Latest / Deep Values: Value Investing Deep Dives / Big Tech‘s Hidden AI Cash Burn Exposed!
Transcript
- 0:00Meta stock is up like nearly 30%.
- 0:02Yeah, and Microsoft and Amazon are up 20%.
- 0:04Right. So I mean, if you are just
- 0:06glancing at your investment portfolio or, you know, reading
- 0:09the headline numbers, Big Tech appears to be in this
- 0:12untouchable golden age of unlimited.
- 0:14Growth. It looks completely bulletproof
- 0:16from the outside. Exactly.
- 0:18But today we are completely ignoring those shiny top line
- 0:22revenue numbers because the data hiding just beneath the surface
- 0:25is pointing to a massive financial hemorrhage.
- 0:29It's actually it's pretty shocking when you dig into it.
- 0:31It really is. I mean, when you look at free
- 0:33cash flow, meta actually plummeted 91% recently.
- 0:36Oh wow. 91%. And Amazon and Google went
- 0:40negative. So today we are doing a deep
- 0:43dive into why companies that literally print money are
- 0:46suddenly bleeding cash. And for this we are looking at a
- 0:49really brilliant recent interview from the Prof G
- 0:52Markets channel. Right.
- 0:54Featuring the legendary valuation expert Professor
- 0:56Aswath Domadaran, our mission today is to unpack Big Tech's
- 1:00astronomical AI spending spree. You know, look at the accounting
- 1:04tricks masking the costs and figure out if these trillion
- 1:07dollar empires actually have a real business model.
- 1:10Or if they're just, you know, throwing hundreds of billions of
- 1:13dollars into the dark hoping something sticks.
- 1:15Exactly. But before we pull back the
- 1:17curtain on these numbers, a quick note for you to access
- 1:20more advanced researches, applications, and our latest
- 1:23Deep Values agents. Welcome to explore and join Deep
- 1:26Values dot AI track total market temperature.
- 1:30Welcome to search and download iOS app Buffett Indicator from
- 1:33Apple App Store or Bubble Watch from Google Play.
- 1:36Disclaimer All analysis is based on publicly available sources
- 1:40and may be subject to revisions or differing interpretations.
- 1:43All contents from this channel is for research and
- 1:45entertainment purposes only and does not constitute any
- 1:48investment or financial advice or service.
- 1:51Conduct your own due diligence before making investment or
- 1:53financial decisions. All right, so let's get into the
- 1:55mechanics of this, because the raw data Damodaran highlights is
- 1:59actively spooking market analysts.
- 2:02The ones who actually bother to look past the hype anyway.
- 2:05Right. We are looking at what could
- 2:08arguably be the biggest fundamental shift in business
- 2:11mechanics sincethe.com bubble. Which is a huge statement.
- 2:15It is because think about it, Five years ago, if you asked an
- 2:18analyst what the invested capital was for these tech
- 2:21giants, they wouldn't even. Cared because it was basically
- 2:24an irrelevant metric, right? Exactly.
- 2:27These companies were the ultimate cash cows.
- 2:29I mean, they built a software platform or a search engine or a
- 2:32social network. And scaling that up to a billion
- 2:35users required relatively little physical infrastructure.
- 2:38Right. They were generating 7080,
- 2:40sometimes even 90% returns on almost zero additional
- 2:43investment. Wow.
- 2:44But today that is totally morphed.
- 2:47They are no longer asset light software companies.
- 2:49No, no, they've become, they're basically capital intensive
- 2:53manufacturing companies now they are building sprawling physical
- 2:58AI factories. Which completely changes the
- 3:00financial math. Right.
- 3:02So I want to make sure I am visualizing the scale of this
- 3:04pivot because to me it sounds like a software company waking
- 3:08up 1 morning and just deciding they want to become a commercial
- 3:11airline. That is actually a really great
- 3:13way to put it. I mean, you used to just write
- 3:15code, put in a comfy office, hit deploy and watch the cash roll
- 3:20in, you know? Yeah.
