Latest / Deep Values: Value Investing Deep Dives / Kaspi Is Worth $0.75 or $97,596? Why the DCF Breaks
Transcript
- 0:00Imagine an artificial intelligence system perfectly
- 0:03analyzing millions of financial data points, running the math
- 0:07Florida State without a single computational error, and
- 0:11concluding that a $93 stock is mathematically proven to be
- 0:16worth either $0.75 or. Or $97,000.
- 0:22Exactly. Yeah, welcome to the absolute
- 0:25off road wilderness of modern value investing.
- 0:28It really is a complete breakdown of traditional metrics
- 0:32because when you take standard, you know, mid century valuation
- 0:35models and just blindly apply them to complex modern financial
- 0:39platforms, the map just completely stops matching the
- 0:42territory, right? And that is exactly what we are
- 0:44navigating for you today. Welcome to a highly customized
- 0:48deep dive because you know you're listening because you
- 0:51have a very specific mission, right?
- 0:53You want to master Buffett style intrinsic evaluation, but you
- 0:56want to do it without drowning in just like an avalanche of
- 1:00information overload. You want the signal, not the
- 1:02noise. Exactly the signal.
- 1:04And today we are dissecting an exclusive freshly generated
- 1:08research package from Deep Value Intelligence, which is the multi
- 1:11agent AI analyst from the Deep Values ecosystem and.
- 1:15Just as a quick note for those of you who want to run these
- 1:18exact metrics on your own portfolios, you can actually
- 1:21pull this data yourself. Oh yeah, definitely.
- 1:23Yeah. For Apple users, there is the
- 1:24Deep Values iOS app for intrinsic valuation or you can
- 1:28use the Buffett Indicator app to gauge the broader market.
- 1:32And if you are on Android you can use Bubble Watch Plus you
- 1:36know the full AI analyst suite is always up at deep values dot
- 1:40AI. Right.
- 1:41Highly recommend checking this out.
- 1:43So our mission today is to parse this new AI analysis on a
- 1:46company called Caspi dot KZ. The AI has basically handed us
- 1:51this report packed with dense financial metrics.
- 1:53And the core question we need to answer is, is the stock a
- 1:57classic Benjamin Graham style deep value play or is it just
- 2:00something else entirely? Right, because the underlying
- 2:03data in this package is phenomenal, but as that you
- 2:08know, 75 cent valuation suggests it is full of structural
- 2:11contradictions. Totally.
- 2:13Contradictions that we really need to untangle if you want to
- 2:15understand what the business is actually worth.
- 2:18Right. But before we jump into those
- 2:20contradictions, we kind of need a hard reality check on the
- 2:22calendar. Oh.
- 2:23The timeline, yeah. This is crucial, right?
- 2:26The AI report notes an important date correction right off the
- 2:29bat because today is Thursday, October 22nd, 2026, and if
- 2:33you're out there scanning the financial headlines, you might
- 2:36be, you know, looking for Caspie's Q3 earnings, but.
- 2:39They aren't out yet. Exactly.
- 2:41The report is very clear, Q3 or the nine month results, they
- 2:44aren't scheduled to drop until November 9.
- 2:46So everything we are analysing today is strictly grounded in
- 2:50the Q2 and first half 2026 reality.
- 2:53Which was published back in August.
- 2:55Right. Yes, back in August.
- 2:57Yeah, because in financial markets, reacting to the wrong
- 3:00timeline is just an easy way to make a catastrophic error.
- 3:03I mean, we really have to draw a hard box around the data we are
- 3:06actually examining, otherwise the the growth trajectories
- 3:10become totally distorted. So let's look at those Q2 growth
- 3:13numbers. Inside that box, the AI breaks
- 3:16down Caspy's performance across 3 segments.
- 3:18You've got marketplace, payments and fintech.
- 3:21Now initially I was thinking of their growth like this massive 3
- 3:24lane highway where every single lane is accelerating.
- 3:27I mean that makes sense on paper, but.
