Latest / Deep Values: Value Investing Deep Dives / [Q3] Broadcom's Dual Engine AI Money Machine
Transcript
- 0:00Welcome back to the Deep Dive. Today we're looking at a company
- 0:03that for a long time really flew just under the radar for most
- 0:07people, yet it's technology is, I mean it's foundational to
- 0:11pretty much every big shift happening in the digital world.
- 0:14It really is. We are peering inside Broadcom
- 0:17Inc ticker symbol AVGO. And we're looking at what you
- 0:21could call a, a profound structural anomaly here.
- 0:25I mean, think about it for a second, OK?
- 0:26How can one single company be totally indispensable to the,
- 0:31you know, the white hot AI revolution?
- 0:34Right, supplying chips to Google and Meta.
- 0:36All the big players, all of them.
- 0:38And at the exact same time, run its huge software division with
- 0:41the, let's say, cold, ruthless efficiency of a private equity
- 0:45firm. That's the perfect way to put
- 0:47it. This is the dive into the
- 0:48Convergent Titan and our mission today is really to unpack
- 0:51Broadcom's unique operational model.
- 0:53We need to get into the mechanics behind this
- 0:56convergence of, you know, bleeding edge silicon on one
- 0:59hand and legacy software on the other.
- 1:01And I think most importantly, figure out if it's trajectory
- 1:06justifies this massive, massive RE rating of its valuation.
- 1:11A RE rating that's just been astronomical.
- 1:13It's gone from a typical sort of industrial tech multiple to a
- 1:17straight up AI growth multiple. It's a whole different ball game
- 1:21now. And the whole thesis really for
- 1:23Broadcom is powered by what we're calling the dual flywheel.
- 1:27The dual flywheel, yeah. You have these two engines, they
- 1:30operate almost independently, but they funnel all that cash
- 1:33flow right back to the corporate center.
- 1:35So what's the first one? The first flywheel is the reason
- 1:38for that high valuation premium, the AI infrastructure super
- 1:41cycle. Broadcom isn't just a player
- 1:43here. They are, you could argue, the
- 1:46second most critical enabler of the entire AI boom, right behind
- 1:49NVIDIA. Wow.
- 1:50OK, so they're the sovereign power in custom silicon and high
- 1:53speed networking. I got it.
- 1:54And the second flywheel. That's the software Capital
- 1:56engine. This is where that private
- 1:58equity playbook really kicks in. The roof, The sufficiency part.
- 2:01Exactly. They go out and buy these
- 2:03mission critical software franchises, you know, LSI,
- 2:07Brocade, CA, Symantec, and now the huge one VM Ware.
- 2:12And they just apply this surgical focus on cost.
- 2:15Absolutely surgical. The goal is simple.
- 2:18Maximize cash and generate these just staggering operating
- 2:23margins. I mean, they routinely get over
- 2:25that 60% mark. 60%, that's incredible.
- 2:27And the numbers, I mean the financials really confirm the
- 2:30scale here. For fiscal year 2024, Broadcom
- 2:33reported $51.6 billion in revenue.
- 2:35Which is a what, 44% year over year?
- 2:37A phenomenal 44%, yeah, mostly because of VM Ware being
- 2:40integrated. But here's the number that you,
- 2:42the listener, really need to hone in on.
- 2:44They delivered $19.4 billion in free cash flow FCF and they
- 2:49maintained a 38% FCF margin while they were shouldering all
- 2:52these massive integration costs. That kind of cash generation is
- 2:56just it's, it's almost unheard of for a company going through
- 2:59that kind of transformation. It is FCF as we always say, it's
- 3:01the truth metric in these big M&A deals.
- 3:04But this dominance, it comes with some serious tension, some
- 3:07real peril. It does.
- 3:09The company is now trading at a forward price to earnings ratio
- 3:12of over 30. And for context, historically,
- 3:15before this whole AI boom, Broadcom was priced like a, you
- 3:19know, a boring industrial company.
- 3:21Right around 15 times PE. That was the norm.
- 3:24So that premium, that 30X multiple, it is explicitly
- 3:27pricing in what flawless execution?