- 3:21Pure profit margin. But now you've suddenly decided
- 3:25to buy $150 billion worth of physical jet engines.
- 3:29Or, well, in this case, immense data centers and hundreds of
- 3:33thousands of NVIDIA chips. Right, the heavy machinery of
- 3:36the AI. Boom, and suddenly you have to
- 3:38prove to your investors that you can actually earn more than the
- 3:41massive cost of that heavy machinery.
- 3:44And that airline analogy hits on the exact burden these companies
- 3:47have just taken on. When you become an airline or a
- 3:49heavy manufacturer, yeah, you are suddenly dealing with
- 3:52depreciation, right, Right. You're dealing with physical
- 3:55maintenance, supply chain bottlenecks, and the sheer
- 3:58massive cost of capital. Which brings us to Dammit
- 4:02Aaron's major warning here. Yes, he focuses heavily on a
- 4:05metric called ROIC, Return on Invested Capital.
- 4:08OK, so unpack that for us. So a few years ago, analyzing
- 4:12the cost of capital for big tech was literally a waste of time
- 4:16because their margins were universally sky high.
- 4:18Right. They didn't really have a huge
- 4:19cost of capital to worry about. Exactly, but today ROIC is the
- 4:25singular metric these companies will live or die by.
- 4:28Because they are pouring unheard of amounts of money into
- 4:31catbacks. Yeah, capital expenditures,
- 4:33which is the money they are spending on buying and
- 4:35maintaining these physical assets.
- 4:37The data centers, the cooling systems, the chips.
- 4:40Right. And if they cannot deliver end
- 4:42user earnings that justify this massive capital outlay, the
- 4:47markets are going to be incredibly punitive.
- 4:48But I have to assume this isn't just like collective stupidity,
- 4:51right? I mean, these are some of the
- 4:52smartest, most data-driven CEOs on earth.
- 4:55You would think so, yeah. So are they seeing a clear path
- 4:58to profitability that the rest of us aren't?
- 5:00Or is this just pure panic driving this, you know, 150 to
- 5:04$200 billion in AI capital expenditures?
- 5:07Interestingly, de Madaron actually leans heavily toward
- 5:10panic. Wait, really?
- 5:12Panic. Yeah, he describes this
- 5:14Emperor's New Clothes dynamic playing out in Silicon Valley
- 5:18right now. You look at the hyperscalers,
- 5:20these massive cloud infrastructure providers like
- 5:23Amazon Web Services, Google Cloud and Microsoft Azure.
- 5:27Right, the big Three. They have not actually
- 5:30enunciated a clear AI business model.
- 5:33They really don't know if they are aiming to build high margin
- 5:36premium products for niche enterprise clients or.
- 5:40If they are trying to build a low margin, massive scale
- 5:43product for everyday consumers. Exactly.
- 5:45The core strategy is completely ambiguous, but the spending is
- 5:49absolute. Because they're terrified of
- 5:51being left behind by history. That's exactly it.
- 5:53The fear of being the CEO who missed the artificial
- 5:56intelligence revolution is what's driving the spending.
- 5:59It's FOMO. Pure $100 billion FOMO.
- 6:02Yeah, it's not based on a clear mathematical path to
- 6:05profitability at all. But wait, is it actually panic
- 6:07if literally everyone is doing it?
- 6:08I mean in tech if you miss major platform shift you die.
- 6:11That's the counter argument, right?
- 6:12Look at Blockbuster ignoring streaming or Nokia ignoring the
- 6:16smartphone. You don't want to be that guy.
- 6:18Nobody wants to be Nokia. Exactly.
- 6:20But in this massive herd of tech giants just blindly throwing
- 6:25hundreds of billions at the wall, there is one glaring
- 6:28outlier. Oh yeah, Apple.
- 6:30Apple While everyone else is spending up to $200 billion,
- 6:33Apple's capital expenditure on AI is sitting around like 11
- 6:38billion. It's a tiny fraction of what the
- 6:40others are doing. Right, so I look at that and
- 6:42wonder, is Tim Cook risking the entire Apple Empire just to save
- 6:46a few billion in the short term? I mean, isn't he missing the
- 6:49biggest gold rush in human history?