- 3:29But looking closer at the mechanics, it's not just
- 3:32parallel lanes moving independently, it's it's an
- 3:35ecosystem. Yeah, it's a closed loop engine,
- 3:38sort of a flywheel effect. Walk us through how you're
- 3:40seeing that interaction in the Q2 data.
- 3:43OK. So first you have the
- 3:44marketplace segment, right? The gross merchandise value,
- 3:47which is the total cash value of all the goods actually sold
- 3:50across their platform, not just like Caspi's cut of it that hit
- 3:542.287 trillion. Kazakhstani tenge trillion,
- 3:58yeah, which is up 15% year over year.
- 4:00Right, up 15%. So that active marketplace
- 4:04generates massive amounts of consumer transaction data.
- 4:07Then you look at the payment segment, which is obviously
- 4:10moving all that money around. The total payment volume is
- 4:13sitting at 12.654 trillion tenge, up 13%.
- 4:18And because they control both the digital marketplace where
- 4:21the, you know, transaction happens and the payment network
- 4:24that settles the funds. I see everything.
- 4:26Exactly, they have total real time visibility into their
- 4:29consumers cash flow and purchasing habits.
- 4:31Which directly feeds the third cylinder, right, The fintech
- 4:34loan portfolio. Because they have all that
- 4:36proprietary behavioral data, they can underwrite consumer
- 4:39loans with incredibly low risk compared to, say, a traditional
- 4:43bank. Because a traditional bank is
- 4:44guessing based on a credit score, but Caspie knows exactly
- 4:48what you buy and when you get paid.
- 4:50Exactly. And you see the result in the
- 4:52data. Their average Fintech loan
- 4:54portfolio is sitting at 7.236 trillion tenge, which is up a
- 4:58massive 18%. What's fascinating here is how
- 5:02that compounding loop actually impacts the bottom line, because
- 5:05this is where honestly a lot of investors get tripped up when
- 5:08valuing rapidly scaling platforms.
- 5:11Right, because here is the anomaly in the report.
- 5:14You have this massive volume moving through the ecosystem.
- 5:16The top line revenue for Q2 was up 15% to 1.112 trillion tenge.
- 5:22Massive top line growth. Huge.
- 5:24But the actual Q2 net income, the profit leftover at the very
- 5:28end of the day was essentially flat year over year at 259
- 5:32billion tenge. Wow.
- 5:33Yeah. So they grew the loan book by
- 5:3618%, but the net cash hitting the bottom line barely moved.
- 5:39Right. And this happens so frequently
- 5:40with financial technology platforms that are scaling at
- 5:43this velocity. I mean, yes, the revenue jumps,
- 5:47but expanding a loan portfolio by 18% requires heavy immediate
- 5:53capital provisioning. Like holding reserves.
- 5:56Exactly. You have to set aside cash
- 5:58reserves to satisfy regulators for those new loans.
- 6:01You are also pouring money into customer acquisition and scaling
- 6:05the, you know, the technological infrastructure to handle that
- 6:0813% bump in payment volume. So if I'm understanding the
- 6:12mechanics right, they are essentially taking the profits
- 6:15generated by the existing loan book and the marketplace fees
- 6:18and instantly plowing that cash back into acquiring new users
- 6:22and backing new loans. Yeah.
- 6:23They are paying for future growth today.
- 6:26I mean, the net income looks flat on this quarter's
- 6:28statement, but the total scale of the enterprise, the actual
- 6:32earning power of that larger loan book moving forward is
- 6:36significantly larger. That makes total sense.
- 6:39And that structural reality is what really sets the stage for
- 6:42our valuation problem here. OK, let's unpack this valuation
- 6:46puzzle, because this is where the AI really drew a line in the
- 6:49sand. Looking at the October 21 close,
- 6:52Caspie's share price was sitting at $93.51, right?
- 6:57That gives it a market cap of roughly $17.77 billion.
- 7:02Now it's trailing price to earnings ratio, or PE, which is
- 7:05essentially how much you pay for every dollar of profit the
- 7:08company makes, is hovering around 8X.
- 7:11And in the broader tech world, an 8X multiple sounds incredibly
- 7:15cheap. It does.
