Latest / AI Investment Edge: Stocks, Markets & Money / NVIDIA: Where’s the Cash? | EP 013
Transcript
- 0:00SPEAKER_01: NVIDIA can report almost $60 billion of quarterly profit,
- 0:04SPEAKER_01: yet collect far less cash in that same quarter.
- 0:07SPEAKER_01: Is that a temporary working capital squeeze inside a brilliant franchise?
- 0:11SPEAKER_01: Or an early sign that the AI build-out is borrowing tomorrow's demand?
- 0:16SPEAKER_01: Welcome to AI Investment Edge, episode 13.
- 0:19SPEAKER_01: Today we are taking a deep dive into the absolute giant of the AI ecosystem.
- 0:24SPEAKER_01: You know, if you follow tech at all, you know NVIDIA.
- 0:27SPEAKER_00: Oh, absolutely.
- 0:28SPEAKER_00: I mean, they design the GPUs,
- 0:30SPEAKER_00: the networking hardware and all the software
- 0:34SPEAKER_00: that basically makes modern AI computing possible in the first place.
- 0:37SPEAKER_01: Right.
- 0:37SPEAKER_01: And they serve everyone from major cloud providers to massive enterprises.
- 0:41SPEAKER_01: But just to set the stage for you listening, let's briefly touch on their ecosystem.
- 0:47SPEAKER_00: Yeah,
- 0:47SPEAKER_00: so developers rely heavily on their CUDA software platform to actually interface
- 0:51SPEAKER_00: with those chips.
- 0:52SPEAKER_00: And then there's their NVLink technology,
- 0:54SPEAKER_00: which connects thousands of individual processors together.
- 0:58SPEAKER_01: Into these massive unified systems.
- 1:00SPEAKER_00: Exactly.
- 1:01SPEAKER_00: Plus, they operate on a fabulous manufacturing model.
- 1:03SPEAKER_01: Which means they design everything,
- 1:05SPEAKER_01: but they rely on external partners to actually physically manufacture the chips,
- 1:09SPEAKER_01: right?
- 1:09SPEAKER_00: You got it.
- 1:10SPEAKER_00: They don't own the billion-dollar fabrication plants themselves.
- 1:13SPEAKER_00: And that brings us to the core mission of today's Deep Dive.
- 1:16SPEAKER_00: We are not just sitting here debating whether AI is transformative.
- 1:19SPEAKER_00: We already know it is.
- 1:20SPEAKER_01: Yeah, that's a given at this point.
- 1:21SPEAKER_00: Right.
- 1:22SPEAKER_00: We are analyzing the actual cash flows that shareholders receive from
- 1:26SPEAKER_00: that NVIDIA ecosystem versus simply owning a piece of the AI productivity wave.
- 1:31SPEAKER_01: Because there is a huge difference there.
- 1:33SPEAKER_00: A massive difference, yeah.
- 1:35SPEAKER_01: Okay, so let's transition right into the numbers with
- 1:37SPEAKER_01: the September iPulse AI Financial Synthesis.
- 1:41SPEAKER_01: This is all refreshed against their August 26 results and the July SEC filing.
- 1:46SPEAKER_01: And looking at the top line, it's just wild.
- 1:50SPEAKER_01: $96.2 billion in quarterly revenue.
- 1:54SPEAKER_00: Which is up 106% year over year.
- 1:55SPEAKER_00: Over 100%.
- 1:55SPEAKER_00: That is nearly $100 billion moving through
- 1:57SPEAKER_01: one company in just three months.
- 2:03SPEAKER_00: The growth is exceptional.
- 2:04SPEAKER_00: I mean, there's absolutely no debating that.
- 2:06SPEAKER_00: But the synthesis highlights a really important cash versus profit gap.
- 2:10SPEAKER_01: Right.
- 2:10SPEAKER_01: The one I mentioned at the top of the show.
- 2:12SPEAKER_01: Let's unpack that.
- 2:13SPEAKER_00: So gap net income, which is just your standard accounting profit, was $59.7 billion.
