Latest / Deep Values: Value Investing Deep Dives / Apple Pays Google Now?
Transcript
- 0:00On January 12th, 2026, we saw something that, I mean it
- 0:04genuinely made the entire tech world just stop and take notice.
- 0:08It really did. Two absolute Titans, Apple and
- 0:11Google, companies that have been, you know, fierce rivals
- 0:13for decades. They confirmed this colossal
- 0:18multi year partnership. And this wasn't just another
- 0:21press release. This felt different.
- 0:22This was a seismic shift structurally.
- 0:24Absolutely. The core of the deal is this
- 0:27Apple, the company that prides itself on controlling, you know,
- 0:30the entire hardware and software stack, is licensing Google's
- 0:33Gemini AI models. The powerhouse models.
- 0:36Exactly. And the goal is to finally power
- 0:38up the next generation of Siri, plus a whole suite of features
- 0:41they're calling Apple Intelligence.
- 0:43And we're going to see that debut with iOS 26.4 right in the
- 0:47spring of 2026. This is basically the moment
- 0:50Apple admitted it needed to outsource its brain it.
- 0:52Is and that's precisely why we're doing this deep dive.
- 0:55Our mission today is to really get under the hood of this
- 0:58thing. We need to unpack the strategy.
- 1:00You know what forced this alliance?
- 1:03And we have to look at the tech, the very unique architecture
- 1:05they built to solve this massive privacy issue at the heart of
- 1:10the deal. And of course, follow the money,
- 1:12because this transaction, at the end of the day, it determines
- 1:15who controls the intelligence layer for the entire mobile
- 1:19Internet. That's the prize.
- 1:20And here's the kicker. Especially if you know the
- 1:22history between these two For years, Google paid Apple.
- 1:27What was it? Something like $20 billion a
- 1:30year. An estimated 20 billion, Yeah,
- 1:32to be the default search engine in Safari.
- 1:34They were buying distribution. Right, but in this new AI deal,
- 1:38the roles have completely reversed.
- 1:40Apple is the one paying Google. A total flip in the power
- 1:43dynamic. It's a fundamental shift.
- 1:45OK, let's unpack this. Where?
- 1:46Do we start? Let's start with the elephant in
- 1:48the room, Apple. Why did the world's most
- 1:51vertically integrated, successful company essentially,
- 1:54well, wave the white flag on building its own foundational
- 1:57AI? What was the breaking point?
- 1:59The pressure was immense. It was an existential crisis
- 2:04that you could really call the Siri problem.
- 2:06The Siri problem. I like that.
- 2:07For more than a decade, Siri was built on these rigid command and
- 2:10control databases. It's basically a glorified look
- 2:13up tool. You ask for the weather, looks
- 2:15up the weather and that's. It exactly, But compared to
- 2:17modern LLMS, you know, the ones that can actually reason and
- 2:20generate text, Siri didn't just look outdated, it looked frankly
- 2:24brittle, unreliable. It felt dumb.
- 2:27It felt dumb, yeah, and that intelligence gap became a huge
- 2:31liability for their premium brand.
- 2:33We know they were trying, right? We heard those code names Ajax
- 2:36and Ferret. Why couldn't they close that gap
- 2:39themselves? Well, it was a mix of cultural
- 2:41and logistical problems internally, Apple's famous
- 2:45culture of extreme secrecy and silo teams, it actually works
- 2:49against the kind of massive, coordinated effort you need to
- 2:52train a frontier model. So the very thing that makes
- 2:54them good at hardware hurt them in AI.
- 2:57Precisely. But the even bigger blocker was
- 3:00data. To train a model to compete with
- 3:03something like Gemini Ultra, you need a colossal real time search
- 3:08indexed understanding of the world.
- 3:10Which Google has in spades. Google and Microsoft have it,
- 3:14Apple just didn't. So this deal, it's officially
- 3:17Apple's short term capitulation. They've given up on the build
- 3:20strategy for now. They're treating the AI model
- 3:23itself as a kind of commodity. Like electricity?
- 3:26Yeah, or cloud compute. They're just plugging into the
- 3:29wall to get the power they need because building their own power
- 3:32plant would take too long. They outsource the heavy
- 3:35lifting. OK so if Apple is admitting they
- 3:37couldn't build it, what does Google get out of selling them
- 3:40the crown jewels? This feels like a massive
- 3:42resurgence narrative for Google. Oh, it's an enormous win.
- 3:45First, it's just pure technical validation, right?
- 3:48When you get Apple, arguably the most demanding consumer
- 3:50electronics company in the world, to sign on, you've proven
- 3:53your model as top tier. It validates Gemini against
- 3:56competitors like HTT 5 and Claude.
