Latest / Deep Values: Value Investing Deep Dives / Tencent Record Profits Fuel AI Arms Race
Transcript
- 0:00Tencent just posted record-breaking profit margins
- 0:03when we were talking about 1/4 where they extracted massive
- 0:05profits from a captive audience of like 1.4 billion users.
- 0:09Yeah, it's a. Staggering number of people.
- 0:11Right. So why did their free cash flow
- 0:13suddenly plunge by 13 billion into the red?
- 0:16Well, welcome to the artificial intelligence arms race.
- 0:19It is. It's really the ultimate
- 0:21corporate paradox. On the surface you have this
- 0:24unstoppable monopolistic cash machine, but underneath they are
- 0:28just hemorrhaging capital to secure their survival in the
- 0:31next computing paradigm and. Untangling that exact paradox is
- 0:34our mission. For today's deep Dive, we are
- 0:36looking at a brand new comprehensive financial research
- 0:39report on 10 set holdings. That's ticker 0700 in Hong Kong
- 0:44or TCEHY over the counter. This is hot off their Q2 2026
- 0:48earnings released just yesterday, August 13th.
- 0:51We're going to breakdown how their AI is actually driving
- 0:54unprecedented ad revenue. Figure out where all that cash
- 0:57suddenly vanishing to run the stock through a brutal value
- 1:00investing stress test. And finally, calculate a strict
- 1:04discounted cash flow valuation to see what this behemoth is
- 1:07actually worse right now. It's a fascinating look at what
- 1:10happens when a legacy software empire is basically forced to
- 1:14become a massive buyer of physical hardware.
- 1:16I mean almost overnight. Totally.
- 1:18Now, before we jump into those underlying mechanics, a quick
- 1:20note for you to access more Advanced Research companion for
- 1:24modern intelligent investors and leverage our latest D Values
- 1:27agents. Welcome to download D Values APP
- 1:30from App Store to track stock market temperature.
- 1:34Welcome to download iOS app Buffett Indicator from Apple App
- 1:38Store or Bubble Watch from Google Play.
- 1:41Disclaimer All analysis is based on publicly available sources
- 1:45and may be subject to revisions or differing interpretations.
- 1:48All contents from this channel is for research and
- 1:50entertainment purposes only and does not constitute any
- 1:53investment or financial advice or service.
- 1:55Conduct your own due diligence before making investment
- 1:58financial decisions. All right, let's look at the
- 2:01sheer scale of the operation before we get into the cash
- 2:03flow. Shocker.
- 2:04Yeah, we have to set the baseline first because the top
- 2:06line numbers for Q2 are just staggering.
- 2:08They really are. Total revenue hit 204.8 billion
- 2:12RMB. That is a solid 11% year over
- 2:15year increase. And what is critical for you to
- 2:17understand is that this wasn't, you know, a flu driven by 1
- 2:20viral game, right? It was synchronized.
- 2:22Exactly synchronized structural growth across their three main
- 2:27pillars. So let's break those pillars
- 2:28down so we can see where the money's actually coming from.
- 2:31The biggest chunk, about 49% of revenue, is value added
- 2:35services. Which is their historical
- 2:36engine, right? Yeah, gaming and social
- 2:38networks, it grew 8% year over year.
- 2:41But gaming in China is supposed to be this this entirely
- 2:46saturated market. How do you squeeze an 8%
- 2:49increase out of a mature demographic?
- 2:52By fundamentally restructuring how they monetize, I mean they
- 2:54saw a 17% resurgence specifically in domestic games.
- 2:58They've optimized the monetization of these massive
- 3:01Evergreen titles, and they paired that with highly
- 3:04successful new releases like Delta Force, Valor and PC, and
- 3:08mobile and Roko Kingdom. World, So they're just
- 3:11exceptionally good at keeping players inside the ecosystem.
- 3:14Very much so. So the legacy gaming machine is
- 3:16humming, even with some currency headwinds dragging down the
- 3:19international side. Then you have the second pillar,
- 3:21fintech and business services, sitting at about 31% of revenue,
- 3:25growing 9%. Yeah, this is essentially the
- 3:27infrastructure layer of Chinese commerce.
- 3:30You have WeChat Pay processing, everyday commercial
- 3:32transactions, actions, their wealth management arms and
- 3:35enterprise cloud services. And crucially, those cloud
- 3:38services have finally moved past the whole cash burn phase.
- 3:41Right. Yes, they've returned to
- 3:42scalable profitability, which is a huge deal.
