Latest / Deep Values: Value Investing Deep Dives / Why Google Rents AI From SpaceX?
Transcript
- 0:00You know, usually when a company releases its earnings report,
- 0:03the math makes intuitive sense. Yeah, generally speaking, you
- 0:06sell more, you make more. Exactly.
- 0:07You sell more products, you make more money, your stock goes up.
- 0:10It's a very simple, comforting equation.
- 0:13Right, the market likes predictability.
- 0:14Right. But then you look at what
- 0:17happened in August of 2026 and suddenly that whole equation
- 0:21completely shatters. Oh, absolutely.
- 0:23It remains one of the most perplexing market reactions in
- 0:26modern tech history. Really.
- 0:28It really is. So today we are pulling from
- 0:30Google's Q 22026 earnings reports, a completely wild SEC
- 0:36regulatory filing from SpaceX, and some recent Bloomberg leaks.
- 0:40Because we need to figure out why the market punished a
- 0:43company for achieving basically absolute perfection.
- 0:46Right, a total disconnect. So let's unpack this for you
- 0:48because you really have to understand the sheer scale of
- 0:51what happened here. It's August 2026, Google
- 0:54releases their Q2 earnings and the numbers are just staggering.
- 0:57I mean, their cloud business grew by 82%.
- 1:00Which is massive. Massive.
- 1:03They hit the highest single quarter profit in the history of
- 1:06human corporate enterprise. Literally a flawless victory on
- 1:09paper. Exactly.
- 1:10On paper, yeah. And yet, the very next day,
- 1:13their stock crashed by 7%. Yeah, a massive sell off.
- 1:17So the mission of our deep dive today is to unravel this
- 1:20mystery. Cause the answer isn't in the
- 1:22profit numbers right? No, not at all.
- 1:24It's buried in a hidden $30 billion contract that explains
- 1:28everything. Right, we are looking at a
- 1:30secret deal with Elon Musk's SpaceX and honestly, a
- 1:35terrifying prisoner's dilemma at the heart of the whole AI
- 1:39industry. It is such a fascinating story.
- 1:41It's fundamentally about the physical limits of
- 1:43infrastructure, corporate desperation, and while some of
- 1:47the most aggressive high stakes financial engineering we have
- 1:49ever seen. It really is.
- 1:51But before we get into the mechanics of this massive
- 1:54financial anomaly, I want to share something with you to
- 1:56access more advanced researches, applications and our latest deep
- 2:00values agent. Welcome to explore and join deep
- 2:03values dot AI. And if you want to track total
- 2:06market temperature, welcome to search and download the iOS app
- 2:10Buffett Indicator from the Apple App Store or Bubble Watch from
- 2:14Google Play all. Right.
- 2:15And as always, a quick note before we proceed.
- 2:18Disclaimer All analysis is based on publicly available sources
- 2:22and may be subject to revisions or differing interpretations.
- 2:26All contents from this channel is for research and
- 2:28entertainment purposes only and does not constitute any
- 2:31investment or financial advice or service.
- 2:34Conduct your own due diligence before making investment or
- 2:37financial decisions. All right, let's get into it.
- 2:40So the stock didn't drop because of the profit.
- 2:42No, it dropped because of some incredibly alarming clues buried
- 2:46deep in the cash flow statement. Yes, the cash flow statement.
- 2:49Help us understand why investors were looking there instead of,
- 2:52you know, the record-breaking income statement.
- 2:54Well, the income statement is essentially A narrative
- 2:56constructed by accounting rules, right?
- 2:58OK, a narrative. Yeah, it can be smoothed out
- 3:00like you can recognize revenue over time or you can delay
- 3:04recognizing certain costs. Oh, I see.
- 3:07But the cash flow statement is brutally honest.
- 3:10It strips away all the accounting magic and tells you
- 3:13exactly what actual dollars are entering and leaving the bank
- 3:16accounts on that specific day. Right cold hard cash.
