Latest / Deep Values: Value Investing Deep Dives / Mag7 Q2 Earnings Deep Dive #MSFT #META #GOOGL #AMZN
Transcript
- 0:00Imagine you're in the world's most high states drag race.
- 0:04You're You're absolutely slurring the pedal, the engine
- 0:07is screaming, and you are burning through premium fuel at
- 0:11just a record pace. But then you glance down at the
- 0:13dashboard and the speedometer is it's barely twitching.
- 0:17Yeah, it's a genuinely terrifying feeling when you're
- 0:20the one footing the bill for that fuel.
- 0:22I mean, in any rational system you expect a direct measurable
- 0:26correlation between the capital you expend and the the forward
- 0:29velocity you achieve. Right, right, exactly.
- 0:32And welcome to our deep dive into the recent Q2 tech earnings
- 0:35for The Magnificent 7. Specifically, today we are
- 0:38looking at Microsoft, Meta, Amazon and Google.
- 0:40We're going to unpack what has essentially been the most
- 0:42violent tech stock pullback in human history over this past
- 0:46month. Yeah, it's been wild.
- 0:48It really has the ultimate race in Silicon Valley right now
- 0:51boils down to artificial intelligence, capital
- 0:54expenditure, CapEx versus actual AI revenue.
- 0:57It's the fuel versus the speedometer.
- 0:59But real quick, before we get into the weeds, I want to
- 1:02mention that to access more advanced researches, follow
- 1:05Gurus's latest movement and leverage our latest Deep values
- 1:08agents. Welcome to download Deep Values
- 1:10APP from App Store. And to track stock market
- 1:13temperature. Welcome to download iOS app
- 1:15Buffett indicator from Apple App Store or Bubble Watch from
- 1:18Google Play. The links are in the
- 1:20description. Yes, check those out and also
- 1:23our standard disclaimer. All analysis is based on
- 1:26publicly available sources and may be subject to revisions or
- 1:29differing interpretations. All contents from this channel
- 1:33is for research and entertainment purposes only and
- 1:36does not constitute any investment or financial advice
- 1:38or service. Conduct your own due diligence
- 1:41before making investment or financial decisions.
- 1:43Absolutely. OK, so let's set the stage here.
- 1:45We didn't just wake up to a sudden drop, right?
- 1:48I mean, we were we were sitting on a powder kick.
- 1:50We absolutely were. To understand the shock we just
- 1:53saw in the markets, you have to look at how information actually
- 1:56moves today. Because traditional investment
- 1:58banks are they're no longer driving the narrative speed.
- 2:02That power has almost entirely shifted to influencers on
- 2:07Twitter, sub stack, specialized Discord channels.
- 2:10So it's much more decentralized. Exactly.
- 2:12Information and more importantly, market sentiment
- 2:15now spreads globally in hours instead of weeks.
- 2:18Wow, so the the learning curve and the panic curve have both
- 2:21become incredibly steep. Very steep, and that speed led
- 2:25to a massive buildup of leverage.
- 2:26By May, you had major funds holding 30, 3 to 35% of their
- 2:31entire portfolios in just semiconductor positions.
- 2:33That is heavily concentrated. Yeah.
- 2:35And then as we moved into June and July, retail leverage, you
- 2:39know, regular people borrowing money to buy options, it hit an
- 2:42all time. High.
- 2:43Oh man. The market was stretched
- 2:44incredibly thin, so we were primed for a violent correction
- 2:48based on the absolute slightest whisper of panic over AI return
- 2:52on. Investment.
- 2:52OK, let's unpack this cause. To understand why that panic
- 2:55actually triggered, we have to we have to move away from the
- 2:57stock charts and walk right into the server.
- 2:59Room right into the data centers, yeah.
- 3:01Yeah, we need to look at the physical, concrete reality of
- 3:03what these companies are actually buying.