- 6:51Well, Damodaran argues, there is actually immense strategic value
- 6:55in Apple's highly cautious approach.
- 6:57They are waiting out the uncertainty.
- 6:59Because jumping in too early can be just as bad.
- 7:01Yes, missing a platform shift is dangerous, absolutely.
- 7:05But building the wrong infrastructure is equally fatal.
- 7:08How so? Well, if you don't know whether
- 7:10the future of AI is going to require high scale, low cost
- 7:13server farms or highly specialized premium product data
- 7:17centers. Then building the wrong factory
- 7:19today is a multibillion dollar mistake.
- 7:21Deficely, you know, it reminds me of the railroad boom in the
- 7:241800s. Oh, that's a good comparison.
- 7:27Right, because you don't want to build thousands of miles of
- 7:30narrow gauge tracks for a train that only goes 25 miles an hour
- 7:35only to realize three years later that the industry standard
- 7:38is high speed wide gauge rail. Because then your entire
- 7:42physical infrastructure is completely useless.
- 7:44Exactly, you just burned billions on the wrong track.
- 7:48And Apple is essentially saying they don't know enough about how
- 7:51this space will monetize yet to justify betting the company's
- 7:54balance sheet on it. Which is honestly pretty
- 7:56refreshing, though I know analysts have punished Apple a
- 7:58bit for this caution, arguing they aren't, you know,
- 8:01innovating fast enough. They have, but a more ambitious,
- 8:04reckless CEO could have easily dragged Apple into 10s of
- 8:08billions in unnecessary debt for physical data centers that might
- 8:12not even fit the eventual market demand.
- 8:14OK, so the hyperscalers are burning immense amounts of cash
- 8:17on physical hardware, but this is what trips me up.
- 8:21Their stock prices are still soaring.
- 8:23Yeah. It seems counterintuitive.
- 8:24The math just doesn't add up for me unless they are masking the
- 8:28bleeding somehow. I mean, why did Amazon and
- 8:31Google's net income absolutely explode in their recent earnings
- 8:34reports if their free cash flow is tanking?
- 8:37All right, so this is where we get into the distortion of
- 8:39modern accounting. Oh, boy.
- 8:40Yeah, the net income numbers you are seeing on the headlines,
- 8:44they are heavily compromised. Compromised how?
- 8:47Let's look at that recent explosion in profits for Amazon.
- 8:5185% of its net income jump was attributable to unrealized paper
- 8:56gains. Wait 85%?
- 8:58Yes, and for Google it was 87%. Just unrealized paper gains,
- 9:02meaning they haven't actually made any actual cash from a
- 9:05customer. That's right, zero actual cash
- 9:08from end users for those games. OK, how does that even work?
- 9:12Well, these companies have taken massive stakes in private AI
- 9:15startups. So Amazon and Google are heavily
- 9:18invested in Anthropic. Microsoft is heavily invested in
- 9:21Open AI. Google also has stakes in
- 9:24SpaceX. Right, we've seen those
- 9:25headlines. And as the theoretical private
- 9:29market valuations of these startups go, U Amazon and Google
- 9:32record those gains on their own corporate income statement.
- 9:35Oh wow, so it artificially bumps up their net income to look like
- 9:39a wildly profitable quarter even though no actual cash has
- 9:44entered their bank accounts from selling a product?
- 9:46Exactly. It's an accounting illusion.
- 9:48Let me test an analogy on you, just to see if I am grasping the
- 9:51sheer absurdity of this. OK, let's say I hand you $100 to
- 9:55buy my old bicycle. OK then I take that exact same
- 9:59$100 sale and record it in my Ledger as absolute proof that my
- 10:03bicycle business is booming and generating massive organic
- 10:05demand. Right, but I literally gave you
- 10:08the money to buy my thing. It's a closed intra company
- 10:11circular loop. That captures the danger of this
- 10:14perfectly. Big Tech gives billions of
- 10:17dollars in investment capital to companies like Anthropic and
- 10:20Open AI, right? Those startups which require
- 10:23massive computing power to train their models, then turn around
- 10:26and use that exact same investment money to buy cloud
- 10:29computing power and data center access from big tech.