- 3:20Exactly. And in Q22026, the sheer volume
- 3:24of actual cash leaving Google's bank accounts was unprecedented.
- 3:28Unprecedented. Almost feels like an
- 3:29understatement, honestly, when you look at the raw numbers.
- 3:32Oh. Absolutely.
- 3:33Their quarterly capital expenditure, what they're
- 3:35physically spending on equipment, servers and
- 3:37facilities hit a mind blowing $44.9 billion.
- 3:42Just an astronomical figure. Right.
- 3:44And because of that massive outflow of cash, their free cash
- 3:47flow actually went negative, negative $5.85 billion to be
- 3:52exact. For money printing machine like
- 3:55Google, seeing free cash flow go negative is incredibly rare.
- 3:58Very rare. I mean, looking at a rolling 12
- 4:01month period, their free cash flow plummeted from 64.4 billion
- 4:05down to just 5.3 billion. And that sudden cash reinforce
- 4:09them into a very unusual financial maneuver.
- 4:11Yeah, this is the part that really surprised me.
- 4:14Right, because Google is a company sitting on roughly $240
- 4:18billion in cash and cash equivalents, right?
- 4:21But they suddenly decided they needed to raise more money.
- 4:24They raised $49.6 billion by issuing stock and another 20.3
- 4:30billion in senior notes. Wow.
- 4:32By doing this, they essentially doubled their long term debt,
- 4:35bringing it to an astonishing 980 billion dollars.
- 4:39Wait, let's pause there, because this makes no sense on the
- 4:41surface. I know it really does.
- 4:43It's. Like a billionaire with a vault
- 4:44full of gold does not suddenly walk into town to take out PD
- 4:48loans and Max out their credit cards, right?
- 4:49Exactly, Yeah. A company that inherently does
- 4:52not lack money opened up both equity and debt financing
- 4:56channels simultaneously. And what exactly are senior
- 5:02notes? Why would they use those?
- 5:03So senior notes are basically high priority Ious.
- 5:07They're a type of debt that takes precedence over other
- 5:10debts if the company were to declare bankruptcy.
- 5:12OK. So they're safer for the lender?
- 5:14Right. They usually carry a lower
- 5:16interest rate because of that safety.
- 5:18As for why a cash rich company would issue them, it really
- 5:21comes down to liquidity and urgency.
- 5:24Urgency, huh? Yeah, a lot of that $240 billion
- 5:27cash pile is likely tied up in investments or held in overseas
- 5:30accounts. And bringing it back to the US
- 5:33would trigger massive tax penalties.
- 5:35Exactly. They needed liquid cash
- 5:37immediately without touching those existing reserves.
- 5:40So what does this tell us? They are borrowing aggressively
- 5:43to fund a $44.9 billion quarterly shopping spree.
- 5:49Right. It proves the absolute
- 5:51desperation driving the AI arms race right now.
- 5:54Desperation is a good word for it.
- 5:55This wasn't a signal that Google score search business was
- 5:58failing. Far from it.
- 6:00It was a glaring signal to the market that the scale of
- 6:03physical investment required. Right now, it's so astronomical
- 6:06that even a quarter trillion dollar cash buffer isn't enough.
- 6:10Wow. During the earnings call, the
- 6:12CFO explicitly referred to this as a supply constrained
- 6:16environment. Supply constrained.
- 6:18Yeah, they are desperate for compute power, and that
- 6:21desperation is what led forensic accountants to discover the
- 6:24hidden contract. The fire contract has been
- 6:27called in the financial press. Yes, this is the $30 billion
- 6:31secret. On June 5th, a regulatory filing
- 6:35from SpaceX revealed something wild.
- 6:38Truly wild. Google is renting 110000, a Gus
- 6:42CUS memory, an related network components directly from SpaceX,
- 6:48and the rent is $920 million per month.
- 6:51Nearly a billion dollars in rent every single month.
- 6:54It's unbelievable. The lease runs from October 2026
- 6:57to September 2029. If you calculate the total over
- 7:01that period, assuming the utilize the full capacity, it's
- 7:03roughly a $30 billion commitment.