- 7:17I mean, it sounds like a steal, but a value investor doesn't
- 7:21just look at the earnings. No, they look at the balance
- 7:23sheet. And that is the friction point
- 7:25Caspi's price to book ratio. The PB is about 3.3 X.
- 7:30Based on that single metric, the deep value intelligence system
- 7:33rendered a very specific verdict.
- 7:35Now what did it say? It concluded that Caspi is a
- 7:38quote high quality compounder trading at a low multiple, but
- 7:42it is not a classic Benjamin Graham deep value stock.
- 7:45Right, because the AI runs a rigorous checklist and Caspi
- 7:48easily passes the test for profitability, right?
- 7:51I mean, it passes the return on equity and return on assets
- 7:54tests. It passes the current free cash
- 7:57flow yield test. But it totally fails the strict
- 8:00Benjamin Graham rule that a deep value stock must have a price to
- 8:03book ratio of 1.5 or less. A 3.3 XPB is like more than
- 8:09double the threshold. Yeah, but I have to push back on
- 8:12this a bit for our listeners. If a company is highly
- 8:15profitable generating cash and passing almost every operating
- 8:18quality test, why are we penalizing it for its price to
- 8:22book ratio? Isn't the strict 1.5 rule a bit,
- 8:26you know, outdated for a modern digital platform?
- 8:29Well, this raises an important question about accounting
- 8:31mechanics and what value actually means.
- 8:34Today, we need to remember why Benjamin Graham cared so much
- 8:38about book value in the first place.
- 8:39Right back in his era, right? Historically, book value was
- 8:43basically liquidation value. If you shut down a textile mill,
- 8:47sold the physical assets, the machinery, the buildings, the
- 8:49land, and paid off all the debt, the book value is the cash
- 8:53leftover for shareholders. OK, Graham wanted to buy
- 8:56companies trading below that number to guarantee an ultimate
- 8:59margin of safety. But Cosby doesn't own heavy
- 9:02industrial factories. They don't.
- 9:04They own network effects software, algorithms,
- 9:07proprietary data and a brand heavily integrated into users
- 9:11daily lives. And here is where traditional
- 9:13accounting completely breaks down.
- 9:16How so? Well, if a legacy company spends
- 9:19$100 million building a physical factory, that factory goes onto
- 9:23the balance sheet as an asset. It directly increases the book
- 9:26value. Got it.
- 9:28But if a tech platform spends $100 million on software
- 9:31development, data analytics and customer acquisition to build
- 9:34out a digital ecosystem. Oh, let me guess, accounting
- 9:37rules force them to expense it? Exactly.
- 9:39Counting rules often force them to treat that as an immediate
- 9:42expense on the income statement. It does not get capitalized as a
- 9:45long term asset, so their true value creating assets are just
- 9:49completely invisible to the balance sheet.
- 9:52That makes perfect sense. So by expensing their growth
- 9:54instead of capitalizing it, their official quote UN quote
- 9:57book value looks artificially tiny compared to their actual
- 10:01earnings power. Right.
- 10:03So when the AI strict filters for a 1.5 price to book ratio,
- 10:07it's essentially hunting for a physical factory to liquidate,
- 10:10missing the entire point of a capital efficient software
- 10:13platform. That is exactly the structural
- 10:15flaw in rigidly applying Graham's PB rule to modern tech,
- 10:19and the AI report confirms this disconnect when it runs a
- 10:22stringent no growth discounted cash flow model, A DCF.
- 10:26Right, because ADCF projects future cash flows and discounts
- 10:30them back to figure out what the company is worth today.
- 10:33Exactly, and the math on that was brutal.
- 10:36Yeah, the AI found that to achieve a 30% margin of safety
- 10:40using a zero growth DCF model, the stock price would have to
- 10:43drop all the way down to $53.46. And since it's currently trading
- 10:48at 9351. It completely fails that safety
- 10:51test too. So the AI is telling us
- 10:53basically do not buy this expecting a distressed
- 10:56liquidation level bargain. You are paying a reasonable
- 11:00price for a compounding engine, not a dying factory.