- 3:44Which brings us to the third pillar, marketing services.
- 3:48This used to just be called online advertising.
- 3:50It makes up 19% of total revenue.
- 3:52And this is where I need you to explain something to me.
- 3:54This division surge 22% year over year to 43.6 billion RMB,
- 4:00but Wikeson and WeChat already have an astonishing 1.439
- 4:04billion monthly active users. Right, the user base is
- 4:08basically tapped out. It's practically the entire
- 4:10connected population of China. The user base is fully
- 4:12saturated, so how on earth do you spike ad revenue by 22%
- 4:17without just bombarding your users with double the amount of
- 4:20ads? That is the multibillion dollar
- 4:23question and the answer is the first major commercial
- 4:27breakthrough of their AI investments.
- 4:29They rolled out an upgraded AI driven ad recommendation model
- 4:33and this automated campaign management solution called AM
- 4:37Plus. OK.
- 4:38So they didn't just increase the ad load.
- 4:40No, not at all. The number of ads you see on
- 4:43WeChat actually remained below the industry average.
- 4:46Instead, they made the ads exponentially smarter.
- 4:48I see. So the artificial intelligence
- 4:51is analyzing user behavior so accurately that it only shows
- 4:54you what you're actually likely to click.
- 4:56On and that completely changes the unit economics of the
- 4:59business. I mean when your ad targeting is
- 5:01that precise click through rates soar.
- 5:03And advertisers notice that immediately.
- 5:04Of course, when click through rates soar, advertisers realize
- 5:07they're getting a much better return on investment, so they're
- 5:10willing to pay a higher premium per thousand impressions.
- 5:13Tencent is literally absorbing marketing budgets away from
- 5:17e-commerce rivals simply because their AI model converts better.
- 5:21That perfectly explains the incredible operating leverage.
- 5:24They are making more money on the exact same amount of digital
- 5:27real estate and that shows up in the profit margins.
- 5:31I mean, their gross margin hit 57.8% this quarter.
- 5:34Which is huge. Just three years ago that margin
- 5:36was 47.5%. A 10 point expansion at this
- 5:39scale is almost unheard of for a mature monopoly.
- 5:43It proves their transition away from, you know, lower margin
- 5:46operations into high margin AI. Targeted advertising and
- 5:50enterprise software is working. Florida State.
- 5:52OK, so let's summarize the narrative up to this point.
- 5:54You have an Apex digital monopoly, they have 1.4 billion
- 5:59captive users, their margins are expanding beautifully, and their
- 6:02new AI ad targeting is basically printing money.
- 6:05Literally printing it. In fact, they generated roughly
- 6:0768.4 billion RMB in non IFRS net profit.
- 6:11And just to define that for a second for you listening, non
- 6:14IFRS net profit is essentially their core operating cash after
- 6:18stripping out weird accounting distortions like stock based
- 6:21compensation. So we are looking at a company
- 6:23generating mountains of cash, which makes the next part of
- 6:26this report absolutely terrifying.
- 6:28How does a company printing that much money suddenly post a
- 6:32negative free cash flow of 13.8 billion RMB in the exact same
- 6:37quarter? It is a phenomenal plot twist in
- 6:39the financial statement. Operating capital expenditure,
- 6:42the money they spend on physical assets skyrocketed by 176 to
- 6:46190% year over year. Yeah, they blew through roughly
- 6:5052.8 billion RMB be in a single three month period.
- 6:54Hold on, you can't just spend 52 billion RMB on like office
- 6:58supplies. Where did the money physically
- 7:00go? Servers, specifically massive
- 7:03front loaded prepayments for compute infrastructure.
- 7:06They are buying application specific integrated circuits
- 7:09known as ASICS and the most advanced GPU's they can get
- 7:11their hands on. Yeah, and management was
- 7:13entirely transparent about this on the earnings call.
- 7:15This is an operational failure where the core business is
- 7:18losing money. If you exclude these massive
- 7:20upfront payments for hardware, they're normal normalized free
- 7:22cash flow would have been a highly +37.6 billion RMB.
- 7:27So it's basically a mechanical distortion on the balance sheet.
- 7:30They essentially just pay the rent for the next 5 years to the
- 7:32AI revolution upfront in one lump sum.
- 7:35That's a really great way to look at it.
- 7:37And they have to pay that rent to train the Hanyuan large
- 7:39model. This is their proprietary large
- 7:42language model and it is an absolute beast.