- 7:05That's a huge commitment for a rental.
- 7:07Right, A Google spokesperson gave a very brief explanation,
- 7:11saying it was a transitional strategy to ensure they have the
- 7:13capacity to meet the demands of their AI agent platform Gemini
- 7:17Enterprise. Transitional strategy.
- 7:19Right. But wait, here's where it gets
- 7:20really interesting for you as a listener, because you have to
- 7:22think about who Google actually is.
- 7:24Exactly. Google literally invented the
- 7:26TPU chip, the Tensor Processing unit.
- 7:29They lay their own undersea Internet cables.
- 7:32They're infrastructure giants, yeah.
- 7:34They operate dozens of their own hyperscale data centers
- 7:37globally. Why on earth is a company that
- 7:40just spent $44.9 billion on its own infrastructure going to a
- 7:46rocket company to rent NVIDIA graphics cards?
- 7:49It seems backwards. Right.
- 7:51Do they just not have the chips? That is the exact assumption
- 7:54most people make, but it's fundamentally incorrect.
- 7:56Really. Yeah, the bottleneck in the AI
- 7:58industry right now is not the chips themselves.
- 8:00Google has plenty of chips. OK, So what is it?
- 8:03The genuinely scarce resources are the electricity and the
- 8:06physical facilities equipped to handle that specific intense
- 8:10power load. Wait I need to push back on that
- 8:13Google has TP US. Why do they specifically need to
- 8:16rent 110,000 NVIDIA GP US? Is there a?
- 8:19Difference. There is a big difference.
- 8:21While Google's TP US are incredibly efficient and custom
- 8:24built for running AI models once they are trained, NVIDIA GP US
- 8:28are currently the universal Swiss Army knife for AI
- 8:30development. Oh, I see.
- 8:32Almost all of the world's leading AI researchers and
- 8:35software ecosystems are deeply entrenched in Nvidia's
- 8:39programming architecture. Exactly CDA.
- 8:42If Google wants to attract top tier enterprise clients and
- 8:45offer the most flexible training environment, they have to offer
- 8:49massive NVIDIA GPU's Lesters. That makes sense.
- 8:52But as I mentioned, the real issue isn't acquiring the GPU's,
- 8:55it's plugging them in, right? Google could buy 100,000 GPU's
- 8:59tomorrow, but they couldn't get a new data center fully
- 9:02permitted, built and powered on by the local grid within a 12
- 9:06month time frame. But SpaceX could.
- 9:09Which is just wild to think about a rocket company beating
- 9:12Google at data center logistics. It really is.
- 9:15I read that SpaceX took an abandoned appliance factory and
- 9:18converted it into a massive AI training center in just 122
- 9:23days. How is that physically possible?
- 9:26It's an engineering marvel. Yeah, I mean according to Jensen
- 9:28Huang, the CEO of NVIDIA, SpaceX managed to install 100,000 GPU's
- 9:33in just 19 days. That is normally a four year
- 9:36process. It is an infrastructure miracle
- 9:39honestly and it comes down to engineering culture.
- 9:42How so? You have to remember SpaceX is
- 9:44an aerospace company used to vertical integration and brutal
- 9:48round the clock timelines. Right, they build rockets.
- 9:51Exactly. When a traditional tech company
- 9:54builds a data center, they hire outside contractors, wait for
- 9:57municipal permits and rely on standard supply chains.
- 10:00Lots of red tape. So much red tape.
- 10:03SpaceX didn't do that. They brought in their own
- 10:05engineers. They prefabricated the complex
- 10:07liquid cooling manifolds, the plumbing needed to keep these
- 10:11chips from literally melting off site and snap them together like
- 10:14Lego bricks. Wow, they treated a data center
- 10:17build like a rocket launch countdown.
- 10:19OK, but what about the electricity you mentioned?