- 10:33So they are just buying server space from their own investors.
- 10:35Yes. So Big Tech gets to record that
- 10:38transaction as cloud revenue, and simultaneously they record
- 10:42the rising paper valuation of the startup as a gain in their
- 10:46own net income. That is wild.
- 10:48It is literally a circular loop of money that creates the
- 10:52illusion of explosive organic growth from end users.
- 10:56But wait, how is that legal? It does sound sketchy.
- 10:58Right, It sounds dangerously close to money laundering or
- 11:01accounting fraud. Well, under standard Generally
- 11:05Accepted Accounting Principles, or GAP, it's perfectly legal
- 11:08because these are technically separate corporate entities.
- 11:11Ah, because big Tech is just a minority investor, not the
- 11:14outright owner. Right, so when Anthropic buys
- 11:17server space from Amazon Web Services, Amazon is legally
- 11:20required to book that as standard revenue.
- 11:22But as an investor trying to analyze the health of a company,
- 11:25relying on that legality totally blinds you to the actual
- 11:28financial mechanics at play. Exactly.
- 11:31Damodarin explicitly warns that net income is now a quote UN
- 11:34quote deadly number for these companies.
- 11:36Because it completely breaks traditional valuation metrics.
- 11:40I mean, if an average retail investor logs into their
- 11:43brokerage account and looks at the PE ratio, the price to
- 11:47earnings ratio of these tech giants?
- 11:49Looks fantastic right now. Yeah, it is artificially dropped
- 11:52below 20. APE below 20 for a massive tech
- 11:55company. Looks like an absolute steal it.
- 11:57Looks incredibly cheap. But that E the earnings part of
- 12:01the ratio is being massively inflated by this circular
- 12:05startup money. Which means investors have to do
- 12:08the hard work now. You cannot just look at the
- 12:10shortcut metrics or the headline PE ratio anymore.
- 12:13You have to dig deeper. You have to strip out these
- 12:15paper gains from the AI startups and look strictly at the
- 12:19operating margins to see the truth of what the core business
- 12:21is actually doing and. What the core businesses are
- 12:23actually doing is taking on massive leverage.
- 12:26Massive. Which brings us to the debt.
- 12:28The off book debt is just staggering.
- 12:30It really is. The source notes that Meta has
- 12:32$140 billion of debt on the books, but they have $420
- 12:38billion in off book shadow debt. And that debt is largely tied to
- 12:43these massive data centers they are building. 420 billion off
- 12:47the books. Yeah, this adds a whole new
- 12:49layer of opacity to the financial health of Silicon
- 12:51Valley. Damon Darren points out that
- 12:54this might be non recourse debt tied specifically to the data
- 12:58centers themselves. OK, what does non recourse mean
- 13:01in this context? It basically means they set up a
- 13:04separate financial entity to build the data center.
- 13:07So if that specific data center fails to be profitable, the
- 13:10creditors can take the physical servers, but they can't
- 13:13necessarily come after meta score assets like Instagram or
- 13:17WhatsApp. Oh I see, so it isolates the
- 13:19risk a bit. Exactly.
- 13:20But regardless of the accounting gymnastics used to keep it off
- 13:23the main balance sheet, it proves that the level of capital
- 13:26leverage happening behind the scenes is astronomical compared
- 13:29to the tech industry of just five years ago.
- 13:31OK, so let's summarize where we are.
- 13:33We have companies acting like heavy manufacturers spending
- 13:36hundreds of billions out of FOMO and hiding the reality of the
- 13:39cash burn with circular accounting and massive off book
- 13:43debt. That's the reality, yes.
- 13:45So what happens when the actual market demand gets tested?
