The bull case for Apple
Tech X (Twitter) is, seemingly every other week, dunking on Apple. First, they were behind to the large language model (LLM) game.
“Where is Apple?” Twitter was asking.
Open AI is publishing models. Amazon has models. Facebook has models. Microsoft has models. Google has models. Where are Apple’s models?
Then, now two months since Vision Pro came out, people are saying they are not using the devices as frequently as they expected. Uh oh! Vision Pro isn’t the next iPhone!
Has Apple lost it?
Tech twitter can sort of turn themselves upside down at times. I don’t own any Apple stock, but I think there are a few reasons to be bullish on Apple.
Reason 1: LLMs
I’ll sing this from the rooftops: Large Language Models ((LLMs) are not competitive advantages. They’re table stakes, and they’ll be open source.((Every business will have their own and it won’t cost a fortune to train.))
The competitive advantage in an LLM is and will be the data that goes into them.
I do not understand why startups are raising $100M+ to buy GPUs to train LLMs. They’re all being trained on the same data!
People go to the Common Crawl, download all the data (it’s free and open source), and then they use that to train the LLM.
Right now, LLMs are scaling based off adding parameters to the model and throwing more hardware at it, but the tech will get more and more refined over the next few years, and we’ll reach the upper limit of what the LLMs can do.
Then, that tech will be open sourced (probably by Meta). That’ll be table stakes.
Then, the challenge in LLM output performance will be the proprietary data that you can put through it.
For Apple, it doesn’t serve a single business purpose to begin competing with Open AI and or Amazon or Meta in the large-scale LLM space. They don’t currently have a cloud offering, they don’t really depend much on ads (their ad algorithm in the App Store is rudimentary at best), and they don’t seem committed to building out a developer cloud (although I think it would be great for their stock price). So why spend the money? Why spend the energy?
Where LLMs will become very useful for Apple will be within Siri and turning Siri into an agent.
Imagine: Apple has access to your email and reads that you are going on a business trip to Nashville in a few weeks. It knows you usually fly Southwest to Nashville and that you like to stay at the Marriott downtown. It asks if you want some travel plans prepared. It finds the hotel in the Marriott app and flights in the Southwest app, of course, corresponding with your calendar. It pre-orders an Uber to the airport leaving you the desired amount of time at the airport before your flight. All you do is press “book.”
Developers are building prototype agents for this type of thing now using large language models. Of course, they’re just prototypes, and making them work would require your giving up all your life’s data, so you probably won’t use just any random company’s service.
But you might use Apple’s.
Apple wants all this to take place on-device((they’ll market it as private and secure)), which means that they need to make the LLMs small enough and energy-efficient enough that, instead of running on an NVIDIA GPU in the cloud, they run on an iPhone in your pocket.
Apple’s strategy isn’t to train large models to run on huge GPUs in the cloud. They need to compress models to run on-device.
These types of updates don’t magically appear. They’re incremental. Incremental hardware upgrades that will drive new software capabilities.
Did you catch that?
To get incredible advancements like the above, they need hardware upgrades.
My expectation here is that Apple isn’t behind at all. Quite the opposite: they’re solving their actual problem.
How to convince people that used to be on 2-year upgrade cycles and are now on 4-year upgrade cycles to go back to 2-year cycles.
For Apple, AI is software that drives consumer purchases and compressed upgrade timelines of hardware. The iPhone isn’t going anywhere. In fact, I’d argue I’m more bullish about it now than I was 3 years ago.
Which brings us to their new product, Vision Pro.
Reason 2: Vision Pro
Tech twitter wants to think that the Vision Pro hasn’t taken off like the iPhone, and is not on a path to, therefore it’ll be DOA. This assumes that Vision Pro is the “next iPhone,” which it’s not. Vision Pro, in its current state, might become the next Mac, but right now it’s being billed as a Mac addition.
People seem to forget that the iPhone wasn’t huge until the 4th gen. Gen 1 iPhone was slow, only worked on AT&T, and didn’t have an App Store. It was, however, still better than most phones on the market, albeit radically different.
The other advantage the iPhone had was that everyone already had a cell phone and had gotten used to using it. The iPhone was a huge leap forward in tech that already existed, so it’s easy to get people to switch.
Vision Pro isn’t that. Nobody outside of gaming is seriously using VR or AR technology. Outside of Meta’s products, I can’t think of any that have come close to being commercial successes. Jason Calicanis, an angel investor in Silicon Valley, refers to VR as being “Try, ‘Oh My!’, Goodbye.” Meaning you try it, are shocked by how cool it is, then put it down and never pick it up again.
This happened to me, too. I had an HTC VR setup which I loved for about 7 days. Then, I put it in my closet and threw it away 3 years later.
Vision Pro as a Mac addition, however, is a different value proposition. It’s a B2B device. Tech hobbyists may have purchased it, but its initial use cases are for businesses.
An elevator technician can do training and collaborate with folks at the home office using Vision Pro. A plumber can do the same.
An expert in London can help troubleshoot a real-world problem in Bali thanks to Vision Pro.
Recently, a surgeon successfully completed the first surgery done fully within Vision Pro.
An executive can use Vision Pro on an airplane to review company financials without anyone peering over their shoulder.
That use case is enough for right now! That’s all that’s needed. It doesn’t need gaming (which I heard is great) or sports (which is super promising) or movies (which I heard is fantastic), it just needs businesses.
Maybe, someday, Vision Pro will become Vision and will be a B2C device, but, right now, all it needs are enough businesses to buy it to continue to further fund hardware development. Mac started as a business device, and it still is. Over time, however, people bought Macs for personal use.
The other aspect to Vision Pro being different than the iPhone is that it’s being geared toward developers first and consumers second, similar to the Mac. The Mac was a platform meant for pro users and dev users to build on top of first, then for consumers to leverage second. The iPhone, however, shipped locked-down. There were zero developer capabilities.
These vastly different strategies have vastly different KPIs. In one, you’re looking for business-focused applications and clients. You compromise quantity sold for price per unit (see: $3500 for the device). With B2C products, your assumption is that you’re already leveraging economies of scale in manufacturing and production, so you aim to make up the balance with quantity sold.
Feature image credit: ChatGPT with the following prompt “An image that’s a clever feature image for a blog titled ‘the bull case for Apple'”