Welcome to Words with Wynn! If this is your first time perusing my content and you’d like more of my weekly musings, subscribe below:
The Acquired podcast changed my perspective on Meta’s investment in Reality Labs (AR/VR):
“I'm pretty sure glasses and holographic presence and AR is going to be a completely ubiquitous product, right, it's just like everyone who had a phone before replaced it with a smartphone and then a lot of more people got smartphones. If all we get is all the people in the world who already have glasses upgrading to glasses that have AI in them, then this is already going to be one of the most successful products in the history of the world, and I think it's going to go a lot further than that. So I know there's that.
There is the thing about controlling our own destiny. It's strategically valuable. You know we did this calculation or estimate at some point where it's like how much money do we lose from our core family of apps to the various taxes that the platforms [Apple’s App Store, Google, etc.] have to like, when they tell us we can't run the ad business the way that we think we should be able to.
When they tell us we can't ship certain products, that way people use the things less or like them less, and it's hard to exactly estimate it. But I think we might be twice as profitable if we own the platform or something. I think from that perspective that's worth a lot, just from a pure dollars perspective, which is not primarily how I come at this stuff…” - Mark Zuckerberg on Acquired
AR/VR was all the hype around the time that Zuckerberg rebranded Facebook to Meta, and Wall Street torched him in the ensuing years for the company’s massive investments into augmented reality technology as the sector fell out of favor. But when you frame it within the context of owning the next platform, their hefty R&D bill starts to make a bit more sense. You can basically run a loose NPV for what that’s worth against the cash sink to develop the technology today. This theoretical number could be so huge that it’s rational to continue your investment as an option on a gargantuan future market. Whether or not AR/VR becomes the dominant platform is anyone’s guess, but if you’re the founder and controlling shareholder of Meta, and your timelines are measured in decades, there is a serious logic in betting on (and willing it into existence) augmented reality if the probability of its ubiquity is high enough.
This is the power of platforms.
Owning your own destiny has proven time and again to be fundamental in cementing network effects with hardware and user lock-in within your ecosystem. It’s tech strategy 101.
Historically (shooting from the hip here):
Microsoft: Had significant market power as the dominant operating system, and integrated its products to both build and capitalize upon network effects. When the internet began to emerge and Netscape was gaining traction as the go-to web browser, Microsoft simply bundled Internet Explorer with its operating system to quash the competition. This type of bundling strategy was so valuable/restrictive that it ultimately led to the company’s antitrust woes in the 90’s.
Apple: Called (and revolutionized) mobile perfectly, and has been able to leverage that ownership ever since. Seamless integration between your iPhone, Airpods, and Macbook creates a hardware ecosystem, and dominating the app store across all these platforms to extract onerous fees indefinitely (See: Epic Games vs. Apple) monetizes the software.
Google: A search engine behemoth saw the mobile revolution, and ensured that it had a horse in the race, reducing its reliance on and gatekeeping by Apple, by acquiring Android. Android has since become the largest smartphone platform today, natively preferencing Google’s product suite (Gmail, Maps, etc.) and helping to proselytize Chrome much like Microsoft’s play before it.
So, platforms are imperative. And beyond consumer hardware (AR/VR), who’s building the next one?
I think the framework above is an interesting lens to contemplate the humanoid robotics landscape. It adds some meat to the various investments and partnerships we’ve seen:
Figure: partnership with OpenAI to develop humanoid AI models + $675M round from Microsoft, OpenAI Startup Fund, NVIDIA, Intel, Bezos Expeditions.
Rainbow Robotics: Samsung’s $181M investment for 35% ownership.
Apptronik: partnership with Google DeepMind.
Agility Robotics: investment from Amazon’s Innovation Fund.
Tesla: speculation around collaboration with X/xAI (supercluster capabilities).
If you consider humanoids the next big wave, however far out, then it is imperative to all the major players to have a ‘piece on the board’ if even to just maintain option value.
This is what came to mind when Apptronik announced its partnership with Google’s DeepMind in late December. I’m sure that Google did an incredible amount of diligence on the various players in the space and likely had carte blanche in their choice of partner (being DeepMind and all), so I’m not trivializing Apptronik’s state of the art technology. But I use this to say that many of the forefront competitors had already been tapped by the other big dogs in AI (OpenAI, etc.), and Google didn’t have a horse in the race. The major tech companies all want a piece on the board, if only to keep a pulse on the industry and maintain an option.
Additionally, the import of these options compounds if you believe that value will ultimately accrue to the intelligence layer (higher margin software). It’s the same playbook as the infra providers investing in LLMs, a la Microsoft <> OpenAI. Sure, they’ve invested significant sums into OpenAI– revenue share aside, and much of the investment was actually compute services instead of cash– but that’s a small price to pay to get ingrained into a potentially massive future customer. All of the protocol level LLM companies are going to require massive amounts of compute, which the hyperscalers are all competing to provide.
I fully expect a parallel with humanoids. Yes, OpenAI is investing and learning and having a dog in the fight. But they’re also securing themselves in pole position to be the vendor of choice for a potentially massive customer (Figure). Embodied intelligence will need just that: intelligence. The value likely accrues to the intelligence layer, and all the providers of that service are happy to ingrain themselves with the corresponding hardware builders.
How this plays out in practice is anyone’s guess, to be fair. Do the humanoids themselves eventually become commoditized, with the ‘sauce’ being the interchangeable AI that powers them? I’m sure it’s likely to be a spectrum of strategies. All of the robotics manufacturers will have some level of AI expertise to be sure, but the massive data processing required to train and perfect models across fleets at scale may lend itself to specialization in the value chain. For example, DeepMind might be the big data + model provider for Apptronik’s edge compute + hardware expertise. Then there will likely be other players that run a vertically integrated strategy, which seems to be Tesla’s approach leveraging its in-house expertise from autonomous driving.
The race has just begun, and I’m eager to see how it plays out. Admitting my ignorance to the particulars of the space, I’m sure that much of what I’ve said here could turn out to be completely off base. Nonetheless, I think that the platform framework helps to rationalize the spend from Big Tech as they look to capitalize on one of the potentially huge next waves.
- 🍋