- 13:48Well, we are already starting to see emerging cracks in the AI
- 13:51demand narrative. Really.
- 13:53Where? A massive red flag is appearing
- 13:55right now in the sub leasing market.
- 13:57A sub leasing market, yeah. Elon Musk with X and SpaceX and
- 14:02Mark Zuckerberg at Meta have both recently started leasing
- 14:06out their compute power and data center space to other companies.
- 14:09Including potential competitors. Including potential competitors.
- 14:12But wait, earlier in the year, the entire narrative in the
- 14:15financial press was all about compute scarcity.
- 14:18You know, everyone was fighting for chips and server space.
- 14:21Exactly. So if you truly believed that AI
- 14:24was going to be an immediate $28 trillion market for your own
- 14:28products, you would never lease your weapons to your
- 14:30competitors. You would hoard every single
- 14:33ounce of compute power you had. And that's why Damodaran deduces
- 14:37that this subleasing trend is a massive tell.
- 14:40It's a huge giveaway you. Only rent out your
- 14:42infrastructure if you are making more reliable money acting as a
- 14:45landlord than you are building your own AI products.
- 14:49It signals that these CEO's just don't have as much faith in
- 14:53their own immediate end user demand as they claim in their
- 14:56big earnings calls. So we might be looking at a
- 14:58potential demand crisis that is currently disguised as a supply
- 15:02premium. Exactly.
- 15:03And then there is the geopolitical factor, which
- 15:06throws absolute gasoline on this whole demand problem.
- 15:09Right, the global competition. The source brings up Chinese
- 15:11LLM's, you know, large language models, the underlying engines
- 15:15powering these AI chat bots. Right, like Open AI is GPT,
- 15:19yeah. And a company called DeepSeek
- 15:21was mentioned, noting that Chinese models have gone from an
- 15:238% market share to over 50% share recently.
- 15:27It's a huge jump. But aren't the US models from
- 15:30Open AI and Google just vastly superior in quality?
- 15:33Like why would market share shift so drastically to Chinese
- 15:37models? Well, quality at the absolute
- 15:40cutting edge might still favor the US, sure, but the shift is
- 15:43really about commoditization. Commoditization.
- 15:46Yeah, China is positioning itself to do to artificial
- 15:49intelligence exactly what it did to physical manufacturing steel
- 15:53and solar panels. Oh I they look at a market,
- 15:56figure out how to produce a quote UN quote good enough
- 15:58version at a massive scale and turn it into a low margin single
- 16:03digit return commodity. Like they did with solar panels,
- 16:06they just flood the market with cheap supply.
- 16:08Exactly. Yeah, which drives the price of
- 16:10the product down for everyone globally.
- 16:12So instead of AI being this super premium, highly expensive
- 16:16software suite that like Fortune 500 companies, pay a massive
- 16:20premium for. It becomes as cheap and common
- 16:23as a basic calculator on your smartphone.
- 16:25That completely, completely wrecks the business model for
- 16:27big tech. The market is likely going to
- 16:29bifurcate. There might be a premium tier
- 16:32where a massive consulting firm pays top dollar for a highly
- 16:36specialized cutting edge AI agent, right?
- 16:39But the vast majority of the global market will be served by
- 16:42low margin mass scale models. And if AI becomes a mass market,
- 16:45low margin business, the hundreds of billions of dollars
- 16:48that big tech just spent building ultra premium cutting
- 16:52edge data center, it will. Be nearly impossible to recoup.
- 16:55They basically built a factory designed to manufacture luxury
- 16:59sports cars, but they are waking up to a market that only wants a
- 17:04$10 bicycle. That's exactly the risk they are
- 17:07facing. Listening to all of this, I mean
- 17:09the parallels to the 1999 and 2000.com bubble are just
- 17:14impossible to ignore. The echoes are definitely there.
- 17:16Back then everyone knew the Internet was the future, so they
- 17:19blindly threw billions at literally any company building
- 17:23fiber optic cables or Internet infrastructure, completely
- 17:26regardless of the actual business model.