- 10:21Power is the main bottleneck. We are seeing Google, Amazon,
- 10:25and Meta face massive delays because municipal power grids
- 10:30just don't have the capacity to hook up these GW facilities,
- 10:33right? So how does SpaceX get around
- 10:35the grid wait times? By completely bypassing the
- 10:38traditional grid. Wait, really?
- 10:40Yeah, SpaceX didn't wait in line for utility companies to spend
- 10:43three years upgrading local substations.
- 10:46They set up massive on site solar arrays and industrial gas
- 10:49generators directly at the factory site away.
- 10:52Yeah, they essentially generated their own power, building a 300
- 10:56MW power island in a matter of months.
- 10:59A power island that is incredible.
- 11:02O what SpaceX is actually selling to Google isn't compute
- 11:06power, they are selling time. In an AI arms race, having
- 11:11capacity available today rather than three years from now is
- 11:13worth billions. I see the logic, but that
- 11:16naturally forces a massive question.
- 11:18What's that? Building a 300 MW power island
- 11:21and racking 100,000 GPU's is incredibly expensive.
- 11:25Yeah. Why does Elon Musk have all this
- 11:27spare capacity just sitting around?
- 11:30Good question. He certainly didn't build the
- 11:31world's largest single AI training facility just to bail
- 11:34out Google. He absolutely didn't, according
- 11:36to a June 12th report from Bloomberg, citing people from
- 11:39familiar with the matter, SpaceX did not build this facility to
- 11:42become a landlord. OK, So what was the goal?
- 11:44The original plan was highly ambitious.
- 11:46They wanted to connect three distinct data centers across
- 11:49different locations together to create one giant unified cluster
- 11:54to train their own cutting edge model, Grok.
- 11:57But they could pull it off. Right.
- 11:58No, they couldn't. The Bloomberg report mentioned
- 12:00it failed because of severe latency issues.
- 12:03Break that down for us. Why is latency fatal to an AI
- 12:06cluster? Normal cloud servers are spread
- 12:09out all over the world, right? Yes they are, but training an AI
- 12:12model is vastly different from hosting 1,000,000 separate
- 12:15Netflix streams. So when you train a massive AI,
- 12:19the chips aren't working independently.
- 12:21They have to constantly share data, weights and parameters
- 12:24with each other billions of times a second.
- 12:26They have to act as a single synchronized brain.
- 12:30If one GPU is waiting even a few extra microseconds for data to
- 12:34arrive from a facility 10 miles down the road, the other 99,000,
- 12:381999 GPU's have to pause and wait for it.
- 12:41Microseconds. That's insane, yeah.
- 12:44The older network infrastructure simply couldn't handle the data
- 12:47transfer speeds required. They couldn't sync the three
- 12:50separate centers. Oh that makes perfect sense.
- 12:53It's like trying to conduct 100,000 piece orchestra, but a
- 12:56third of your musicians are sitting in a different concert
- 12:59hall being broadcast in via satellite.
- 13:02Even a half second delay turns the entire Symphony into
- 13:05complete noise. The music just stops.
- 13:08That is an excellent way to visualize it.
- 13:10The synchronization is everything.
- 13:12So when the latency issues proved insurmountable, the
- 13:15massive distributed cluster plan essentially collapsed.
- 13:20So essentially must try to build a sprawling hyper connected mega
- 13:24mansion. But when they turn the water on,
- 13:27the plumbing across the different wings just didn't
- 13:30connect right. The pressure was all off.
- 13:32Yeah, it was a mess. So instead.
- 13:33Of taking a total loss on the property, he immediately chopped
- 13:36the mansion up into fully furnished luxury apartments and
- 13:40now he is renting those apartments out to his biggest
- 13:42rivals for an absolute premium. That's wild.
- 13:45It's a brilliant pivot honestly, instead of sitting on a sump.
- 13:48Cost Musk turned that failed cluster into a highly lucrative
- 13:53infrastructure business. Yeah, I'll say.
- 13:55In May, they rented out a massive portion of the compute
- 13:58to another competitor for $1.25 billion a month.