- 3:58It's the ultimate stamp of approval.
- 4:00And the second reason it has to be money.
- 4:02It's infrastructure monetization.
- 4:04Google has spent 10s of billions of dollars on their custom TPU
- 4:08chips and data centers. This deal with Apple guarantees
- 4:12high volume usage of that infrastructure immediately.
- 4:15So it accelerates their return on that massive investment.
- 4:19Massively, they go from building capacity to selling it at
- 4:22incredible scale. We should probably clarify what
- 4:25about Open AI? Apple did have a partnership
- 4:27with them first. Where do they fit into this
- 4:29picture now? The sources are pretty clear on
- 4:31this. The Gemini deal basically pushes
- 4:34Open AI into a sort of supporting role.
- 4:38A niche player. Exactly, for niche queries,
- 4:41maybe something a user has to actively opt in to use.
- 4:45Gemini is now the default core reasoning engine.
- 4:48In hindsight, that Open AI partnership looks like a
- 4:50stopgap. Or maybe leverage to get a
- 4:52better deal out of Google. That's also very possible
- 4:54leverage until Google could offer the privacy tech we're
- 4:57about to get into. Let's circle back to the money,
- 4:59because that reversal is just stunning.
- 5:01The old deal, the ISA. Google pays Apple 20 billion.
- 5:05Right. And now the new AI licensing
- 5:07agreement, the money flows the other way.
- 5:10Apple pays Google an estimated $1 billion as an initial annual
- 5:15value. They're not buying distribution
- 5:16anymore. No, they're buying raw
- 5:18capability. Pure intelligence. $1 billion
- 5:21sounds like a huge number, but I suspect for Apple it's a
- 5:25bargain. It's an incredible bargain, our
- 5:27analysis suggested, for Apple to build, train and maintain a
- 5:31competitive model and the server cluster to run it.
- 5:34It could have easily cost them 50 to 100 $4 billion / 5 years.
- 5:38So for 1 billion a year they get instant world class AI without
- 5:44even touching their massive cash reserves.
- 5:47It's brilliant financial engineering.
- 5:48It lets them keep doing those huge shareholder buybacks.
- 5:51So I'm guessing Google isn't just letting them, you know,
- 5:53slap an Apple logo on Gemini. What are the branding rules
- 5:56here? This is key.
- 5:58It's what the deal calls the white label mandate, meaning the
- 6:01Gemini integration has to be totally invisible to you, the
- 6:04end user. Apple is obsessed with
- 6:07maintaining the illusion of seamless vertical integration.
- 6:11So when the new Siri answers you, it just says Siri.
- 6:13It protects the brand. At all costs, and the timeline
- 6:16for this magic trick is fast. They're targeting the iOS 26.4
- 6:20update in March of April 2026, which lines up perfectly with
- 6:25their spring hardware marketing push.
- 6:26OK, this brings us to what had to be the biggest hurdle, the
- 6:30core contradiction. Apple, the champion of privacy,
- 6:34is getting into bed with Google, the data advertising giant,
- 6:37right? How on earth did they solve
- 6:39that? This must have been the sticking
- 6:40point for months. It absolutely was, yeah.
- 6:43And they solved it with a really ingenious complex 3 tiered
- 6:46hybrid architecture. And it's all managed by a new
- 6:48layer they call the Semantic router.
- 6:50The semantic route, What does it do?
- 6:52It decides based on how sensitive and how complex your
- 6:55request is, where that request gets processed.
- 6:57OK, so let's break down the tiers.
- 6:59Tier 1. Tier one is all on device if you
- 7:02ask Siri to, say, set a timer or launch an app, that's handled by
- 7:07small language models running right on the neural engine chip
- 7:09in your phone. So that data never leaves your
- 7:11device. Never.
- 7:13Privacy is absolute and the latency is 0.
- 7:15That makes sense for the simple stuff.
- 7:17What about more personal tasks? That's Tier 2 private Cloud
- 7:21compute, or PCC. This is for intermediate tasks
- 7:24that are highly sensitive. Think about it, indexing your
- 7:27personal photos or analyzing your health data from the health
- 7:30app. So where does that go?
- 7:32That request goes to Apple's own private cloud compute.
- 7:36These are custom Apple servers running an ephemeral OS.
- 7:39The process is cryptographically guaranteed to delete your data
- 7:43right after it's used. So tiers one and two keep all
- 7:46your really personal data inside Apple's walled garden.
- 7:50But the truly heavy lifting the frontier level reasoning that
- 7:54has to go to Google. That's Tier 3, right?
- 7:56That is tier 3 Gemini cloud injection.