- 3:30Absolutely flawless execution. It implies the market is
- 3:33completely certain they're going to hit their their really
- 3:36audacious AI targets. They're projecting something
- 3:38like $50 billion in AI revenue by 2026 or 2027.
- 3:43And we have to look at the other side of that efficiency coin.
- 3:46The other big tension point is the the customer fallout, the
- 3:50potential churn from this really aggressive monetization strategy
- 3:54they're using on VM Ware's customers.
- 3:56Broadcom's making a huge multi billion dollar bet that
- 3:59operational lock in that the sheer cost of migration is just
- 4:02powerful enough to completely override how their customers
- 4:05feel. It's a fascinating high stakes
- 4:07bet. OK, let's unpack this and I
- 4:09think we have to start with the man who built it, CEO Hoktan and
- 4:13his his unique private equity playbook.
- 4:16What exactly is this operational doctrine that's turned Broadcom
- 4:21into this, what, $700 billion powerhouse?
- 4:24The whole thing, this whole structural anomaly, we're
- 4:26talking about the mix of cutting edge tech and legacy software
- 4:30extraction. It is entirely the creation of
- 4:33CEO Hock Tan. Ever since he took over the
- 4:36predecessor company, Avago way back in 2006, the strategy has
- 4:39been just relentlessly consistent, predictable
- 4:43discipline. Broadcom operates less like a
- 4:45typical tech company. Much less they're not focused on
- 4:48some 10 year tech road map or slow organic growth.
- 4:51It's much more like a systematic, financially
- 4:54engineered roll up. So that doctrine really dictates
- 4:56everything, what they buy, how much they pay for.
- 4:58It and crucially what happens in the 1st 90 days after a deal
- 5:01closes. It's also really important to
- 5:03understand what they don't buy. OK, what don't they buy?
- 5:05Broadcom fundamentally avoids the high growth, high risk
- 5:09moonshots. You know the of R&D stuff, they
- 5:12are not interested in the next startup that might be a huge
- 5:15success or a huge failure. They want predictability.
- 5:18Precisely, they only target established mature mission
- 5:22critical franchise assets and these targets have to have three
- 5:26qualities non negotiable, which are high switching costs and
- 5:30entrenched market position and critically very low risk of
- 5:34being disrupted. And when you look at their
- 5:35shopping list, LSI, Brocade CA, Symantec VM Ware, they all fit
- 5:41that profile perfectly. Every single one.
- 5:43The market often saw these companies, especially the
- 5:45software ones. As you know, legacy assets slow
- 5:48growth. Right, because they weren't as
- 5:49sexy as the cloud companies. Exactly.
- 5:51But Tan saw something different. He saw these assets with
- 5:54incredible protected revenue streams that were, in his view,
- 5:58just mismanaged. Played by bloated R&D and too
- 6:00much sales and marketing spend. You got it.
- 6:02He saw cash cows just waiting to be optimized.
- 6:06And that optimization happens through this operational surgery
- 6:08playbook, a three-step process. A meticulous 3 step process that
- 6:13starts, I mean the minute the ink is dry on the deal.
- 6:16So step one is swift and decisive divestiture.
- 6:20Right away they immediately sell off anything that's not core,
- 6:23anything that doesn't contribute to that high margin sticky
- 6:26revenue. Like with Symantec, they sold
- 6:29the consumer business with VM Ware.
- 6:31They got rid of the end user computing and carbon black
- 6:33units. Immediately, it's all about
- 6:35ruthlessly narrowing the focus to maximize profit and just kill
- 6:39the overhead. OK, then.
- 6:40And Step 2 is the one they're famous for cost rationalization.
- 6:44This is where that private equity mindset really, really
- 6:46shines. They just slash selling general
- 6:49and administrative expenses, S&G to the bone, to the absolute
- 6:53bare minimum. I mean, reports suggest they cut
- 6:55it down to what's needed to service only their top
- 6:58customers, the top 500 or 2000 global accounts.
- 7:01So if you're a smaller customer, a long tail customer, your
- 7:04support just vanishes. It evaporates overnight.
- 7:07Your relationship with the company is gone.
- 7:09That must be a brutal culture shock for the company they
- 7:12acquire. I mean are there stories about
- 7:14what that feels like when Hawk Tans team comes in?