- 14:03Then in June, they signed the deal with Google for 920 million
- 14:08a month. That is over $2.1 billion a
- 14:11month in rental income across just two contracts.
- 14:15That's staggering. Musk capitalized on the exact
- 14:18infrastructure bottleneck his competitors are trapped in.
- 14:21OK, so Google is paying a massive premium to rent Spacex's
- 14:25luxury apartment because their own mansion is still under
- 14:29construction and waiting for the power company, right?
- 14:31But when you look closer at the actual terms of this contract,
- 14:34the $920 million one, it gets even crazier.
- 14:38It doesn't read like a standard partnership at all.
- 14:40No, it really doesn't. The terms of the agreement
- 14:43reveal an emergency contingency plan with heavy distrust on both
- 14:46sides. Exactly, I want to highlight the
- 14:48escape hatches built into this thing.
- 14:50Let's do it first. If SpaceX doesn't deliver the
- 14:53GPU's by the end of September, basically after a one month
- 14:57grace period, Google has the right to immediately terminate
- 15:01the contract or demand a significantly discounted
- 15:04delivery rate. Right, covering in their bases.
- 15:06But the real kicker is what happens after December 31st,
- 15:102026. After that date, either party
- 15:13can just walk away with a mere 90 day notice.
- 15:16Just pack up and leave. Exactly.
- 15:17You can just pack up and leave. Why structure a multibillion
- 15:20dollar deal like that? Because structurally what Google
- 15:23is doing here is executing an options trade, not signing a
- 15:27traditional long term lease. An options trade.
- 15:29Yes, think of that $920 million a month as a premium paid to buy
- 15:33a call option on AI demand. A call option.
- 15:36Let's make sure that's clear for you listening.
- 15:37A call option gives you the right, but not the obligation,
- 15:41to buy an asset at a certain price.
- 15:43Correct. In this scenario, the asset is
- 15:45compute capacity. Got it?
- 15:47If the enterprise demand for AI models continues to explode,
- 15:50well, Google has machines ready and waiting to capture that
- 15:52revenue. They exercise their option.
- 15:55Makes sense. But if the AI bubble bursts, or
- 15:58if the consumer demand doesn't materialize the way they expect,
- 16:01Google can bail out with just 90 days notice.
- 16:04I see, so it's like paying a massive non refundable deposit
- 16:08to hold a luxury venue for a giant wedding that might get
- 16:12cancelled. You're paying a huge premium
- 16:15just for the right to use the space, guaranteeing you aren't
- 16:19left out in the cold if the wedding happens right?
- 16:21But if the whole thing falls apart, you just walk away from
- 16:23the deposit without being forced to actually buy the building.
- 16:26Precisely, they have successfully dumped all the
- 16:30heavy asset long term risk on to SpaceX.
- 16:34Think about the hardware itself. The servers have a depreciation
- 16:37life of five to six years. The contract is for 33 months.
- 16:41Even if Google stays for the whole term, when they leave,
- 16:44those 110,000 GPU's will still have half their accounting life
- 16:48left. And what to do with aging
- 16:50rapidly obsolescing assets becomes the landlord's problem,
- 16:53not Google's. Google bought themselves
- 16:56certainty and agility and a highly certain market.
- 16:59OK, so if Google was smart enough to build in this 90 day
- 17:02escape hatch and they successfully pushed the long
- 17:05term risk onto SpaceX, why did the market still punish them?
- 17:11Why did the stock drop 7% if this was such a clever defensive
- 17:15maneuver? Because the market isn't just
- 17:17looking at the SpaceX contract. They're looking at the big
- 17:20picture. Right, the 920 million a month
- 17:23is a drop in the bucket compared to the $44.9 billion in capital
- 17:29expenditure that Google is spending on its own permanent
- 17:32infrastructure, right? That $45 billion doesn't have an
- 17:36escape hatch, and it represents a massive depreciation time
- 17:40bomb. A depreciation time bomb.