- 7:59This is for the really complex stuff.
- 8:02Multi step planning, writing a poem, anything that needs
- 8:05Gemini's huge world knowledge. The semantic router hands that
- 8:09off. And here's where they solve the
- 8:11privacy paradox. OK, I'm listening.
- 8:14Google developed a new system called Private AI Compute and
- 8:17the core of it is this new hardware they built called
- 8:20Titanium Intelligence Enclaves or TIE.
- 8:23Titanium intelligence enclaves. You can think of a TIE as a
- 8:26secure armored vault built deep inside Google's massive data
- 8:31centers. So Apple is basically renting a
- 8:33secure room inside Google's house.
- 8:36That's the perfect analogy. Apple encrypts your prompt on
- 8:39your phone, sends it to Google's TIE and it is only decrypted
- 8:42inside that hardware secured enclave.
- 8:43And then it's processed by Gemini inside that isolated box.
- 8:46Exactly and the answer is immediately re encrypted and
- 8:48sent back to your phone. But how do we know Google isn't
- 8:51peeking? How do we know their own
- 8:52engineers aren't logging that data to train their models?
- 8:55That's the no training covenant, and it's not just a promise,
- 8:59it's enforced by the hardware itself.
- 9:01The Thai architecture is built to be technically incapable of
- 9:05logging the prompt or the answer.
- 9:07The data is ephemeral by design. So the hardware guarantee.
- 9:10It's a hardware guarantee. That technological shield is the
- 9:13linchpin. It's what finally allowed this
- 9:15deal to happen. OK, let's bring this back to the
- 9:17user. All this amazing complexity,
- 9:19what does it actually mean for you when you pick up your phone?
- 9:22What's the Holy Grail experience they've unlocked here?
- 9:25The Holy Grail, in a word, is context.
- 9:28The new Siri, powered by Gemini's reasoning, can now
- 9:32index your entire digital life, your photos, your emails, your
- 9:36calendar to build a semantic map of you and.
- 9:39What does that enable? Give me an example.
- 9:40OK, personal context. Remember the old Siri
- 9:42frustration? You'd ask when is my mom's
- 9:45flight landing? And Siri would say I can search
- 9:48the web for that, yes. Infuriating.
- 9:50The new Siri uses Gemini to find the flight confirmation e-mail
- 9:54in your inbox, extract the flight number, query a real time
- 9:58flight API, and then just give you the answer all in one go.
- 10:01It knows where the information is, so you don't have.
- 10:03To exactly then you have on screen awareness.
- 10:07This is multi modality. Siri can now see what you're
- 10:10seeing. So if I'm looking at a picture.
- 10:12Let's say you're on Instagram looking at a picture of a
- 10:14landmark in Paris. You can just ask Siri.
- 10:17Book a table at a good Bistro near here for tonight.
- 10:20The Vision model sees the landmark, knows it's Paris, and
- 10:24kicks off the reservation. It becomes a true agent.
- 10:27The most impressive thing to me is chaining those actions across
- 10:30different apps. That's cross app app intense.
- 10:34Gemini is the conversational glue.
- 10:36You can give a really complex, messy command like find the
- 10:40photo of me at the beach, make it look more cinematic, and then
- 10:43e-mail it to Jamie and. It just.
- 10:45Does it? It translates that one sentence
- 10:47into a precise sequence of API calls that the operating system
- 10:50executes. It opens photos, then an editor,
- 10:53then mail. That kind of multi step planning
- 10:55was just impossible before this. Let's quickly touch on the
- 10:58financial fallout again. We said Apple avoids maybe 50 to
- 11:02100 billion in CapEx. And that capital efficiency is a
- 11:05core reason their stock is so resilient.
- 11:08They buy world class capability for a billion dollars and use
- 11:12the rest for massive shareholder returns.
- 11:14It's their playbook. And for Google, that billion
- 11:17dollar contract is more than just revenue.
- 11:19It's foundational. It gives Google Cloud a stable
- 11:22recurring high margin revenue stream that diversifies them
- 11:26away from the volatile advertising market.
- 11:29And the market noticed that deal pushed Alphabets market cap pass
- 11:33$4 trillion, briefly overtaking Apple.
- 11:36Because investors saw Google grabbing control of the core AI
- 11:39infrastructure layer. Exactly.
- 11:41But when you have that much consolidation, it lands you
- 11:44right in the middle of an antitrust minefield.
- 11:46We saw people like Elon Musk calling it an unreasonable
- 11:49concentration of power. What's the real regulatory risk
- 11:53here? The concern is legitimate.