- 3:05The economics of the physical AI data center are really the root
- 3:09of this entire situation, OK? Because if we look back at early
- 3:122024, the narrative was actually quite comforting for Wall
- 3:15Street. How so?
- 3:17Well, AI annual recurring revenue was growing faster than
- 3:20the CapEx. Investors were seeing the
- 3:22absolute dollar gap between spending and earning start to
- 3:26shrink. OK, so it felt like it was
- 3:28paying off. Right.
- 3:29But moving into Q2, massive supply chain inflation
- 3:33completely changed the math. Wait, are we talking about the
- 3:35physical parts here? Like the memory and the the
- 3:37GPU's themselves. Primarily, yes.
- 3:40Let's look at the cost of a one GW data center.
- 3:43OK, what is? What does one GW even mean in
- 3:45real terms? Just to put that in perspective,
- 3:47one GW is enough electricity to power a medium sized city
- 3:51continuously. Are you serious?
- 3:53A medium city just for one data center.
- 3:55Yes, it's a massive amount of power and at the end of last
- 3:58year, A1 GW data center cost about $35 billion to build from
- 4:02the ground up. OK, 35 billion, that's already a
- 4:05lot. It is, but today, factoring in
- 4:08the transition to new silicon, that exact same capacity cost
- 4:12between 55 and 60 billion. A $25 billion increase in less
- 4:19than a year, Yeah. What is what is driving that?
- 4:23I mean, the sources mentioned things like B300 and Ruben
- 4:25Architectures, but what actually are those?
- 4:27Were they just the newest chips? Yeah, they are the next
- 4:29generation of silicon from manufacturers like NVIDIA and
- 4:33AMD. But it's not just the chips
- 4:36themselves that are more expensive.
- 4:38Because these new architectures are exponentially more powerful,
- 4:42they run significantly hotter. So you have to cool them
- 4:45differently. Exactly.
- 4:46You can't just plug them into an old rack, you have to install
- 4:49completely new liquid cooled infrastructure, massive new
- 4:53power delivery systems and and literally reinforced physical
- 4:57rafts just to hold the weight. The weight of the coolant and
- 4:59everything. Yeah, the entire ecosystem
- 5:01around the chip has to be upgraded.
- 5:02That is a 50% jump in total project cost.
- 5:06I mean it's like budgeting to build it a massive luxury
- 5:10skyscraper and halfway through construction the price of steel,
- 5:14concrete and labor absolutely double s.
- 5:17That's a perfect analogy. But your rental income
- 5:19projections haven't changed at all.
- 5:20How do you how do you model a return on a $60 billion building
- 5:24if you can't raise the rent? You can't.
- 5:26And the the rent in this analogy is the premium these tech
- 5:30companies can charge for their proprietary closed AI models,
- 5:33right? Right.
- 5:33Which brings us to the second massive pressure point causing
- 5:36the panic, which is the open source threat.
- 5:39Right, because if someone's giving it away for free.
- 5:41Exactly. Open source models like Metas
- 5:43Lama 3 have rapidly closed the capability gap.
- 5:46How close are we talking it? Used to be that closed
- 5:49proprietary models from companies like Open AI or
- 5:51Anthropic were six to nine months ahead of open source
- 5:55models in terms of raw capability.
- 5:57Which is a lifetime in tech. Right.
- 5:58But now that gap has shrunk to just three to four months.
- 6:02Oh wow, so your skyscraper cost twice as much to build and
- 6:07suddenly the city opens a completely free, beautifully
- 6:11landscaped public park right across the street that does, you
- 6:16know, 90% of what your luxury building does?
- 6:17That is the exact fear keeping investors awake at night right
- 6:21now. I bet.
- 6:23If free open source models are nearly as good, how can the big
- 6:27tech giants charge the massive subscription premiums required
- 6:30to pay off a $60 billion data center?