- 7:18Oh, absolutely. And the reports are all
- 7:20consistent. The culture shift is immediate.
- 7:22It's jarring. I can imagine.
- 7:24For a company like VM Ware, which was built on this idea of
- 7:27wide distribution and high touch engineering support, people
- 7:31described it as a clinical amputation.
- 7:35Entire layers of middle management, non essential R&D
- 7:38staff, they're just eliminated, sometimes in weeks.
- 7:41The message is crystal clear. If your project isn't generating
- 7:44measurable near term revenue, it's gone.
- 7:46And R&D gets refocused. Strictly, it's refocused on
- 7:49protecting the core mode, just making sure the current products
- 7:52stay best in class. They get rid of what Broadcom
- 7:55internally calls Science Project Research.
- 7:57So invest just enough to keep the franchise, but not a penny
- 8:00more on speculation. That's the playbook.
- 8:02Which brings us to Step 3, which is leveraging that new cost
- 8:05structure ricing power assertion.
- 8:08Exactly. Once they've confirmed the
- 8:10operational lock in those high switching costs, they use that
- 8:14leverage to impose really aggressive price hikes and move
- 8:17customers to higher margin contracts.
- 8:19And the VMware integration is the ultimate test of this entire
- 8:23doctrine, isn't? It it's the multibillion dollar
- 8:25final exam, they didn't just buy VM Ware, they bought the control
- 8:29point for the entire private cloud infrastructure of the
- 8:32world's biggest companies. And the financial goals they set
- 8:34were just stunning. Broadcom said they could almost
- 8:37double VM Ware's EBITDA from about $4.5 billion to what, $8.5
- 8:42billion in three years? An incredible target.
- 8:45That kind of margin expansion requires, well, surgical
- 8:48precision. So how are they doing?
- 8:50The early results from Q3 2025 show they're running ahead of
- 8:54schedule. The infrastructure software
- 8:55gross margins have already shot up to an elite 93%. 93% you only
- 9:00get there through that radical cost cutting and forced
- 9:03bundling. That's the only way.
- 9:04And their strategy is based on that aggressive segmentation we
- 9:07talked about. They made an explicit pivot to
- 9:09focus only on the top 2000 enterprise customers.
- 9:12Period. So this isn't just a preference,
- 9:14it's a systematic abandonment of everyone else.
- 9:17Small and medium businesses are just out.
- 9:20Out the termination of all those reseller agreements for smaller
- 9:24accounts was the proof. They did the math and figured
- 9:27out the cost of supporting hundreds of thousands of small
- 9:30businesses just dragged down the overall profit margin.
- 9:33So focus only on the big inelastic customers to maximize
- 9:37profit per employee per support ticket for everything.
- 9:41It's brutal, but from a purely financial perspective, it's
- 9:45undeniably effective and. What's so fascinating here is
- 9:48how this incredibly profitable software engine, the cash
- 9:51machine, is what funds the innovation and drives this huge
- 9:55market re rating in their other business.
- 9:57The semiconductor segment, it's a perfect synergy.
- 9:59Absolutely. The semi division is the
- 10:01innovation engine. It's the reason the stock's
- 10:03valuation is spiked. If the software story is about
- 10:05financial mechanics, the semiconductor story is 1 of
- 10:08technological necessity and pure market dominance.
- 10:11Right. And it's still the majority of
- 10:12the business, 58% of total revenue, yeah, or about $30.1
- 10:15billion in fiscal 2024. This is where Broadcom really
- 10:19becomes that second sovereign of AI infrastructure, just behind
- 10:23NVIDIA. And they have a deep Moat in two
- 10:26really key areas, yeah, custom silicon or Asic's and the high
- 10:30speed networking fabric that makes AI clusters work OK.
- 10:34Let's start with the custom silicon, the ASIC hegemon.
- 10:37NVIDIA cornered the market on general purpose GPU's for AI
- 10:40training, but Broadcom commands something like a 70% market
- 10:44share in custom ASICS. It's a.
- 10:46Quiet mode, but it's an incredibly defensible 1, and to
- 10:50really get it, you have to understand the basic problem
- 10:52that the hyperscalers are facing.
- 10:54Google, Meta, all of them. They're running billions of very
- 10:57specific, very repetitive AI operations every single.