- 7:44It's a 389 billion parameter model built on what's called a
- 7:49mixture of experts or Moe architecture.
- 7:52Let's unpack Moe for a second because it's a huge buzzword in
- 7:55AI right now, but the mechanics are crucial.
- 7:58Why does a mixture of experts architecture matter when it
- 8:01comes to the actual hardware they are buying?
- 8:03So think of a traditional AI model like a single massive
- 8:07brain. Every time you ask it a
- 8:09question, even a simple one, the entire brain has to fire up.
- 8:11Which uses an incredible amount of computing power.
- 8:14Exactly a mixture of experts. Architecture is more like a
- 8:17massive hospital. You don't need the entire board
- 8:19of directors and 50 surgeons to look at a sprained ankle.
- 8:23The triage nurse, which is a routing algorithm, sends you
- 8:26directly to the specific expert who handles ankles.
- 8:29Oh, I get it. Meaning only a small fraction of
- 8:32the 389 billion parameters is actually activated for any given
- 8:36query. So it's an efficiency play to
- 8:38save on compute costs. Which is an absolute matter of
- 8:41survival because Tencent isn't just researching AI in a lab,
- 8:45they are deploying it at scale. They have applications running
- 8:48live right now, like Work Buddy, which is the most widely used
- 8:51office production activity agent in China, and Code Buddy for
- 8:54software engineers. Every single time a user asks
- 8:58Work Buddy to summarize an e-mail that requires an
- 9:01inference token. And an inference token is just
- 9:04the physical computing cost to generate that specific answer.
- 9:07Correct. The marginal cost to deliver an
- 9:10extra digital sword in one of their video games is practically
- 9:130. It's just pixels.
- 9:14But the marginal cost to deliver an AI inference token is very
- 9:18real. It requires electricity, cooling
- 9:21and really expensive of silicon. So they had to buy the servers
- 9:24now to handle the traffic tomorrow.
- 9:25Exactly. It's an inelastic cost.
- 9:27If they want to maintain their monopoly, they have no choice
- 9:29but to pay up. Which frames the central dilemma
- 9:32for anyone looking at this stock.
- 9:34I mean, we're looking at a company spending massive amounts
- 9:36of capital. So you have to ask, is this a
- 9:39value trap like a legacy company burning cash just to tread water
- 9:44or is it a misunderstood compounder?
- 9:46Right. And the sources provided
- 9:48actually run 10 cent through a classical deep value investing
- 9:52framework to figure this out. They do, and they use the
- 9:54stringent metrics pioneered by people like Benjamin Graham.
- 9:57They put $0.10 through a 19 point stress test, looking at
- 10:01everything from debt loads to historical earnings consistency
- 10:05and surprisingly 10 cent passes 13 out of the 19 metrics.
- 10:10Passing 13 out of 19 is incredibly strong for a tech
- 10:13company undergoing A multibillion dollar hardware
- 10:16transition. But let's look at what it fails,
- 10:18because that usually tells you more about a company's real
- 10:20condition. True, it completely fails the
- 10:23Benjamin Graham style price to book ratio.
- 10:25It fails it miserably. Traditional value investing
- 10:28looks for a price to book ratio of 1.5 or less, meaning you are
- 10:32paying very close to what the company's physical assets are
- 10:35actually worth. Earth if you liquidated them
- 10:37$0.10. PB ratio is currently sitting
- 10:40around 3.0. But the research argues that
- 10:42metric is completely broken when applied to a modern software
- 10:45monopoly right. Exactly because book value only
- 10:48measures tangible assets. Factories, warehouses, inventory
- 10:52tractors. Tencent is an intangible heavy
- 10:55company. Book value completely ignores
- 10:58the network effect of having 1.4 billion users locked into
- 11:01WeChat. It assigns 0 value to their
- 11:04massive intellectual property portfolio in gaming, and it
- 11:07completely ignores the proprietary data reserves they
- 11:10use to train the Hanyuan model. Right.
- 11:13Measuring a software giant by its physical book value is like
- 11:15measuring the genius of a mathematician by, I don't know,
- 11:20weighing their brain on a scale. You're measuring the physical
- 11:23container, but missing the actual value completely.
- 11:25That's the perfect analogy because when you look at the
- 11:27deep value metrics where 10 cent does pass, the underlying
- 11:31economics are just phenomenal. We are looking at a 10 year
- 11:34average return on equity of roughly 20.9%.
- 11:36It's. Huge.