- 17:42Let's dig into the accounting mechanics here, because this is
- 17:44the exact reason the stock crashed despite record profits,
- 17:47right? In Q2, Google spent 44.9 billion
- 17:51on equipment and facilities. Yeah, but on their income
- 17:54statement, they only recorded 7.1 billion in depreciation.
- 17:57Which is a ratio of over six to one.
- 18:00That's huge. To understand why this is
- 18:02terrifying to investors, we have to look at how accounting rules
- 18:06treat infrastructure. OK, when you buy a massive
- 18:09server farm or construct a data center, you don't take the
- 18:13financial hit for that entire cost on day one.
- 18:16You spread that cost out over the useful life of the asset,
- 18:20right? As you mentioned, servers
- 18:22depreciate over 5 to 6 years and the physical buildings
- 18:25depreciate over 25 to 40 years. So what does this all mean for
- 18:29you? Listening.
- 18:30Think of it like buying a fleet of incredibly expensive
- 18:33supercars on credit to start a luxury rental business.
- 18:36OK, good analogy. On your monthly budget, you're
- 18:39only recording the first month's insurance premium and a tiny
- 18:43fraction of the loan payment, right?
- 18:44Looking at that Ledger, it seems like you have a 34% profit
- 18:47margin today. You look like a genius.
- 18:49Yep, but those massive multi $1,000,000 bills for the cars
- 18:53themselves are just slowly creeping up the calendar.
- 18:56The money is spent, the debt is incurred, but the pain hasn't
- 18:59hit your profit and lost achievement yet.
- 19:01That is exactly how the math works out.
- 19:04The 34% operating margin Google reported is somewhat of an
- 19:08illusion. An illusion.
- 19:09Yeah, it's constructed entirely on the premise that the vast
- 19:12majority of their newly purchased assets haven't started
- 19:16depreciating yet. The costs haven't matured.
- 19:20Wow. And this isn't just a Google
- 19:21problem, it is an industry wide phenomenon.
- 19:23Really. Everyone is doing this, Yes.
- 19:26If you look at the top 4 hyperscalers over the past four
- 19:29quarters, they have a combined capital expenditure of roughly
- 19:33$433.9 billion, OK, but they have only recorded about 149
- 19:40billion in depreciation. That is nearly a $300 billion
- 19:44gap, a 3X difference. And that gap will not magically
- 19:47disappear. It is simply delayed.
- 19:49During the earnings call, management was very direct about
- 19:53this reality. They stated that this massive
- 19:55increase in infrastructure investment will inevitably press
- 19:58down on the income statement in the form of.
- 20:00Higher depreciation and energy costs in the near future.
- 20:03Free cash flow will continue to be under immense pressure.
- 20:07The market isn't pricing in Google's record earnings today.
- 20:11Right, they're looking ahead. The market is forward-looking.
- 20:14They are pricing in the absolute certainty that future profit
- 20:18margins are going to be crushed by the delayed depreciation of
- 20:22today's $45 billion spending sprees.
- 20:26And here's the really scary part, right?
- 20:28Those depreciation costs are going to hit the Ledger whether
- 20:32the AI revenue actually materializes or not.
- 20:34Exactly. The bill comes due.
- 20:36Those costs are locked in by the calendar, not by customer
- 20:39demand. The bill comes due regardless.
- 20:42So when the stock dropped 7%, the market wasn't reacting to
- 20:45Google renting chips from Elon Musk.
- 20:48The market was asking, you know, how much more money do you have
- 20:51to spend to secure the next wave of growth?
- 20:53And when, if ever, are you actually going to earn that
- 20:56money back? And the unsettling truth is that
- 20:59neither the earnings report nor the management team provided a
- 21:01clear answer to those two questions.
- 21:03Why not? Because they don't know.
- 21:05And that uncertainty is what caused the sell off.
- 21:08Which brings us to the ultimate take away from all of this.