- 11:01Day billions, search ranking, content recommendations, image
- 11:04tagging, Real time inference. All of it and general purpose
- 11:07GPU's are powerful, obviously, but they're incredibly expensive
- 11:11both to buy and to operate because of the power
- 11:13consumption. Plus you can't always get enough
- 11:15of them from NVIDIA. Exactly.
- 11:16So for the biggest players the driving factor becomes TCO total
- 11:19cost of ownership for a huge, predictable, very specific
- 11:23workload. Spending hundreds of millions
- 11:26upfront to Co design a custom ASIC with Broadcom pays off
- 11:30massively over a five year period.
- 11:32Because that custom chip is way more efficient and cheaper to
- 11:36run for its one dedicated task than a generalized GPU.
- 11:39Infinitely more efficient. That focus on TCO is why
- 11:43Broadcom's ASIC business is absolutely soaring.
- 11:47And Broadcom isn't just a vendor in this relationship, they're a
- 11:50Co design partner. I mean, the biggest one is with
- 11:52Google, right? The tensor processing unit, the
- 11:54TPU. That's the most critical
- 11:56partnership. The TPU is the silicon heart of
- 11:59Google's entire AI infrastructure, and Broadcom
- 12:02provides the absolutely critical proprietary IP blocks that make
- 12:06it possible. What kind of blocks are we
- 12:08talking about? We're talking about specialized
- 12:09SRTS, that's the serializer, D serializer for high speed data
- 12:12movement memory controllers and the advanced packaging
- 12:15specifically Co S chip on wafer on substrate packaging.
- 12:18Let's stick with the SRTS for a second, because this is a
- 12:20technical what most people don't really appreciate.
- 12:23They don't. The Certus is what dictates how
- 12:25fast data can physically move, both inside a chip and between
- 12:29chips. And if you can't move the data
- 12:31fast enough, your whole AI cluster just stalls.
- 12:33It doesn't matter how powerful the compute engine is, the data
- 12:36has to move and Broadcom has a multi decade lead in 112G and
- 12:40224G Certus tech. And that IP is incredibly hard
- 12:45to replicate. Billions in R&D and decades of
- 12:48expertise. So when Google or Meta designs a
- 12:50new custom chip, I have to go to Broadcom for that high speed
- 12:53data layer. It's not optional.
- 12:54So as Google scales up its Gemini models, that directly
- 12:58benefits Broadcom. Directly through IP licensing
- 13:01and manufacturing revenue. And the big valuation shift in
- 13:042024 was really driven by the market realizing this ASIC
- 13:08dominance wasn't just about Google anymore.
- 13:10That's the third customer story. We already knew they were
- 13:12ramping up production for Meta's custom AI chip, the MTA.
- 13:16That's already a multibillion dollar business, right?
- 13:18But the reports in late 2025 confirming this massive new
- 13:22engagement with a third undisclosed customer.
- 13:25Widely rumored to be open AI. Widely reported to be open AI,
- 13:28Yeah, for a custom inference chip, that was the real
- 13:31catalyst. And that single deal is
- 13:33projected to add, what, another $10 billion in revenue?
- 13:36Over its life cycle, yes, and it fundamentally validates the
- 13:39market's theory about bifurcation.
- 13:41Meaning the world needs NVIDIA for general purpose training.
- 13:45But it desperately needs Broadcom for hyperfilant custom
- 13:48inference, the part that actually scales to billions of
- 13:51users every day. It confirms Broadcom is a must
- 13:54have supplier. OK, so let's pivot to the second
- 13:57pillar of their AI dominance networking.
- 14:01If the ASIC is the brain, the network is the nervous system.
- 14:05A perfect analogy because the main bottleneck in modern LLM
- 14:08training isn't compute power of 1 chip, it's the speed at which
- 14:12thousands of chips can talk to each other, what engineers call
- 14:15East West traffic. And Broadcom basically has a
- 14:17monopoly on this switching and routing silicon for these
- 14:20clusters. They do.
- 14:21Their flagship products are the industry standard.
- 14:24Specifically you have to look at the Tomahawk series.
- 14:26The Tomahawk five at 51.2 terabits per second is the
- 14:30fundamental building block. It's the chip that makes sure
- 14:32the massive data flows for LLM training don't just choke the
- 14:36whole system. Exactly.