- 2:19SPEAKER_00: But the actual operating cash flow was only $24.1 billion.
- 2:24SPEAKER_01: Wait, hold on.
- 2:24SPEAKER_01: The profit is nearly $60 billion, but the operating cash flow is only $24 billion.
- 2:29SPEAKER_00: Yeah.
- 2:30SPEAKER_00: It is roughly 40% of the net income.
- 2:32SPEAKER_01: Wow.
- 2:33SPEAKER_01: So where is all that other money?
- 2:35SPEAKER_00: It's tied up.
- 2:36SPEAKER_00: And if you look at the half year numbers, it's a very similar story.
- 2:39SPEAKER_00: You have $74.4 billion in operating cash flow against $118.0 billion in net income.
- 2:46SPEAKER_00: So roughly 63%.
- 2:47SPEAKER_01: Okay, so let me try to ground this for a second.
- 2:51SPEAKER_01: It's kind of like a restaurant booking these huge corporate banquets, right?
- 2:54SPEAKER_00: Exactly.
- 2:55SPEAKER_00: Exactly.
- 2:56SPEAKER_01: Like the sales are completely real,
- 2:58SPEAKER_01: the food is delivered,
- 2:59SPEAKER_01: but the corporate clients have up to a year to actually pay the invoice.
- 3:03SPEAKER_01: So the restaurant looks crazy profitable on paper,
- 3:06SPEAKER_01: but their bank account is getting squeezed because they still have
- 3:08SPEAKER_01: to buy ingredients for the next banquet.
- 3:10SPEAKER_00: That is a perfectly accurate analogy.
- 3:13SPEAKER_00: Receivables and inventory are heavily consuming cash right now.
- 3:17SPEAKER_00: It's real revenue, yes, but the cash collection lags.
- 3:20SPEAKER_01: And what about their margins?
- 3:23SPEAKER_01: Because they have immense pricing power right now.
- 3:25SPEAKER_00: Oh, incredible pricing power.
- 3:28SPEAKER_00: They hit a 75% gross margin.
- 3:31SPEAKER_01: 75% for a hardware-centric business is just unheard of.
- 3:34SPEAKER_00: It is historic, really.
- 3:36SPEAKER_00: But the synthesis does caution that the mid-60s future margins you see in
- 3:40SPEAKER_00: the models are forecasts, not achieve results.
- 3:43SPEAKER_01: Right.
- 3:44SPEAKER_01: Okay, let's look at the balance sheet.
- 3:46SPEAKER_01: And I want to correct something here because there is this idea floating around
- 3:49SPEAKER_01: that NVIDIA is entirely debt-free, right?
- 3:52SPEAKER_00: Yeah, the archive actually had an error there.
- 3:53SPEAKER_00: They are definitely not debt free.
- 3:55SPEAKER_00: As of the July quarter,
- 3:57SPEAKER_00: they had fifty six point six billion dollars in cash and marketable securities,
- 4:01SPEAKER_00: but they also carry thirty three point four billion dollars in funded debt.
- 4:05SPEAKER_01: OK, so not debt free,
- 4:06SPEAKER_01: but doing the math that still leaves them with a very healthy net liquidity
- 4:10SPEAKER_01: of twenty three point two billion.
- 4:11SPEAKER_00: Yes, a substantial twenty three point two billion.
- 4:14SPEAKER_00: But we need to talk about capital expenditures.
- 4:17SPEAKER_01: Well, let me push back a little on that because we just established they are fabulous.
- 4:21SPEAKER_01: Taiwan Semiconductor builds the stuff.
- 4:24SPEAKER_01: So shouldn't NVIDIA's direct capex be pretty light?
- 4:27SPEAKER_00: You would think so, yeah.
- 4:28SPEAKER_00: And their direct quarterly capex is only $2.68 billion.
- 4:31SPEAKER_00: But here is the catch.
- 4:33SPEAKER_00: They have a staggering $279 billion in future supply and capacity commitments.
- 4:39SPEAKER_01: $279 billion?