- 6:32So if the cost of the building has doubled and a free park
- 6:35opens across the street, yeah, how is anyone surviving this
- 6:38market? I mean, the crazy thing in the
- 6:40sources is that Microsoft actually found a limp hole.
- 6:42They. Really.
- 6:43Did they completely? Flipped the narrative and
- 6:45emerged as the the clear winner this quarter.
- 6:47Yeah, their copilot breakthrough this quarter was undeniably
- 6:50significant. We moved away from the
- 6:52experimental trial phase and into massive committed
- 6:55enterprise deployment. Yeah, the number is definitely
- 6:57backed that up. I mean the sources highlight
- 6:58that one multinational corporation just bought over
- 7:01100,000 copilot seats. We aren't talking about a, you
- 7:05know, a 5000 seat pilot program anymore.
- 7:09Buying 100,000 licenses shows real tangible trust from a
- 7:14massive enterprise that this tool actually works and.
- 7:17That trust wasn't accidental. It came from a distinct
- 7:19architectural change under the hood.
- 7:21What do you mean? Well, the secret to Copilot's
- 7:23recent success is that Microsoft actually stopped using standard
- 7:27language models for tasks like PowerPoint generation.
- 7:30Wait, really? Yeah, they swap them out and
- 7:32replace them with coding models. Wait, like like Python And C++?
- 7:36Yeah. How does a coding model help me
- 7:38center a text box on a slide or or format a spreadsheet?
- 7:42That sounds entirely counterintuitive.
- 7:44It it does sound strange at first, but you have to
- 7:46understand how coding models are architected compared to standard
- 7:50conversational AI. Think of a standard AI like an
- 7:54improvisational actor. It it predicts the next word
- 7:58that sounds right based on the prompt, but it doesn't
- 8:00necessarily plan ahead, right? It just goes word by word.
- 8:02Yeah, but a coding model is like an architect.
- 8:05It inherently possesses longer reasoning chains.
- 8:08When a model writes software code, it has to anticipate how
- 8:11variable A effects function B 10 steps down the line.
- 8:16Otherwise the code just breaks. Oh I see, so it has a built in
- 8:20mechanism to check its own math before it actually builds the
- 8:23output. Precisely.
- 8:24It features much better pre execution error checking that.
- 8:26Makes so much sense. Right.
- 8:28So when Microsoft applied that specific architecture to a
- 8:30complex multistep task like generating A20 slide PowerPoint
- 8:34presentation or writing macros for a massive Excel spreadsheet,
- 8:38the failure rate dropped drastically it.
- 8:40Stopped hallucinating, bad formatting.
- 8:42Exactly. It stopped being a frustrating
- 8:44gimmick and became a highly reliable enterprise utility
- 8:47that. Is brilliant product engineering
- 8:49just changing the brain inside the tool to suit the task?
- 8:53And their financial engineering seems just as aggressive, right?
- 8:55I mean, Azure, Microsoft's cloud platform, showed a 43% growth
- 8:59rate. Massive.
- 9:01I know the cloud is booming, but how are they pulling off those
- 9:04specific margins? A few different levers are being
- 9:07pulled simultaneously. First, there's an accounting
- 9:09shift that is finally bearing fruit.
- 9:11OK, what? Kind of shift open.
- 9:13AI's training revenue is finally being booked as Azure revenue.
- 9:17Oh I see. As Open AI scales their model
- 9:19training beyond their initial supercluster agreements, those
- 9:22new massive training compute hours are officially hitting
- 9:26Azure's books as top line revenue.
- 9:29So Microsoft is finally getting public credit for the massive
- 9:32amount of compute they are providing behind the scenes,
- 9:35Yes. And furthermore, Azure is
- 9:37actively and somewhat quietly raising prices.
- 9:41Just across the board they're. Raising prices on GPU's through
- 9:44both spot pricing, which fluctuates like Uber surge
- 9:48pricing based on demand and through locking in higher rates
- 9:51on long term contracts. They're also raising prices on
- 9:54standard CPUs mostly through subtle methods like reducing
- 9:57corporate volume discounts. Wow.