- 14:37And then for routing, they have products like the Jericho Three
- 14:40AI. That chip is specifically
- 14:43engineered for AI scale up needs.
- 14:45It has perfect load balancing across thousands of ports.
- 14:48So no congestion, no latency spikes which are fatal for
- 14:52training. Fatal.
- 14:53Which brings us to this key strategic battle that's
- 14:56happening right now. Optimized Ethernet versus
- 15:00Nvidia's proprietary InfiniBand. Right, NVIDIA pushes InfiniBand
- 15:04saying it's low latency is essential.
- 15:05For training and Broadcom is leading the counter attack
- 15:08through the Ultra Ethernet consortium, the UEC.
- 15:11Their argument being that optimized Ethernet can basically
- 15:14match Infiniband's performance. But with way better economics,
- 15:18better interoperability with different vendors and more
- 15:20flexibility. And is this just marketing or is
- 15:23there a real technical architecture behind?
- 15:25It oh, it's very real. The UEC is proposing new
- 15:28congestion control mechanisms, specialized data protocols, all
- 15:31built on top of Broadcom's Tomahawk switches.
- 15:34So the strategic goal is to keep the entire AI network layer and
- 15:38open market running on their chips instead of a closed
- 15:41proprietary NVIDIA ecosystem. You've got it and the market
- 15:45trends are showing that the hyper scalers are increasingly
- 15:48choosing Ethernet to avoid that vendor lock in and high cost of
- 15:52InfiniBand. It's a huge tailwind for
- 15:54Broadcom. And finally, as we move past
- 15:55400G and 800G networking, copper cables just can't keep up.
- 16:00The future is optical. The future is absolutely
- 16:03optical, which brings us to Broadcom's leadership in digital
- 16:06signal processors, or DSPS. These are the complex chips
- 16:10inside the optical transceivers. They convert the electrical
- 16:13signals to light and back. Again, and the move to 800 G and
- 16:161.6 T networking is driving exponential demand for these
- 16:20high margin DSPS. It's a vertically integrated
- 16:22Moat. So Broadcom sells the switch the
- 16:25Tomahawk. The router, The Jericho.
- 16:27And the critical optical component, the DSP to connect it
- 16:30all. They're selling the entire
- 16:31nervous system. Right.
- 16:32And we can't forget the cash cows that anchor the whole
- 16:35business, wireless and broadband.
- 16:37Apple is still Broadcom's single biggest customer, about 20% of
- 16:40their revenue historically right they.
- 16:42Supply the FBIR filters and the Wi-Fi Bluetooth chips for the
- 16:45iPhone, and for years the market has been terrified of this
- 16:51replacement risk. Apple's Project Everest to
- 16:54design those chips in house. But the FBR Moat has been
- 16:57incredibly hard for them to cross.
- 16:59Incredibly resilient, the manufacturing requires this
- 17:02extreme precision in thin film deposition and material science
- 17:07that's just proven much harder to replicate than anyone
- 17:10thought. So that led to the big D risking
- 17:12event in 2023 when they signed that multi year extension with
- 17:15Apple. Exactly.
- 17:17That secured this high margin revenue stream through at least
- 17:192026 or 2027. It basically neutralized the
- 17:23single biggest risk in their semiconductor business for the
- 17:27near future. And broadband is just a stable
- 17:29duopoly providing predictable cash flow.
- 17:31A stable cash cow that helps them pay down debt while the AI
- 17:34engine really Revs up. OK, here's where it gets really
- 17:37interesting for me, how they're generating these huge software
- 17:40profit margins while at the same time creating just enormous
- 17:44customer outrage. It's the VM Ware paradox.
- 17:46It's the ultimate tension between financial engineering
- 17:49and customer happiness. The whole $69 billion VM Ware
- 17:53acquisition was built on a a counter narrative, the belief in
- 17:57the private cloud as a durable long term.
- 18:00Thing, right? A thesis that really pushes back
- 18:02against the idea that everything is just going to end up on AWS
- 18:06or Azure. And for big companies, the logic
- 18:08is pretty compelling. Not all workloads can move to
- 18:11the public cloud. You've got cost volatility,
- 18:13latency issues for critical apps, and all these data
- 18:16sovereignty rules, especially in finance or healthcare.