- 4:40SPEAKER_00: Yeah.
- 4:41SPEAKER_01: That is a terrifying bar tab to leave open for a company that doesn't own factories.
- 4:45SPEAKER_00: It's a massive obligation.
- 4:47SPEAKER_00: Outsourcing moves the risk, you know, it doesn't eliminate it.
- 4:49SPEAKER_01: Okay, so let's connect all these fundamental numbers to the fixed numeric card and
- 4:53SPEAKER_01: the future endpoints because I want to know what this means for the stock.
- 4:56SPEAKER_01: Today's price, and we're using the Friday, October 9 close, is $229.28.
- 5:01SPEAKER_01: Right.
- 5:03SPEAKER_00: And from that $229.28, the synthesis gives us an evidence-weighted base path.
- 5:08SPEAKER_00: We're talking cumulative gains here.
- 5:10SPEAKER_01: Non-annual.
- 5:11SPEAKER_00: Exactly.
- 5:11SPEAKER_00: Not annual.
- 5:12SPEAKER_00: So the one-year endpoint landing on September 18th, 2027 is $243.76.
- 5:19SPEAKER_00: That is up 6.3%.
- 5:21SPEAKER_01: Only 6.3% for NVIDIA.
- 5:25SPEAKER_01: I mean, people are going to hear that and think the boom is completely over.
- 5:27SPEAKER_00: Well, it just models a slower absorption phase.
- 5:30SPEAKER_00: The five-year endpoint on September 18th, 2031 is $349.94.
- 5:33SPEAKER_00: That is up 52.6%.
- 5:38SPEAKER_01: But we also have to factor in dividends, right?
- 5:40SPEAKER_01: They paid a 25 cent per share dividend on October 1st?
- 5:43SPEAKER_00: They did.
- 5:44SPEAKER_00: And assuming an unchanged dividend rate, that annualizes to $1.
- 5:48SPEAKER_01: So over four quarters, that's $1.
- 5:51SPEAKER_01: And over 20 quarters, that's $5 in cumulative cash dividends.
- 5:55SPEAKER_00: Right.
- 5:55SPEAKER_00: So if you add that,
- 5:57SPEAKER_00: the cash inclusive returns are roughly 6.8% for the one year and 54.8% for
- 6:00SPEAKER_00: the five year without reinvestment.
- 6:05SPEAKER_01: Got it.
- 6:06SPEAKER_01: Now, what about the extremes, the bull and the bear cases over five years?
- 6:10SPEAKER_00: So one forensic AI advisor details some pretty wide scenarios.
- 6:15SPEAKER_00: The five-year bull is $650.
- 6:16SPEAKER_00: Whoa, $650.
- 6:17SPEAKER_00: That is a 183.5% upside.
- 6:19SPEAKER_01: Yeah, and it's based on hitting $1.2 trillion in revenue,
- 6:21SPEAKER_01: a 50% net margin, and a 26 times multiple.
- 6:23SPEAKER_00: And the bear case?
- 6:32SPEAKER_00: The bear case is $120.
- 6:33SPEAKER_00: That is a 47.7% drop.
- 6:37SPEAKER_01: A nearly 48% drop.
- 6:40SPEAKER_00: Yeah.
- 6:41SPEAKER_01: Wow.
- 6:41SPEAKER_00: Yeah.
- 6:41SPEAKER_00: Based on only $500 billion in revenue, a 32% margin, and an 18 times multiple.
- 6:48SPEAKER_01: And just briefly for you listening,
- 6:49SPEAKER_01: I want to note that the one-year bull and bear targets are simply not quantified
- 6:53SPEAKER_01: in this data.
- 6:54SPEAKER_01: We are just looking at those fixed endpoints.
- 6:56SPEAKER_00: Right.
- 6:57SPEAKER_01: So what exactly pushes us toward that $650 bull or drags us down to the $120 bear?
- 7:04SPEAKER_01: Let's talk about the four investment quadrants.
- 7:07SPEAKER_00: Sure.