- 9:59They're squeezing every drop, yeah.
- 10:00But here's the real revelation from Microsoft's earnings, and
- 10:03it ties directly back to our discussion about the open source
- 10:06threat. The ROI bombshell from The
- 10:08Outline. Exactly.
- 10:10Microsoft revealed that hosting open source models on Azure
- 10:13actually yields a 30% return on investment.
- 10:16That equates to roughly a three-year payback period on the
- 10:20hardware Wait. I'm not sure I followed the math
- 10:22there. How is hosting a free open
- 10:25source model so profitable for them compared to hosting a
- 10:28proprietary 1? Because it changes who holds the
- 10:32leverage when Microsoft hosts Open AI's flagship models.
- 10:37Open AI has massive bargaining power.
- 10:39They demand a huge cut of the API revenue, right?
- 10:43Because it's their model, but. When Microsoft hosts a free open
- 10:46source model like Llama, there is no model maker demanding a
- 10:50cut of the profits. Microsoft gets to set the pure
- 10:53computing price for the customer, and they capture the
- 10:56entirety of that margin for themselves.
- 10:58Wow. So the open source park across
- 10:59the street doesn't hurt them because they own the robes
- 11:02leading to the park and they're charging a toll it.
- 11:04Is a massive structural advantage, to the point where
- 11:06Microsoft is the only major player in this group confidently
- 11:09projecting positive cash flow next year.
- 11:11That's incredible. Microsoft proved that if you own
- 11:15the cloud infrastructure and the user interface, you win.
- 11:18But what if you don't have a traditional cloud to sell?
- 11:21That seems to be exactly where Meta and Google are struggling.
- 11:24This is where the CapEx anxiety really takes hold of the market,
- 11:27especially with the companies we call the builders.
- 11:30Let's let's start with Meta. Their earnings presented this
- 11:35wild contradiction. I mean, their core business
- 11:38looks incredibly healthy. Ad revenue was great, user
- 11:41engagement was up, but the stock got punished anyway.
- 11:44Because the market looked right past the current quarters ad
- 11:47revenue and stared directly at the 2025 CapEx projections.
- 11:52And what do those look like? Meta's projected capital
- 11:54expenditure is approaching 200 billion dollars two. 100
- 11:57billion, yeah. To put that into perspective,
- 11:59that is 70% of their projected revenue and double their annual
- 12:03profit. The existential quotient
- 12:06institutional investors are asking is can you realistically
- 12:08build a second meta with this money which.
- 12:11Brings up something I saw in the tech press a few weeks ago.
- 12:14There was a a massive rumor that Meta was going to rent out its
- 12:17extra compute to outside developers to make some cash on
- 12:20the side. You need to offset that 200
- 12:22billion. The stock even spiked on the
- 12:25rumor, but then Mark Zuckerberg reportedly refused to do it.
- 12:28But wait, if they are spending 200 billion, why not just hire a
- 12:32dedicated Cloud team to handle it?
- 12:34Isn't he leaving billions on the table just out of stubbornness?
- 12:38It's. Not stubbornness, it's a cold
- 12:40realization of their own corporate DNA.
- 12:42It comes down to the realities of enterprise cloud
- 12:44architecture. Meta lacks the service level
- 12:47agreement or SLA capabilities of a true cloud provider like AWS
- 12:51or Azure. What?
- 12:52Goes into an SLA that medic can't handle.
- 12:54I mean they run Facebook and Instagram well.
- 12:56In the cloud world, an SLA is a a legally binding contract where
- 13:00you have to guarantee what's called 4.
- 13:02Nines of uptime 4. Nines, yeah.
- 13:04That is 99.99% guaranteed availability.
- 13:07Wow. If your servers drop below that,
- 13:09you owe your enterprise clients massive financial compensation.
