
From the factory floor to the living room
Here is where it gets really big.
The factory use case is obvious. Warehouses, manufacturing, logistics — these are controlled environments with known variables. But the real scale comes when robots enter the home.
Samsung, LG, and Dyson are all building household robots. Not Roombas. Full-size machines that can fold laundry, load dishwashers, and carry groceries. Samsung showed off Ballie at CES 2025 — a rolling home assistant that projects screens, controls smart home devices, and moves around autonomously.
When a household robot costs $15,000–$25,000, it's priced like a car. And it will be financed like a car.
Think about what that means. Auto loans are a $1.6 trillion market in the US alone. Robot financing could be a comparable category within a decade. Monthly payments. Interest rates. Credit scoring. Lease-to-own. Trade-ins.
And then insurance. A car can hit a pedestrian. A household robot can break your television, knock over your grandmother, or malfunction while holding a kitchen knife. The liability models for home robots don't exist. Someone has to build them. That someone will be in financial services.
Wealth management needs new models
This is the part I think about most, because this is where I work.
If robots take over a significant portion of physical labor — and the trajectory says they will — then income distribution changes in ways that wealth management has never had to deal with. Combined with the longevity revolution, the entire financial planning model needs to be rebuilt from scratch.
The person who owns ten warehouse robots and leases them to Amazon makes money while sleeping. The person who used to operate the warehouse forklift does not. This isn't a new concept — capital vs. labor is the oldest tension in economics. But the speed and scale are new.
A single company could deploy 100,000 humanoid robots in 5 years. Tesla says it will build them at a rate of millions per year by the end of the decade. Even if they hit 10% of that target, the economic displacement is massive.
Wealth managers will need to advise clients on robot fleet investments the way they currently advise on real estate portfolios. Robot equity funds. Robot leasing REITs. Robot maintenance bonds. None of these exist today. All of them will exist within five years.
The regulatory vacuum
Europe is way ahead on AI regulation with the EU AI Act. But the Act is focused on software AI — language models, decision systems, biometric surveillance. It has almost nothing to say about physical robots operating in workplaces and homes.
Who certifies a household robot for safety? Under what framework? Is it CE marking, like a kitchen appliance? Or is it more like automotive certification? What happens when a robot with German hardware, American AI, and Chinese sensors causes damage in Italy?
This regulatory vacuum is a massive opportunity for whoever fills it first. In financial services specifically, the institution that builds the risk models, insurance products, and compliance frameworks for physical robots will own a category.
Why finance should care right now
Not in 2030. Not when the first robot walks into a bank lobby. Now.
Because the financial infrastructure for robotics needs to be built before the robots ship at scale. Underwriting models take years to develop. Insurance actuarial tables need data that doesn't exist yet. Leasing frameworks need legal precedents that haven't been set.
The auto industry had decades to build its financial ecosystem. Robots will have maybe five years.
I work at the intersection of AI and financial services. Every week, I see how slow finance moves when it comes to software AI — and software AI has been around for years. Physical AI is newer, more complex, and more capital-intensive. If finance struggled to adapt to ChatGPT, it is completely unprepared for a humanoid robot that shows up on a loading dock and starts working.
The robots don't need bank accounts. But the companies that build them, deploy them, insure them, finance them, and profit from them absolutely do. The financial services industry that figures this out first won't just serve the robotics economy. It will own a piece of it.
That's not a prediction. That's an inevitability on a timeline. The only question is who moves first.