- 11:03Exactly. Here's where it gets really
- 11:05interesting, though, because spreadsheets and accounting
- 11:08rules only tell you part of the story.
- 11:10If you want to know what's really happening inside the
- 11:13machine, you follow the insiders.
- 11:16Oh, absolutely. Insider allocation of capital is
- 11:19often the most honest signal of management team broadcasts to
- 11:22the market. Well, the baseline signal here
- 11:23is massive confidence. At the start of 2026, founders
- 11:27Mikhail Lamtadze and Viatra Slav Kim held incredibly concentrated
- 11:32stakes. We're talking 22.62% and 20.78%
- 11:38respectively. That is huge.
- 11:40That's over 40% of the entire enterprise held by just two
- 11:44people. Their personal net worth is tied
- 11:46directly to this platform's long term survival.
- 11:49And high insider ownership like that provides a real structural
- 11:51safety net. I mean, it usually prevents
- 11:53management from making reckless short term decisions just to
- 11:57engineer a temporary bump in the stock price for the next
- 12:00quarterly earnings call. They are playing a multi decade
- 12:04game. Exactly.
- 12:05We also saw a hugely bullish move in 2026.
- 12:09Tencent, along with Lumtadze and senior management, bought a
- 12:12colossal 6,000,000 ADS block directly from Bearing.
- 12:16Bering was liquidating a massive chunk, and management stepped up
- 12:20alongside a major global tech player to just absorb it.
- 12:24And putting hard Personal Capital behind a block purchase
- 12:27of that scale, that indicates A profound belief in the intrinsic
- 12:30value of the ecosystem. But then we have the bearish
- 12:32countersignals. Right.
- 12:33It's not all one sided, no. The report notes that Co founder
- 12:36Kim actually trimmed his position slightly, dropping from
- 12:4020.78% to about 20% by midyear. And honestly, more importantly,
- 12:45we have a complete buyback drought.
- 12:47Yeah, this is the tricky part. Cosby finished its seventh share
- 12:50buyback program back on February 27th, 2026, repurchasing almost
- 12:551.3 million shares for over 50 billion tenge.
- 12:58And since then? Nothing.
- 12:59The AI scraped all investor relations data through today,
- 13:02October 22, and found a 0 news of a new repurchase
- 13:06authorization. If we connect this to the bigger
- 13:08picture, this is a classic capital allocation dilemma.
- 13:12The AI categorizes this as mixed recent trading and it really
- 13:16forces you to weigh opposing strategic theories.
- 13:19Right, because if you're listening to this, you're
- 13:20probably thinking, OK, if the company is such a great
- 13:23compounding engine, why would a founder trim his shares?
- 13:26And why would the company suddenly stop buying its own
- 13:28stock? Well, let's look at the
- 13:30underlying business strategy. A founder trimming a fraction of
- 13:33a percent off a 20% stake. I mean, that might just be
- 13:36personal estate planning or like portfolio diversification.
- 13:39Right, exactly. It's the buyback pause that is
- 13:42far more analytically significant, because when a
- 13:46highly profitable company abruptly stops repurchasing its
- 13:49own shares, it generally points to one of two strategic
- 13:53realities. OK.
- 13:54So either management thinks the stock at $93 is fully priced,
- 13:57meaning buying it is just a bad return on investment for
- 14:00shareholders. Or they see a massive capital
- 14:04intensive growth opportunity internally that requires them to
- 14:07hoard every single dollar of cash today.
- 14:10Which ties perfectly back to that 18% growth in the fintech
- 14:14loan portfolio we discussed in Segment 1.
- 14:16Yes, you see the connection. Right, because if you are
- 14:20aggressively expanding a loan book in a complex interest rate
- 14:23environment, you need immense regulatory capital.
- 14:27Every dollar spent on a share buyback is a dollar you cannot
- 14:29use to back a new, highly profitable loan.
- 14:32And that is the most likely strategic explanation.
- 14:35The buyback pause isn't necessarily a warning siren
- 14:38about the stock being overvalued.
- 14:40You know, it could simply be the necessary cost of feeding that
- 14:43rapidly accelerating fintech cylinder.