- 13:12OK Meta's internal data centers are built for a consumer app
- 13:16culture. You know, move fast and break
- 13:18things, right? Right.
- 13:19If Instagram goes down for 10 minutes, it's embarrassing, but
- 13:22it doesn't violate a corporate contract.
- 13:25Their infrastructure simply isn't hardened to guarantee that
- 13:28level of external customer facing reliability.
- 13:31OK. So they literally just aren't
- 13:32wired culturally or structurally to be a landlord.
- 13:35Exactly, and furthermore there's a hardware mismatch.
- 13:38What do you mean Meta wants to hoard its absolute best newest
- 13:43chips, the B series we talked about earlier for training its
- 13:47own frontier AI models? Oh, so they.
- 13:49'D only rent out the older stuff.
- 13:50Right. If they were to rent out
- 13:51compute, they would only want to rent out their older, less
- 13:55efficient H series chips and. Nobody wants those but.
- 13:58The external market for renting compute wants the bleeding edge,
- 14:01so Meta is stuck they they can't provide the legal reliability
- 14:05and they won't provide the tier one hardware.
- 14:07That paints a much clearer picture.
- 14:09Now let's look at Google. On paper, they look like they
- 14:13are in a better spot than Meta. They had 82% growth in their
- 14:17cloud division, but their flagship AI model Gemini 3.5 Pro
- 14:22is significantly delayed. What is the bottleneck inside
- 14:26Google? Google is facing a very unique
- 14:29hurdle. It is not a lack of engineering
- 14:32talent, and it's certainly not a lack of financial resources.
- 14:34OK, what is it? Their primary bottleneck is
- 14:37compliance. Compliance.
- 14:38You mean they're they're bogged down by internal lawyers
- 14:40essentially? Yes, Google is culturally and
- 14:43structurally terrified of legal and brand safety blowback,
- 14:47especially regarding purchasing and utilizing training data.
- 14:51The copyright stuff, exactly. Think about the copyright
- 14:53lawsuits hitting the industry from publishers like the New
- 14:56York Times, right? Right.
- 14:57Competitors like XAI or the open source community are scraping
- 15:01data aggressively, but Google's internal procurement and legal
- 15:05review processes are so rigorous that they are moving at a
- 15:09glacial pace in acquiring new high quality data sets.
- 15:12They're just. Terrified of ingesting A
- 15:14copyrighted book or private user data getting sued so they're
- 15:17they're over filtering everything and.
- 15:19Because of that strict data diet, they couldn't utilize
- 15:21their full computing power. The sources note they only had
- 15:25half the compute allocated for Q2 training compared to what
- 15:28Open AI used. Half.
- 15:29Half Less approved data meant less compute utilized, which
- 15:32resulted in delayed models. But despite those model delays,
- 15:36their balance sheet isn't collapsing right, because Google
- 15:39has a a massive secret weapon that doesn't rely on software at
- 15:43all. Yes, Google's saving grace right
- 15:45now is hardware sales. Hardware.
- 15:48Specifically they are directly selling their custom AI chips
- 15:52called TP Use or Tensor processing units to their cloud
- 15:55customers. OK.
- 15:57I know NVIDIA dominates the chip market, so how is Google making
- 15:59money selling their own chips? Through unprecedented hardware
- 16:03margins, OK. Break that down.
- 16:04ATPU cost Google roughly $10,000 to manufacture because they
- 16:09design it entirely in house so. They completely skip NVIDIA.
- 16:12They. Completely skip the massive
- 16:14markup NVIDIA charges for its proprietary ecosystem.
- 16:18Then Google turns around and sells access to these Tpus
- 16:21directly to clients for the equivalent of 20,000 dollars.
- 16:23Double. That is a massive 50% margin
- 16:26business that is generating billions in single quarter
- 16:30revenue. So.
- 16:31It's effectively subsidizing their massive AI research costs.
