Roku Loses Money on Every Device It Sells. That's the Plan.

Buy a Roku stick for the price of a few sandwiches, plug it into the back of your TV, and you have just done Roku a favor it would happily have paid you for. The company sells that hardware at a loss on purpose - in fiscal 2024 its Devices business ran a gross margin of about negative 14%, with the cost of making and shipping the boxes ($670 million) running higher than what the boxes brought in ($590 million).¹ Every device sold made the loss bigger. And the bigger the loss, the better Roku liked it - because the box was never the product. You are.

The official story is that Roku is a streaming-device company. It is the most persistent misread of the business, because the device segment doesn't make a cent of profit - it deliberately bleeds, so that something far larger can grow. Roku is not a gadget maker that dabbles in ads. It is a toll booth that gives away the road to manufacture the traffic it then charges for.

The box is bait. The household is the catch.

Strip the income statement down and the architecture is unmistakable. In 2024, Roku's Platform segment brought in $3.52 billion - about 86% of total revenue - at a gross margin north of 53%.¹ The Devices segment, by contrast, was $590 million of revenue and a negative margin: a money-losing line item by design.¹ One leg of the business is a software machine that prints cash; the other is a hardware machine that burns it. They are not two businesses competing for the company's attention. They are one machine, and the loss-making leg exists to feed the profitable one. The device's only job is to plant Roku's operating system in front of a household and keep it there.

Once the household is captured, Roku monetizes it two ways, and the distinction matters more than most coverage admits. Platform revenue rests on two legs: advertising - direct and programmatic video, sponsorships, and the ads woven into Roku's home screen - and streaming-services distribution, where Roku takes a cut of subscriptions sold through its interface, sells premium subscriptions, and even charges content owners for the branded buttons on the remote.³ Roku does not publicly split the two, so anyone who tells you Roku is 'mostly an ad company' is guessing at the size of one leg of a stool they can only see the seat of.

This is why Streaming Households is the number that actually governs the business. Roku ended 2024 with 89.8 million of them, up 12% year over year, and crossed 90 million in the first week of January 2025.⁴ Each new household is a fresh seat in front of which Roku can sell attention for the lifetime of the device. The hardware loss is, in effect, a customer-acquisition cost paid upfront and recouped over years of platform revenue - a subscription you didn't know you signed up for, financed by a stick you thought you bought.

A failed Netflix gadget that became the better business

The model's origin is its own quiet tell. Roku, LLC was incorporated by Anthony Wood back in 2002,⁶ but the thing we now call Roku began inside Netflix as a secret hardware effort codenamed Project Griffin. Weeks before its planned late-2007 launch, Reed Hastings killed it - he worried that a branded Netflix box would look like competition to the third-party hardware partners Netflix needed.⁵ Wood, who had been leading the project, then relaunched it independently, and the first Roku player, the DVP N1000, was unveiled in May 2008.⁶

Netflix walked away because hardware was a strategic liability for a subscription company. Roku embraced the same hardware - and then did something Netflix never wanted to: it turned the box into a neutral storefront for everyone's streaming service and took a toll on the traffic. The thing Netflix discarded as a distraction, Roku rebuilt as the foundation. Selling the box at a loss only makes sense if you're not in the box business at all.

Isn't a business that loses money on hardware just a slow bleed?

The honest objection is that this model spent years not working. Selling devices below cost to chase a future toll is a leap of faith, and for a long stretch Roku posted operating losses - $218 million in 2024 alone.¹ A skeptic could fairly say: a company that loses money on its product and loses money overall is not running a clever flywheel, it's running on investor patience. That is a real risk, and the model only earns the benefit of the doubt at scale - when the household base is large enough that platform revenue dwarfs the hardware subsidy. Roku has been approaching that line. Q4 2024 was the first quarter Platform revenue exceeded $1 billion, up 25% year over year,² and the company reported its first full year of positive net income in 2025 as Streaming Households marched toward 100 million.⁷ The bleed, in other words, finally had a floor under it - but it took two decades and ninety million households to reach.

There is a second crack worth naming. ARPU is going sideways, not up, and Roku admits why: a growing slice of its households are international, where monetization is still in its early stages.⁸ So the growth engine is the one that pays less per seat. The toll booth keeps adding lanes, but the newer lanes collect smaller tolls - and whether they ever fill up at U.S. rates is the open question the whole valuation rests on.

Roku makes its money the way a casino makes money on the buffet: the cheap thing at the door is not where the profit lives, and it was never meant to be. The stick is the buffet. The platform is the floor. Every device sold at a loss is one more household seated in front of a screen Roku controls, monetized a little at a time, for years. The genius was never a better gadget. It was the nerve to lose money on the gadget on purpose - and the patience to wait until ninety million living rooms made the loss the cheapest customer Roku ever bought.