- 18:18Exactly. And VM Ware Cloud Foundation VCF
- 18:21is the standard on premise operating system for running
- 18:24that kind of secure internal cloud.
- 18:27So what was Broadcom's strategy after the acquisition?
- 18:30A forced and very aggressive transition of the business
- 18:32model. Step 1 was just killing
- 18:34perpetual licenses, the old buy it once CapEx model, everything
- 18:38is now subscription an OpEx model.
- 18:40And at the same time, they're grabbing way more wallet share
- 18:42by forcing customers to bundle products.
- 18:44You got it. They take things that used to be
- 18:47separate if esphere, VSN, NSX and force customers to buy them
- 18:50all together in one big expensive VCF skewers.
- 18:54And subscription revenue is higher margin, more predictable
- 18:57and feeds that massive FCF generation.
- 19:00Which brings us right to the paradox.
- 19:02The huge customer outrage versus the cold, hard financial reality
- 19:07of lock in. And let's start with the
- 19:09outrage, because the stories are just wild.
- 19:12They are. Industry forums are filled with
- 19:13people talking about renewal quotes going up 304 hundred,
- 19:17even 500%. Simply because they're now
- 19:19forced to buy a bundle with products they don't need or
- 19:22want. And a small and mid market
- 19:23customers have been, I mean completely abandoned.
- 19:26Systematically, they terminated thousands of smaller reseller
- 19:29agreements, leaving 10s of thousands of organizations just
- 19:32orphaned. I get the financial logic, you
- 19:34know, cutting low margin support, but at what point does
- 19:37a 300% price hike make a CIO say, OK, that's it, we're
- 19:41leaving? When does the lock in start to
- 19:42crack? That is the critical risk.
- 19:44The Broadcom's bet, which has been correct so far, is on the
- 19:48sheer inertia of their biggest accounts.
- 19:50Churn among those top 2000 enterprise customers is still
- 19:54remarkably low. They just correctly calculated
- 19:57that for a giant bank or a hospital, the operational risk
- 20:00of trying to RIP out and replace their core virtualization layer
- 20:03is just it's catastrophic. The migration cost, both in
- 20:08dollars and an operational disruption, is higher than the
- 20:11outrageous price hike. It's incredible.
- 20:13It doesn't matter if the new price is 500% higher.
- 20:16The alternative, a multi year multi $1,000,000 project that
- 20:20could paralyze your business, is just too scary.
- 20:24And reports show that about 70% of those large accounts have
- 20:27already transitioned grudgingly to the new VCF contracts.
- 20:30So the playbook worked. The bet on lock in paid off.
- 20:34It's a stunning validation of the strategy.
- 20:36But let's talk about the alternatives for a minute.
- 20:38For that mid market that they abandoned, what's gaining
- 20:41traction there? Could a competitor build up
- 20:43enough scale there to eventually challenge the enterprise core?
- 20:46That's certainly the hope of the competitors in that mid market.
- 20:49We are seeing more adoption of things like Nutanix which is a
- 20:52hyper converged platform or even open source solutions like KVM
- 20:56or Proxmox. Are they technically mature
- 20:59enough? For smaller environments, yes,
- 21:01but the challenge is always enterprise compatibility.
- 21:04They're really complex features. Right.
- 21:05Advanced site recovery, deep integrations with legacy
- 21:09systems, specific security certifications.
- 21:12The open source stuff often can't provide that with the same
- 21:15level of guaranteed support. So until a competitor can offer
- 21:18a seamless low risk migration path for all those features, the
- 21:22Enterprise core stays locked in. Got it.
- 21:25And to round out the software engine, the legacy stuff, CA,
- 21:27mainframe, Symantec, they're just run for cash basically.
- 21:31Pure cash flow generation, minimal reinvestment, low single
- 21:34digit growth, but those best in class 60% plus operating
- 21:39margins. They are the funding mechanism
- 21:41that lets Broadcom pay down all that M&A debt.
- 21:44So if we connect this to the bigger picture, Broadcom's key
- 21:47financial advantage isn't just the revenue they generate, it's
- 21:50the massive structural cost they systematically avoid.