- 16:34Exactly. So Meta and Google are sweating
- 16:37over building the best models, wrestling with compliance Slas
- 16:41and insane CapEx bills. Yeah, and then there was Amazon.
- 16:45Amazon has somehow found a way to profit massively without
- 16:48having the smartest proprietary model in the room.
- 16:50They're playing a completely different game.
- 16:52Oh yeah. Amazon Web Services or AWS had a
- 16:54phenomenal quarter. They accelerated to 36.7%
- 16:58growth, adding $4.6 billion in single quarter.
- 17:02New revenue and how? Exactly, are they doing it if
- 17:04they don't have a G PT4 or a Gemini trying people in through?
- 17:08A highly strategic partnership with Entropic AWS sells access
- 17:12to Entropic's AI model models to their existing cloud customers,
- 17:15and the mechanics of this deal are incredibly lucrative.
- 17:18Why was the cut? AWS takes a massive 50% cut of
- 17:21the API revenue generated. Wait. 50% Amazon takes half the
- 17:24revenue just for hosting it. Yeah that is wild.
- 17:27So Amazon isn't mining for gold, they just own the only highway
- 17:32to the gold mine and they are charging a 50% toll it.
- 17:36Is the ultimate toll booth strategy as Anthropic's usage
- 17:39doubled from March to June, AWS reaped massive financial rewards
- 17:44with almost zero additional research and development
- 17:47expenditure on the model itself, they.
- 17:49Just sit back. Yeah, Anthropic does the hard
- 17:51math. Amazon collects the check but.
- 17:54You know what's even more interesting about this toll
- 17:55booth strategy? It's what the source is called
- 17:58the spillover effect. Oh yeah, it's not just the
- 18:00AIAAPI calls making Amazon money.
- 18:03No. This spillover effect is a
- 18:05crucial metric for understanding the entire industry.
- 18:09Every single dollar spent on AI processing on AWS inevitably
- 18:13pulls up traditional cloud spending.
- 18:15How? Does that work?
- 18:16Think about it, you can't just run an AI model in a vacuum.
- 18:19You need surrounding infrastructure.
- 18:22Like what? You need massive data storage,
- 18:23you need networking bandwidth, and you need traditional CPU
- 18:26usage to run the application logic that talks to the AII.
- 18:29See. It's like a shopping mall.
- 18:31Amazon doesn't own the cool new anchor store.
- 18:33Anthropic does. Right?
- 18:35Because everyone is coming to the mall to visit Anthropic.
- 18:39Amazon is making an absolute fortune on the parking garage,
- 18:42the food court, and all the smaller traditional shops around
- 18:46it that. Is a perfect analogy.
- 18:47Amazon is seeing a massive uptick in the use of their
- 18:50traditional Graviton CPU chips which are their custom
- 18:53processors for standard non AI web hosting, entirely driven by
- 18:58the AI boom. Wow, AI is quite literally a
- 19:01rising tide lifting the entire AWS ecosystem.
- 19:05OK. So we've looked at the micro
- 19:07level. You know how Microsoft, Meta,
- 19:08Google and Amazon are all playing their hands differently,
- 19:11right? But we have to zoom out to the
- 19:12macro to wrap this up. We have to look at the total
- 19:15bill for the industry and. This is where the math gets
- 19:17truly daunting hit. Me with the numbers the.
- 19:19Entire tech sector's data center capital expenditure out over the
- 19:23next few years is currently heading toward a staggering $1.7
- 19:27trillion one. .7 trillion, that is the GDP of a major country
- 19:31just being poured into servers and cooling systems.
- 19:34Yeah, yes. And to justify a standard
- 19:37acceptable return on investment for that level of expenditure
- 19:41across the industry, the tech sector needs to generate
- 19:44somewhere between 600 billion to 800 billion in AI annual
- 19:47recurring revenue. But.
- 19:48I'm looking at the revenue figures for the biggest players
- 19:51and there is a massive gaping hole there, a huge.
