valve is very differently structured and has had a very tough time pivoting to hardware. We can see this obviously with steam machines where they just didn't want to get involved. Then the steam controller, steam link, and now steam deck...
they're getting there but they're not exactly there yet. I think oculus taking their research tech and turning it into a $2b acquisition and a threat to steam because of its closed garden was a big wake up call.
It takes consistent organizational effort spanning years to build hardware, maintain vendor relationships, merchandize the product (advertise, sell, ship, liquidate), inventory management, and continuously develop new hardware to bring to market.
Valve has traditionally focused on their Steam platform, which has a much tighter feedback loop and none of the aforementioned effort that comes with physical products.
Not to mention Ben Krasnow's comment about the hardware culture at Valve probably meant a lot of talented hardware people wouldn't be interested working somewhere where their accomplishments are seen as worthless. Not having good talent and good culture are not good for developing good hardware.
Hardware is, from a business perspective, completely different from software.
Iteration cycles are way longer and more expensive. Find a bug in a piece of software / firmware? Push out an update and the bug is gone. For hardware bugs you have to pray you can fix it in firmware somehow, or you have to call back hardware from the customer and send them replacements in the worst case, or deal with the fallout of class action lawsuits or regulatory punishments (e.g. if you mess up something that causes inappropriate RF emissions).
Additionally, developing for hardware is way harder than developing software. In software, you have clean, somewhat-well designed APIs to get you up and running, and there's bazillions of StackOverflow posts and open source code you can have a look at if you have problems. In hardware, you have to deal with half-assed documentation, reference designs that do not work / are buggy / cause unwanted RF emissions / produce signals that are outside of standards, tight NDAs / stuff that isn't even available under NDA but is vital in debugging issues, BSPs (board support packages) with outright fossilized and Frankenstein'd bootloaders and Linux kernels, not to mention binary blob stuff such as early stage bootloaders, WiFi/BT/GPU firmware and other completely intransparent and barely-working crap.
And once you do have a working prototype in your hand, you have to deal with more bullshit: certifications (UL/TÜV electrical safety, FCC RF emissions, CE, and whatever specific local markets require) primarily (and the findings of the certifications may well send you back to the drawing board, which means more expenses), ridiculous minimum-order quantities, supply chain establishment and upkeep (=preventing counterfeit components in your chain) in general, manufacturing QA, logistics of getting the hardware to consumers, returns/warranty claim/repair/spare parts logistics, keeping track of components getting EOL'd or outright being unavailable due to some component availability crunch, keeping track of recalls of components before your product ends up setting someone's house on fire (=the usual trouble with Lithium batteries and shoddy power supplies), dealing with insurance to cover your butt in case your product does end up setting someone's house on fire or electrocuting someone...
Hardware is ugly and it's rare to have hardware, firmware and software be matched in quality (which is also why so many hardware Kickstarter/Gofundme projects fail or under/late deliver). The only vendor where that is closest to reality is Apple, and they command a hefty premium for that.
they're getting there but they're not exactly there yet. I think oculus taking their research tech and turning it into a $2b acquisition and a threat to steam because of its closed garden was a big wake up call.
We'll see how they do in the future I guess.