- 17:28Right, it was all about building for the future.
- 17:30But they built the infrastructure for a future that
- 17:33took another 10 years to actually monetize.
- 17:35It is a profound historical parallel, and it's important to
- 17:39rememberhowthe.com bubble actually popped.
- 17:41It wasn't just overnight, was it?
- 17:42No, it wasn't one single catalyst.
- 17:45It wasn't one Tuesday where everyone just woke up and
- 17:47realized it was over. It was a staggered slow burn
- 17:51collapse. How did it start?
- 17:52Well, individual companies started missing their earnings
- 17:55estimates by small margins. Then telecommunications
- 17:59companies realized nobody was buying the broadband they just
- 18:02laid in the ground. Right, there just weren't enough
- 18:04users yet. Small write offs happened, so
- 18:06David Aaron suggests we should be watching for a massive
- 18:11unignorable CapEx write off today.
- 18:13Like the moment a major tech giant steps forward and admits
- 18:17they messed up. Exactly.
- 18:19If someone says we made a mistake on demand forecasting,
- 18:22we are writing off $10 billion in data center investments that
- 18:27could be the catalyst that finally breaks the momentum
- 18:30effect we are seeing in the stock market right now.
- 18:32Because right now, the momentum is just overpowering the reality
- 18:35of the balance sheets completely.
- 18:37Every time there is a slight dip in these tech stocks, money from
- 18:40the sidelines just pours right back in because retail and
- 18:43institutional investors like are just terrified of missing the AI
- 18:46wave. But eventually, financial
- 18:48gravity takes over. When the core fundamentals
- 18:51break, the catalyst effects will overwhelm the momentum, just
- 18:54like they did in 2000. So what does this all mean for
- 18:58you, the listener? I mean, if you hold index funds
- 19:01or you have your retirement tied up in the S&P 500, you need to
- 19:05realize that the fundamental nature of your investments has
- 19:08changed beneath your feet. Yes, it is a totally different
- 19:11landscape now. You are no longer investing in
- 19:14those high margin asset light software Wizards of the 20 tens
- 19:18who could scale to a billion users with a few lines of code.
- 19:22No, you are heavily invested in highly leveraged capital heavy
- 19:26infrastructure builders. Builders who are currently
- 19:29hoping a profitable business model magically appears out of
- 19:32the ether to justify their historic spending spree.
- 19:35That is the sober reality of the market today, and investors
- 19:38really need to look at operating margins, not just a headline net
- 19:42income. Absolutely.
- 19:43But you know, I want to leave you with a final lingering
- 19:45thought to Mull over, building on this idea of AI becoming a
- 19:49cheap commodity. Oh, I'm ready.
- 19:51Lay it on us. We've spent this entire deep
- 19:54dive worrying about the tech giants, right?
- 19:56Analyzing their return on investment?
- 19:58Stressing over their physical data center.
- 20:00Right, the corporate side of it. But let's imagine the
- 20:02commoditization strategy actually succeeds.
- 20:05Let's imagine AI intelligence becomes a low margin commodity,
- 20:09essentially as cheap and ubiquitous as electricity.
- 20:13If that happens, what happens to the concept of the knowledge
- 20:16worker? Oh wow.
- 20:17If answers, complex analysis, and basic coding cost absolutely
- 20:21nothing to generate, the real crisis won't be on big tech's
- 20:24balance sheet. It'll.
- 20:25Be on ours. Exactly.
- 20:27The real crisis will be figuring out the economic value of human
- 20:31expertise in a world where intelligence is free.
- 20:34Man, that is a wild and slightly terrifying thought to end on,
- 20:38but it's so true. I mean, if the foundation of the
- 20:40economic skyscraper is shifting that drastically, maybe it's
- 20:44time we look at what kind of ground we are standing on in our
- 20:47own careers. Definitely something to think
- 20:48about. Keep questioning the numbers,
- 20:50look past the headlines, and stay curious.
- 20:52We will catch you on the next deep dive.