- 11:37Yeah. And they have historical free
- 11:39cash flow margins that consistently range between 19
- 11:42and 39% over the last decade. And despite the massive AI
- 11:46spending spree, they still have a fortress balance sheet with
- 11:4958.2 billion RMB in net cash. Which totally explains why
- 11:53legendary value investors, you know, Charlie Monger, Lelu
- 11:57Monish Pabrai, Michael Bury have historically circled this stock.
- 12:01They look at the numbers and see an unassailable toll bridge
- 12:03monopoly. But there's a reason this stock
- 12:06isn't trading at a massive premium.
- 12:08There is a geopolitical discount applied to Chinese equities and
- 12:11we really have to address the long term structural risks, the
- 12:14research points out. We do.
- 12:16The sources outline 3 permanent risks that really cannot be
- 12:19ignored. The first is demographic decline
- 12:21in China and aging population and plummeting birth rates mean
- 12:25the core demographic for high engagement competitive mobile
- 12:28gaming is permanently contracting.
- 12:30It's just simple math. Fewer teenagers means fewer
- 12:33people with the reflexes and free time to buy battle passes
- 12:36in honor of kings. Precisely The second risk is
- 12:39state regulatory oversight. The government maintains strict
- 12:43control over the domestic Internet.
- 12:45That means arbitrary gaming approval freezes, playtime
- 12:48restrictions for minors, and intense algorithmic transparency
- 12:52laws. It constantly threatens to put a
- 12:54ceiling on their total addressable market.
- 12:57But the third risk is arguably the most dangerous for this
- 13:00specific AI transition, which is US expert controls on advanced
- 13:04GPU's. Right, this is the bottleneck
- 13:07for their entire AI ambition. The United States is actively
- 13:11restricting China's access to the frontier chips made by
- 13:14companies like NVIDIA. If Tencent is permanently
- 13:17handicapped from acquiring the best silicon in the world, it
- 13:20could severely limit their ability to train models at the
- 13:22same speed as Western competitors.
- 13:24Even if they optimize their MO E architectures and pivot to
- 13:27domestic silicon, I mean, being cut off from the global hardware
- 13:30supply chain is a severe structural disadvantage.
- 13:32Yeah. So we have a domestic Moat boxed
- 13:34in by international geopolitics. How is management actually
- 13:38dealing with this because I know process their largest and oldest
- 13:40shareholders mechanically dumping shares on the open
- 13:42market every single day. Yeah, Process is running an
- 13:45open-ended buyback program of its own.
- 13:47So they are selling 10 cent shares to fund it.
- 13:49It creates immense constant downward pressure on the stock
- 13:53price. But 10 Cent isn't just letting
- 13:55the price collapse. This is where their capital
- 13:57allocation gets brilliant. They are acting like a massive
- 14:00financial sponge. They're weaponizing that 58
- 14:04billion RMB cash pile to absorb the selling pressure with
- 14:08aggressive price insensitive buybacks.
- 14:11Just in the second quarter alone, they repurchased 37.4
- 14:15million shares for 16.9 billion Hong Kong dollars.
- 14:19Management looks at the geopolitical discount, they know
- 14:22the reality of their cash flow, and they realize the highest
- 14:24return investment they can make is simply buying their own
- 14:27company at a discount. Retiring equity when it's cheap
- 14:30is exactly what you want to see as a long term shareholder.
- 14:33Which brings us to the ultimate question for you listening to
- 14:35this deep dive. After factoring in the cash
- 14:38flow, the AI spend, the risks in the buybacks, what is a single
- 14:42share of this company actually worth today?
- 14:43The sources provided A comprehensive discounted cash
- 14:46flow or DCF valuation to find the intrinsic value.
- 14:51Let's walk through this carefully.
- 14:52ADCF simply calculates the present value of all the future
- 14:55cash a company is expected to generate.
- 14:58To be intellectually honest, the analysts laid very conservative
- 15:01groundwork here. They didn't base the math on
- 15:03some optimistic future peak. They establish a baseline of
- 15:0724.0 billion U.S. dollars in normalized free cash flow.
- 15:12OK, let's pause there. Why 24 billion?
- 15:14Because that strips out the temporary distortion of the
- 15:18massive GPU prepayments we discussed earlier, giving us a
- 15:21true picture of the underlying business as it operates today.
- 15:24They assume 9.08 billion shares outstanding, which accounts for
- 15:28the current buyback rate, and they apply a 9.0% discount rate,
- 15:32which is the required rate of return an investor would demand
- 15:34to take on the risk of holding the stock.