- 21:10The big picture. What we are witnessing among the
- 21:13tech giants is not just standard corporate competition, it is a
- 21:17classic prisoner's dilemma on a multi $100 billion scale.
- 21:22The AI prisoner's dilemma explain how that dynamic is
- 21:25playing out right now because it feels like they are all.
- 21:28Well, in game theory, a prisoner's dilemma occurs when
- 21:31rational actors are forced into a mutually destructive outcome
- 21:35because they cannot afford to trust each other.
- 21:38Right now, every single major tech company, Google, Amazon,
- 21:42Meta, Microsoft knows that continuing to spend $45 billion
- 21:47a quarter is financially toxic. It's unsustainable.
- 21:50They know it will crush their future free cash flow, they know
- 21:53it will artificially inflate their depreciation cost for the
- 21:55next decade, and they know there is a very real possibility that
- 21:59they might not see a return on this investment.
- 22:00OK. But if they all know it's
- 22:02financially toxic, why wouldn't one of them just say, well, you
- 22:05know what, we're going to pause, we'll let the others burn their
- 22:08cash and we'll sit out for 1/4 to stabilize our margins.
- 22:11Because of the physical laws of compute scaling and AI.
- 22:14OK, tell me more. In traditional software, a
- 22:17smaller team with better code could be a tech giant.
- 22:21AI doesn't work like that right now.
- 22:23AI capability scales linearly with compute power.
- 22:26Meaning. Meta trains a model on 100,000
- 22:30GPUs and Google decides to save money and only train on 50,000
- 22:34GPUs. Meta's model will be objectively
- 22:37smarter, faster, and more capable.
- 22:40Oh wow, there is no second place in foundational AI capability.
- 22:44Oh wow, so you can't out algorithm A sheer lack of
- 22:47compute. Exactly.
- 22:48The moment a company steps off the gas pedal, even for a single
- 22:51quarter, they risk losing everything.
- 22:54Everything. They immediately lose their
- 22:56enterprise clients who demand the hardest models.
- 22:58They lose their models capabilities and they hand the
- 23:00future of the Internet over to their rivals.
- 23:03Ma'am, they are locked in. They can complain about the high
- 23:05capital expenditures on earnings calls, but they cannot stop.
- 23:09It is no longer about greed or capturing new markets.
- 23:11It is purely about survival. Which is such a dark way to view
- 23:15a $45 billion investment. It really is.
- 23:17It's not a victory March. It's a hostage situation.
- 23:20The money isn't the bottleneck anymore.
- 23:22The electricity is a bottleneck. The physics of latency is the
- 23:25bottleneck, but the hardest constraint of all is this
- 23:29psychological trap. Everyone knows continuing to
- 23:32throw money into this might not yield a return, but stopping
- 23:36guarantees failure. And that dynamic fundamentally
- 23:39changes how investors have to evaluate these companies moving
- 23:43forward. We are about to witness a brutal
- 23:47filtering process in the market. A filtering process.
- 23:50Yeah, the ultimate winner of this arms race will be the one
- 23:53who can successfully convert these massive asset investments
- 23:57into future pricing power and scale advantages.
- 24:00Right. And the losers, but the.
- 24:01Losers. The losers are going to find out
- 24:03that they just spent hundreds of billions of dollars building
- 24:05incredibly expensive infrastructure for someone.
- 24:07Else man. So as you watch this tech arms
- 24:10race unfold and you see these mind boggling numbers cross the
- 24:15news ticker, ask yourself this. If the ultimate winner of the AI
- 24:19wars spends hundreds of billions of dollars, ruins their free
- 24:22cash flow, and navigates all these physical bottlenecks just
- 24:25to become an overpriced utility provider for someone else's
- 24:28software, are they actually winning or just surviving?
- 24:32That is the multibillion dollar question that the market is
- 24:36desperately trying to answer. Thank you for joining us on this
- 24:38deep dive into the hidden layers of the tech economy.
- 24:41Keep questioning those headlines, keep looking closely
- 24:44at the cash flow statements, and we'll see you next time.