- 7:07SPEAKER_00: Let's start with drivers.
- 7:09SPEAKER_00: You have the CUDA and NVLink lock-in, which is huge.
- 7:12SPEAKER_01: It's like trying to switch out a house's entire electrical foundation instead
- 7:16SPEAKER_01: of just changing a light bulb.
- 7:18SPEAKER_00: That is a great way to put it, yeah.
- 7:20SPEAKER_00: Other drivers include AI agents boosting inference demand and, of course, stock repurchases.
- 7:26SPEAKER_00: But you also have frictions.
- 7:27SPEAKER_01: Like what?
- 7:28SPEAKER_00: Custom silicon developed by competitors,
- 7:30SPEAKER_00: electricity and transformer constraints slowing down data centers,
- 7:34SPEAKER_00: and export restrictions.
- 7:36SPEAKER_01: Makes sense.
- 7:37SPEAKER_01: What about the tail opportunities and risks?
- 7:39SPEAKER_00: For tail opportunities,
- 7:41SPEAKER_00: physical AI and robotics are massive, plus dedicated power infrastructure.
- 7:46SPEAKER_00: But the tail risks?
- 7:47SPEAKER_01: The stuff dragging us to the bear case.
- 7:50SPEAKER_00: Exactly.
- 7:51SPEAKER_00: A sudden procurement slowdown that collides with those $279 billion in capacity commitments?
- 7:57SPEAKER_00: And potential Taiwan supply disruptions.
- 7:59SPEAKER_01: Here's where it gets really interesting to me.
- 8:02SPEAKER_01: We have this alpha gap.
- 8:04SPEAKER_01: The easy narrative out there is just, you know, NVIDIA is a permanent toll road on AI, period.
- 8:09SPEAKER_00: Right.
- 8:09SPEAKER_00: But the synthesis viewpoints out a major gap in that thinking.
- 8:12SPEAKER_00: Investors might be underestimating the software mode,
- 8:16SPEAKER_00: but they are overestimating how long peak hardware margins will last.
- 8:21SPEAKER_01: Ah, so you can have an amazing product but still get a disappointing stock return if
- 8:25SPEAKER_01: the current price assumes absolute endless perfection.
- 8:28SPEAKER_00: Exactly.
- 8:29SPEAKER_01: So how did the AI models actually arrive at these completely conflicting conclusions?
- 8:34SPEAKER_00: Well, there are exactly 14 AI-generated research contributions,
- 8:39SPEAKER_00: these three model backends, 11 analyst configurations, and nine personas.
- 8:45SPEAKER_01: That is a lot of cooks in the kitchen.
- 8:46SPEAKER_00: It is.
- 8:47SPEAKER_00: But I want to be clear, these are not independent human analysts.
- 8:50SPEAKER_00: Their assumptions are correlated across those back ends.
- 8:54SPEAKER_01: So what is the actual debate among all these personas?
- 8:57SPEAKER_00: The debate hinges on the duration of growth and cash quality.
- 9:01SPEAKER_00: Capital cycle skeptics are really worried about margin compression,
- 9:05SPEAKER_00: while the forensic models heavily weigh cash conversion.
- 9:08SPEAKER_01: OK, let's map out the timeline on this.
- 9:10SPEAKER_01: How does this play out quarter over quarter?
- 9:12SPEAKER_00: The path shows a slow digestion phase.
- 9:15SPEAKER_00: December 2026 is roughly 0.7% above today.
- 9:16SPEAKER_00: March 2027 is 4.7% and September 2027 is 6.3%.
- 9:18SPEAKER_00: So very flat for a while.
- 9:20SPEAKER_00: When does the recovery really hit?
- 9:21SPEAKER_00: The recovery turning point is mid to late 2028.
- 9:23SPEAKER_00: It is never mid 2026.
- 9:36SPEAKER_01: Never mid-2026.
- 9:37SPEAKER_01: Got it.
- 9:37SPEAKER_01: Yeah.