- 21:53That's the key. That's the whole financial
- 21:55architecture. Let's dig into that.
- 21:56Broadcom's growth is all about these big step functions from
- 21:59M&A, not slow organic growth. Revenue has gone from what, $6.8
- 22:04billion in 2015 to a projected $63.4 billion in 2025?
- 22:09It's an aggressive expansion and it needs an equally efficient
- 22:12financial structure to support it.
- 22:14And that efficiency is all measured in free cash flow FCF.
- 22:17Their FCF margin has historically been elite, around
- 22:2049%, and even during the chaos of the VM Ware integration it
- 22:24only dipped to 38%. Generating almost $20 billion in
- 22:28cash during a peak integration year.
- 22:31That tells you everything about the quality and stickiness of
- 22:33their revenue. And that FCF is the lifeblood.
- 22:36It supports the dividend, the debt pay down and the huge share
- 22:39buybacks. So what's the critical engine
- 22:41driving that cash conversion? The fabulous advantage.
- 22:44This is so important to understand why they are
- 22:46structurally better at this than their peers.
- 22:48Broadcom outsources 100% of its chip manufacturing.
- 22:52Mostly to TSMC. Almost entirely.
- 22:54And because they don't own, build or maintain these insanely
- 22:58expensive semiconductor fabs, their capital expenditures,
- 23:01their CapEx are negligible. We're talking two 3% of revenue,
- 23:05like $150 million in 1/4 where they make almost $16 billion in
- 23:09sales. Exactly.
- 23:10Now contrast that with an integrated competitor like Intel
- 23:13and IDM, an integrated device manufacturer.
- 23:16Intel has to spend 10s of billions every year, 2030% of
- 23:20the revenue on CapEx just to keep up.
- 23:23Broadcom avoids that massive annual cash strain entirely.
- 23:26That 15 to $20 billion difference in CapEx for Broadcom
- 23:30converts to directly into free cash flow.
- 23:32It is a massive structural advantage.
- 23:34Which is why FCF, not GAAP net income is really the best way to
- 23:38value this company. Absolutely.
- 23:39And that focus is vital because of another tricky financial
- 23:42item, stock based compensation or SBC, right?
- 23:45Which jumped a lot after the VM Ware deal from about $2.2
- 23:49billion to $5.7 billion. Mostly from assuming VM Ware's
- 23:53equity and the retention grants needed to keep key engineers.
- 23:56And while they exclude that from their non GAAP numbers, SBC is a
- 23:59real cost to shareholders through dilution.
- 24:02Yeah, your slice of the pie gets smaller.
- 24:03So Broadcom's counter move is to use that massive SES stream for
- 24:07an aggressive programmatic share buyback program.
- 24:10It's been $4.2 billion on buybacks in just one quarter.
- 24:14Which effectively neutralizes the SBC dilution and keeps the
- 24:17share count stable over time. It's very disciplined.
- 24:19OK. Shifting to the balance sheet,
- 24:21let's talk that the VM Ware deal meant taking on a lot of
- 24:24leverage. Total debt balloon to around $66
- 24:27billion. And their gross debt to EBITDA
- 24:30ratio spiked to about 3.5 times, which is high, but the market
- 24:35was OK with it because those software cash flows are so
- 24:37predictable. And the plan is to aggressively
- 24:39pay that down. Very aggressive, yeah.
- 24:41Use the FCF to get that ratio below 2.5 X within two years.
- 24:45They have to do that to maintain their investment grade credit
- 24:48rating. OK.
- 24:48And finally, the big consequence of the Hock Tam playbook is a
- 24:52balance sheet just dominated by intangible assets.
- 24:55It is. As of 2024, the company carries
- 24:57$138.5 billion in goodwill and another $45 billion in
- 25:03intangible assets. And that goodwill is just the
- 25:06premium they paid over the book value of the companies they
- 25:08bought. Which is the big structural risk
- 25:10of the whole roll up strategy if the VM Ware integration fails to
- 25:14hit its targets or if the Apple deal suddenly falls apart.