- 11:55One company, Google, is now the AI brain for the 2 operating
- 11:59systems, iOS and Android, that have 99% of the market.
- 12:03Regulators are worried they're locking out smaller competitors.
- 12:06So how do they defend this in court?
- 12:08They draw a really critical line between this deal and the old
- 12:12illegal search agreement. In that deal, Google paid for
- 12:16exclusivity. And here.
- 12:18Here, Apple is the buyer, and they can argue they're just
- 12:21choosing the most capable supplier based on technical
- 12:24merit. And crucially, the deal is non
- 12:27exclusive. You can still use the Open AI
- 12:29integration. So that non exclusivity is their
- 12:32legal shield. It's a very powerful one.
- 12:35It allows them to argue they're fostering competition, not
- 12:37stifling it. So when 2 giants like this get
- 12:39together, someone always gets squeezed out.
- 12:42Who are the biggest losers in this new landscape?
- 12:45Well, the biggest loser, without a doubt is Open AI.
- 12:48Really, they've been marginalized to a secondary
- 12:50role, and by losing that default status on the iPhone, they lose
- 12:55access to that constant high frequency stream of queries.
- 12:58Which is the data you need to make your model better.
- 13:00It's the lifeblood. It's the feedback loop for
- 13:02reinforcement learning. Now they have to rely on their
- 13:04own app and Microsoft for distribution.
- 13:06It's a major blow. Who else takes a hit?
- 13:08I think Samsung does. They put a lot of marketing into
- 13:11Galaxy AI, but now that's a bit diluted because Galaxy AI is
- 13:15also powered by Gemini. So the intelligence is
- 13:17commoditized. It is.
- 13:20If the top Samsung phone and the new iPhone are both running on
- 13:23Gemini, Samsung is forced to compete on hardware alone on
- 13:27things like folding screens. They lose that AI
- 13:30differentiator. And meta seems completely
- 13:33trapped. They don't have an operating
- 13:35system. They are trapped.
- 13:36Meta with their open source Llama models is stuck inside
- 13:39their own app silos. They depend completely on Apple
- 13:43and Google for distribution. Their assistants will likely be
- 13:46blocked or throttled from being true system wide agents.
- 13:50They're confined to Instagram and WhatsApp.
- 13:52So if we step back, this whole thing is a classic marriage of
- 13:55convenience. It solves Apple's existential
- 13:58Siri crisis overnight, and at the same time it secures
- 14:01Google's relevance and future revenue in this new agent
- 14:05economy. Exactly.
- 14:06But here's the final provocative thought, the tension that's
- 14:10going to define the next decade. This Gemini deal buys Apple,
- 14:13what, two or three years of breathing room?
- 14:15But we know Apple is still quietly working on its own
- 14:18ferret models in house. And critically, they're now
- 14:21collecting all this anonymized user data via the Semantic
- 14:24router. And using it to fine tune their
- 14:26own models. Precisely.
- 14:28It is highly likely that by the time we get to iOS 28 or 29,
- 14:32Apple will start to silently swap out the Gemini back end for
- 14:35its own model, query by query, slowly weaning itself off that
- 14:39licensing fee. So the ultimate goal is still
- 14:41vertical integration. The defining conflict is that
- 14:44tension Apple's deep seated desire to own the stack versus
- 14:48Google's desire to be the permanent utility provider.
- 14:51And that deeper shift from the app economy where you go to
- 14:54websites to the agent economy where your device just does
- 14:56things for you, that is the profound change that affects how
- 15:00you will use your phone every single day.
- 15:02Something for all of us to watch as that battle for the stack
- 15:06plays out. Wilson.
- 15:07And before we wrap up, just a quick note for you.
- 15:09If you are interested in tracking this kind of market
- 15:13temperature and these huge financial shifts we've been
- 15:15talking about, we encourage you to search for and download the
- 15:19iOS app Buffett Indicator from the Apple App Store.
- 15:22It's a tool we developed and it can give you daily updates on
- 15:25market. Tonya, it's our latest tool
- 15:28giving you daily updates and monitoring of the market bubble.
- 15:31We welcome you to check it out. Thank you for joining us for
- 15:33this deep dive. Disclaimer All analysis is based
- 15:37on publicly available sources and may be subject to revisions
- 15:40or differing interpretations. The content is for educational
- 15:44purposes only and does not constitute investment advice.
- 15:47Conduct your own due diligence before making investment
- 15:49decisions. If you want to listen to our
- 15:52full deep dive about this topic or access more advanced
- 15:55contents, welcome to join Deep Value Club membership or
- 15:58subscribe us on Apple and Spotify podcasts channel Deep
- 16:01Value investingandluluvc.com.