- 19:54Hole even. If we take the most wildly
- 19:56optimistic projection, say Open AI and Anthropic absolutely
- 19:59crush it and hit 200 billion or even 300 billion in revenue
- 20:03combined. We are still hundreds of
- 20:06billions of dollars short of making that 1.7 trillion back.
- 20:09Yeah, who is going to pay the rest of that bill?
- 20:12That is the $1.7 trillion question.
- 20:14Where does that unseen revenue come from?
- 20:17To justify this massive build out, the market will have to
- 20:20rely on adoption we haven't seen yet.
- 20:22Like what? We are talking about massive,
- 20:24unprecedented spending from quantitative trading firms
- 20:27deploying AI for algorithmic trading.
- 20:30It will require nearly every major enterprise on Earth
- 20:33deploying and paying for open source models internally.
- 20:36And crucially, it relies on unseen internal efficiencies,
- 20:41meaning companies using AI to cut their own operational costs
- 20:45so dramatically that the savings basically count as new revenue
- 20:49IT. Sounds like the entire tech
- 20:51sector is taking a massive leap of faith.
- 20:53They. Really.
- 20:53Are they're? Building the infrastructure for
- 20:56a future economy that hasn't fully materialized yet.
- 20:59They are. The transition is clear.
- 21:01We have moved from a pure exciting race for AI capability
- 21:05like who has the smartest chatbot into a brutal,
- 21:09unforgiving math equation of CapEx versus Roy.
- 21:13Microsoft is navigating it by monetizing the interface and
- 21:16capturing the open source margin.
- 21:18Amazon is navigating it by being the ultimate tollbooth and mall
- 21:21owner. Google is saving their balance
- 21:23sheet by selling hardware, and Meta is betting the entire farm
- 21:26on their internal ad engine and future capabilities.
- 21:29They are all playing different angles of the same massive
- 21:32gamble. So.
- 21:33What does this all mean for the future of this tech race?
- 21:35Well, I won't. Leave you with a thought that
- 21:36goes beyond the balance sheets in the silicon.
- 21:39We spend so much time obsessing over whether these companies can
- 21:42physically build the compute, secure the electricity, and
- 21:45manufacture the chips to reach this $1.7 trillion scale right?
- 21:50That's the whole focus, but. What if the ultimate bottleneck
- 21:53to achieving 800 billion in AI revenue isn't the hardware at
- 21:58all? If.
- 21:58It's not the hardware or the electricity.
- 22:00What else could it be? The human capacity to adapt.
- 22:04Think back to that multinational corporation that just bought
- 22:07100,000 copilot seats, right? The big win for Microsoft The.
- 22:11Intelligence has been built, the software is successfully
- 22:14deployed on their network, but if regular employees, middle
- 22:18managers, and traditional businesses cannot fundamentally
- 22:21change their daily ingrained workflows fast enough to
- 22:25actually use those 100,000 seats effectively, then what happened?
- 22:28The subscription won't get renewed.
- 22:29The rent won't get paid. The machines might be ready to
- 22:32change the world, but the humans might simply be too slow to
- 22:35adapt to them. Wow.
- 22:37If we can't physically change how we work on a day-to-day
- 22:40basis, the whole $1.7 trillion drag race hits a brick wall and
- 22:46suddenly that speedometer not moving makes a whole lot of
- 22:48terrifying sense. Thank you all for joining us on
- 22:51this deep dive. Keep questioning the consensus.
- 22:54Keep looking past the headlines and before we go, our standard
- 22:58disclaimer. All analysis is based on
- 23:00publicly available sources and may be subject to revisions or
- 23:03differing interpretations. All contents from this channel
- 23:06are for research and entertainment purposes only and
- 23:08do not constitute any investment or financial advice or service.
- 23:12Please conduct your own due diligence before making
- 23:14investment or financial decisions.
- 23:16We will catch you next time.