- 14:46OK, which brings us to the most mind bending part of this entire
- 14:50deep value Intelligence Report. The discounted cash flow stress
- 14:53tests. Oh boy, the DCF roller coaster.
- 14:56This is where the AI perfectly illustrated how dangerous
- 14:59standard math can be. It is a master class in the
- 15:02limitations of algorithmic modeling without human
- 15:04structural context. So the AI runs its DCF models
- 15:08using a trailing 12 month free cash flow of about $1.98
- 15:12billion. It uses 190 million shares and a
- 15:1713.66% discount rate, which it cross checked for October 21.
- 15:22Very standard, rigorous inputs. It's a standard playbook, yeah.
- 15:26Right. But when it runs the strict
- 15:28conservative scenario based on historical minimum cash flow
- 15:31growth, which was a -92.4%, it concludes the intrinsic value of
- 15:37CASPIE is exactly $0.75 per share.
- 15:41A $93 stock mathematically valued at under a dollar.
- 15:46But wait, it gets better. If you're listening to this and
- 15:49thinking, OK, that's just a broken worst case scenario,
- 15:51let's look at the strict nominal scenario.
- 15:54The other extreme? The AI takes the historical
- 15:57average free cash flow growth a +338%, plugs it in, and spits
- 16:03out an intrinsic value of $97,596 per share.
- 16:09So the AI tells us this asset is worth either pocket change or a
- 16:13luxury sports car. It's wild.
- 16:15Wait, how can cash flow be this distorted?
- 16:17I mean, if a company has cash coming in the door, isn't that
- 16:20universally a good thing? Not if you're a bank, and that
- 16:23is the fundamental misunderstanding here.
- 16:26The math the AI ran is flawless, but the underlying business
- 16:29structure breaks the formula. Because of the fintech side,
- 16:32right? CASPI is essentially a deposit
- 16:35funded financial institution and for a bank, traditional free
- 16:39cash flow from operations is just hopelessly noisy.
- 16:43Noisy how? Like how does that actually
- 16:45work? Well, think about a bank like a
- 16:47high security storage facility for cash.
- 16:50If 1000 customers suddenly deposit $1,000,000 today, your
- 16:54operational cash flow looks incredible.
- 16:57Sure, cash came in. Right.
- 16:59A standard algorithm sees that massive influx and says wow
- 17:02growth is up 338%. But the bank doesn't own that
- 17:06money. It is a liability.
- 17:09You owe it back to the depositors.
- 17:11And when the bank actually puts that money to work to make a
- 17:13profit, like when they originate a consumer loan, cash literally
- 17:18leaves the building. Exactly.
- 17:19When you originate a loan that will pay you healthy interest
- 17:22for years to come, cash flows out the door.
- 17:25Today, the traditional algorithm sees that massive outflow and
- 17:29panics, calculating a 92% drop in cash flow.
- 17:33Wow. But in reality, you just traded
- 17:35A liability, the deposits, for a highly profitable asset, the
- 17:39loan. So this constant wild swinging
- 17:42between taking in customer deposits and aggressively
- 17:44writing new loans creates these absurd 300% spikes and 90% drops
- 17:50in standard free cash flow metrics.
- 17:52And if you plug those wildly distorted inputs into a highly
- 17:55sensitive DCF algorithm, the outputs become economically
- 17:59meaningless. $0.75 from 97 grand.
- 18:01Exactly. That is why the Deep Value
- 18:03Intelligence Report explicitly warns against trusting
- 18:07conventional DCF models for deposit funded institutions.
- 18:10It suggests that alternative valuation methods like a
- 18:13residual income model, a dividend discount model, or a
- 18:16some of the parts approach are far superior because they
- 18:19account for the cost of equity and banking capital structures.
- 18:23Though I should mention the AI did provide a fourth quote UN
- 18:26quote normalized scenario. It did try to smooth it out,
- 18:29yeah. Right, it attempted to smooth
- 18:31out the banking noise by using a projected 10 year cash flow
- 18:35compound annual growth rate of 24.28%.