- 15:36Got it. So 24 billion in cash divided
- 15:38among 9 billion shares discounted at 9%.
- 15:41Now they ran this baseline through three very distinct
- 15:44future scenarios. Let's start with scenario one,
- 15:46the no growth case. This is the absolute doomsday
- 15:50model. It assumes Tencent never grows
- 15:52its cash flow by a single penny ever again.
- 15:55It just spits out exactly $24 billion a year, forever.
- 15:59So if the core business flatlines completely and AI adds
- 16:020 value, where does the stock price land?
- 16:05Under that 0% growth assumption, the intrinsic value is $30.24
- 16:10per share. OK, so 30 bucks is the absolute
- 16:13floor. Let's move to scenario 2, the
- 16:15conservative case. For this scenario, they looked
- 16:18at the worst historical growth phase Tencent has endured over
- 16:21the last 10 years. That was between 2021 and 2023,
- 16:25during the severe regulatory crackdowns.
- 16:27They grew at just 3% annually during that time, right?
- 16:30So, the analysts asked, what if that sluggish 3% growth becomes
- 16:34the permanent new normal for the next decade?
- 16:36So if they only managed to scrape together 3% growth, what
- 16:39is the value? The intrinsic value jumps to
- 16:41$41.85 per share. Finally, we have scenario 3, the
- 16:45nominal case. This assumes their transition
- 16:48actually works and plus advertising continues to command
- 16:52premium pricing. Work Buddy monetizes heavily in
- 16:55the enterprise sector and they return to their historical 10
- 16:58year average growth rate of roughly 12%.
- 17:00Right. If they execute and hit that 12%
- 17:03historical growth, the intrinsic value more than double s to
- 17:06$91.51 per share. So we have a clear map, $30 if
- 17:11growth dies completely, $41.00 if growth is severely stunted,
- 17:15and $91.00 if they return to their historical average.
- 17:19Where is the stock actually trading right now?
- 17:21It's hovering around 56 to 58 U.S. dollars, which is roughly
- 17:25350 to 400 Hong Kong dollars. Which means the market is
- 17:28pricing $0.10 squarely in no man's land, right between the
- 17:31conservative and the historical scenarios.
- 17:33It's offering you a margin of safety compared to their
- 17:35potential upside, but it's far above the absolute floor if the
- 17:39business stops growing. The market is incredibly
- 17:41rational here. It is acknowledging that the
- 17:43core mode is unassailable, but it is aggressively discounting
- 17:47the future growth because of the demographic risks, the
- 17:50geopolitics and this massive mandatory AI capital
- 17:54expenditure. It's the cost of crossing the
- 17:56bridge into the next era of tech, really.
- 17:59You have a company experiencing unprecedented margin expansion
- 18:02in its core software day job, but all that extra cash is
- 18:06immediately being diverted to fund a heavy physical hardware
- 18:10transition. Securing those ASICS and GPU's
- 18:12today to train the Hanyuan model is really the only way to ensure
- 18:17their 1.4 billion user monopoly survives the next 10 years.
- 18:21The capital drain is massive, yes, but it deepens an already
- 18:24immense Moat. Which leaves us with a
- 18:26fascinating macro level question.
- 18:27Based on these sources, we're talking about an apex predator
- 18:30spending 10s of billions just to stay in the game.
- 18:33I want to hand it over to you for a final thought on what this
- 18:35means for the rest of the industry.
- 18:37It really makes you reconsider the entire landscape.
- 18:40I mean, if a dominant giant like $0.10 with practically infinite
- 18:43resources, 1.4 billion captive users, and virtually 0 marginal
- 18:48cost for its legacy products has to endure negative free cash
- 18:51flow just to secure enough compute power for the generative
- 18:54AI era, what does that mean for the survival chances of smaller
- 18:58tech companies trying to compete in this space?
- 19:00It's a great point. The financial gravity of AI
- 19:03might just be too heavy for anyone but the absolute biggest
- 19:06players to survive. That is exactly the kind of
- 19:09question we need to be asking as this technology unfolds.
- 19:13And that wraps up our deep dive on the 10 Cent Q2 2026 financial
- 19:17report. Thank you for joining us as we
- 19:19unpack the numbers, the mechanics, and the strategy
- 19:21behind this digital giant. Keep asking questions, keep
- 19:24looking past the headline metrics, and keep exploring the
- 19:26data. We'll catch you on the next one.