- 9:37SPEAKER_00: So hitting 12.1% by September 2028,
- 9:39SPEAKER_00: 25.6% by September 2029, and eventually that 52.6% by September 2031.
- 9:50SPEAKER_01: We also have some major dated events to factor in.
- 9:52SPEAKER_01: On August 26, they guided for $108 billion next quarter, sans China Compute.
- 9:59SPEAKER_00: Correct.
- 9:59SPEAKER_01: And then on September 28,
- 10:01SPEAKER_01: they announced this massive repurchase authorization increase of $150 billion,
- 10:06SPEAKER_01: which leaves them with $235 billion available through fiscal 2028.
- 10:11SPEAKER_00: Yes.
- 10:12SPEAKER_00: But keep in mind, authorization does not guarantee a stock price floor.
- 10:16SPEAKER_00: It just means they have the option to buy.
- 10:18SPEAKER_01: Right.
- 10:19SPEAKER_01: They don't have to spend it.
- 10:20SPEAKER_01: So how has this overall view evolved since the last model run?
- 10:24SPEAKER_00: Well,
- 10:24SPEAKER_00: we don't have a prior podcast baseline since this is NVIDIA's first episode here,
- 10:29SPEAKER_00: but July's batch had 12 contributions while September has 14.
- 10:32SPEAKER_01: Did the focus change?
- 10:34SPEAKER_00: It did.
- 10:35SPEAKER_00: July focused on CapEx digestion.
- 10:37SPEAKER_00: September keeps that tension,
- 10:39SPEAKER_00: but it adds intense scrutiny on cash conversion,
- 10:42SPEAKER_00: grid limits, and those $279 billion capacity commitments.
- 10:47SPEAKER_01: Okay, so bringing this to the listener,
- 10:49SPEAKER_01: if you're watching the news, what are the invalidation tests?
- 10:53SPEAKER_01: What actually breaks the base case?
- 10:55SPEAKER_00: You need to watch two main metrics.
- 10:57SPEAKER_00: Sequential data center growth needs to stay above 10% and gross margins need to stay above 72%.
- 11:04SPEAKER_01: And what are the warning signs?
- 11:06SPEAKER_00: If you see stretching receivable days,
- 11:08SPEAKER_00: inventory pressure,
- 11:09SPEAKER_00: or a stall in cash collections while those massive commitments turn
- 11:12SPEAKER_00: into actual financial charges.
- 11:14SPEAKER_01: So cash conversion is a huge test.
- 11:17SPEAKER_00: It is one test alongside margins, valuation and customer demand.
- 11:21SPEAKER_00: It's not the only thing that matters.
- 11:23SPEAKER_01: Right.
- 11:24SPEAKER_01: So to synthesize all this,
- 11:26SPEAKER_01: NVIDIA is this extraordinary computing business,
- 11:29SPEAKER_01: but they're facing a very ordinary shareholder test.
- 11:33SPEAKER_00: Exactly.
- 11:34SPEAKER_00: How much of the AI boom becomes durable, collected cash and how much is already priced in?
- 11:39SPEAKER_01: That is the big question.
- 11:41SPEAKER_01: So I want to leave you listening with a final thought to ponder.
- 11:45SPEAKER_01: If you look back at historical tech booms,
- 11:47SPEAKER_01: the infrastructure builders made the initial fortunes.
- 11:50SPEAKER_00: Oh, absolutely.
- 11:50SPEAKER_00: Every time.
- 11:51SPEAKER_01: But the ultimate durable returns shifted to the secondary layer of companies,
- 11:56SPEAKER_01: the Googles and Amazons that figured out how to monetize that new infrastructure
- 12:00SPEAKER_01: over decades.
- 12:01SPEAKER_01: So is NVIDIA immune to the cycle?
- 12:04SPEAKER_00: That is a fascinating question for people to explore on their own.
- 12:07SPEAKER_01: To explore these scenarios,
- 12:09SPEAKER_01: the financial evidence, and the full quarterly path, visit iPulseAI.com.
- 12:14SPEAKER_01: Thanks for listening to AI Investment Edge.