- 18:39Under that smoothed out view, it calculated an intrinsic value of
- 18:43roughly $226.27 per share. Which is a very different
- 18:47picture. Very different.
- 18:48That would mean achieving a 30% margin of safety requires a
- 18:51price of $158.39, making the current $93 price look
- 18:56incredibly cheap. But again, even the AI says to
- 19:00treat that normalized number with extreme caution given the
- 19:03structural noise we just outlined.
- 19:05It's a helpful guidepost, you know, but it is not a guarantee.
- 19:08You have to understand the business, not just the math.
- 19:11So what does this all mean? I mean, we've covered a ton of
- 19:14complex ground today. We established that Caspi isn't
- 19:17just growing in parallel lanes, it's operating a compounding
- 19:20flywheel where the marketplace data feeds the payment network,
- 19:24which ultimately DE risks a massive 18% growing loan
- 19:28portfolio. And we've also seen that while
- 19:31it fails the classic Benjamin Graham price to book test, that
- 19:34failure is largely due to modern accounting mechanics penalizing
- 19:37tech platforms for expensing their ecosystem growth rather
- 19:41than capitalizing physical assets.
- 19:43Right. We untangled the mixed insider
- 19:45signals, realizing that pausing A buyback program might just be
- 19:48a strategic move to hoard cash for that exact loan portfolio
- 19:51expansion. And finally, we saw first hand
- 19:54how applying standard cash flow formulas to a deposit taking
- 19:57institution results in valuations ranging from $0.75 to
- 20:01$97,000. It is a stark reminder that
- 20:05knowing the math is only half the battle.
- 20:07You have to understand the underlying business generating
- 20:09the numbers. Definitely.
- 20:11Now, as always, this deep dive is just scratching the surface.
- 20:15The Deep Value Intelligence package we use today isn't just
- 20:18a text summary. It includes a zip file with 39
- 20:21Bilingual Basic Black charts, a full markdown report covering
- 20:24management buybacks and scorecards, plus the actual
- 20:28Python research scripts. That's the best part honestly,
- 20:31being able to see the code. Yeah, totally.
- 20:34If you want to run this data yourself, you can track these
- 20:37exact metrics daily by downloading the Deep Values app
- 20:40on iOS, the Buffett Indicator app for a market wide analysis,
- 20:44or Bubble Watch on Android. You can also explore the AI
- 20:47analyst directly at deep values dot AI and find more life
- 20:51philosophy and investing podcasts on our Spotify and
- 20:54Apple feeds. All those links are just they're
- 20:56essential tools for a modern value investor.
- 20:59The tools are incredibly powerful, but as we saw today,
- 21:02they still require a human to interpret them correctly.
- 21:06And Speaking of human interpretation, here is our
- 21:08required disclaimer. This deep dive is for
- 21:10educational and research purposes only.
- 21:13It is not investment advice. It is based on public sources
- 21:16that may be revised or interpreted differently.
- 21:19You must do your own due diligence before making any
- 21:21investment decisions whatsoever. Trust but verify.
- 21:24Always. We started this episode talking
- 21:27about the off road wilderness of modern investing, so I want to
- 21:32hand it over to you for one final thought.
- 21:34As you, our listener, turn this off and head back into the
- 21:37markets, what should they be pondering?
- 21:39Well, this raises an important question that I think every
- 21:42modern investor needs to wrestle with.
- 21:44We just analyzed A flawless, computationally perfect AI model
- 21:49that processed years of data, and that perfect system told us
- 21:54with a straight digital face that depending on which
- 21:57historical metric you select, this company is either
- 22:00essentially worthless or worth more than most people make in a
- 22:03year, right? The question is this.
- 22:06If the most advanced artificial intelligence in the world can be
- 22:09mathematically precise but completely economically blind
- 22:13without structural context, does that finally prove that true
- 22:16value investing will always remain as much a human art as it
- 22:20is a mathematical science? Well.
- 22:23The algorithms can build the map, but you still have to know
- 22:25how to navigate the terrain. Thank you for joining us on this
- 22:28journey and we will see you on the next deep dive.