- 25:16They could be forced to take massive non cash impairment
- 25:19charges. And while that wouldn't hit cash
- 25:21flow immediately, it would hammer their gap earnings and
- 25:24more importantly, it would signal to the market that they
- 25:26overpaid. It's the shadow hanging over all
- 25:30the success. OK, let's.
- 25:31Pull all these threads together now and talk about valuation.
- 25:34Broadcom spent years trading at a conglomerate discount 12 to 15
- 25:40times forward PE. Right, because the market didn't
- 25:42know what to make of it. Mainframe software, Wi-Fi chips,
- 25:46AI, networking. It was a weird.
- 25:48Mix, but the AI boom has completely shattered that
- 25:50discount. Shattered it.
- 25:52The market has re rated Broadcom from a complex value stock to an
- 25:56AI growth enabler. They now trade at a forward PE
- 25:59near 35 times. Putting them right in the same
- 26:02group as high growth players like Marvel.
- 26:04Though still a bit behind NVIDIA of course, but this shift it
- 26:08requires a lot of conviction in their future cash flow.
- 26:10So what does that price actually imply?
- 26:12I mean when analysts run ADCF at discounted cash flow model, what
- 26:16does the math say? The math says that to justify
- 26:18today's market cap, Broadcom has to grow.
- 26:20Its free cash flow at AI Mean a staggering CEJR of 15% to 20%
- 26:25for the next five years. And that growth has to come from
- 26:28perfect execution on all fronts. All fronts, the AI segment has
- 26:32to hit that $50 billion revenue target and the VM Ware margins
- 26:35have to hold up without major churn.
- 26:37It's a very, very ambitious forecast.
- 26:39So let's breakdown the scenarios analysts are using to model that
- 26:42What's the path to a trillion dollars?
- 26:44Or what's the risk of a big correction?
- 26:46OK, so you have scenario A, the AI bull case.
- 26:49Analysts give this about a 40% probability.
- 26:52This is the flawless execution scenario.
- 26:54AI revenue hits $50 billion by 2026, driven by Google, Meta and
- 26:59the Open AI deal, and VMware lock in proves to be absolute.
- 27:04In that case, the outcome is just exponential.
- 27:06Broadcom could be generating over $60 billion in free cash
- 27:10flow by 2027. And if you put a 30 XFCF
- 27:13multiple on that. You are looking at a market cap
- 27:15of 1.8 to $2.0 trillion. That's the path.
- 27:18Wow. OK, then there's the base case.
- 27:20Scenario B, the base case also about a 40% probability here.
- 27:24AI revenue grows strongly but maybe stabilizes closer to $35
- 27:27billion. CapEx cools a little, VM where
- 27:29it works, but they see some churn in that mid market they
- 27:32abandoned. So in that case, FC ETF grows
- 27:34steadily to maybe $45 billion. The stock still compounds
- 27:39nicely, maybe 1012% total annual returns from dividends and
- 27:42buybacks. It's a solid, high quality
- 27:45return, but it doesn't really justify the dramatic premium
- 27:48that's priced in today. And finally, the bear case.
- 27:51Scenario C The bear case maybe a 20% probability.
- 27:56This is where multiple things go wrong at once.
- 27:58Hyperscalers cut CapEx. Apple finally succeeds with
- 28:01Project Everest and brings its chips in house.
- 28:04And critically, a mass migration of VM Ware customers to a real
- 28:07alternative like Nutanix or a mature open source stack.
- 28:11In that worst case scenario. AI revenue stalls over our
- 28:13revenue contracts because they lose Apple and the market
- 28:16multiple just violently compresses back to its
- 28:18historical 15X. That would mean a big
- 28:20correction, 3040% down from the highs.
- 28:23Before we wrap, let's touch on governance and Insider
- 28:26alignment. Hock Tan's pay structure is
- 28:28pretty unique. It's a master class in
- 28:29alignment. His big 2023 grant was a
- 28:32performance stock unit award that vests only if the stock
- 28:35price hits some formidable targets over five years.
- 28:37And he gets no annual cash bonus during that time.
- 28:40It's an all in package. It forces him to deliver that
- 28:43bull case scenario to get paid. His interests are completely
- 28:47aligned with shareholders. What about the insider selling
- 28:50we've seen? There has been notable selling
- 28:52by Tan and others in 2024 and 2025.
- 28:56Now some of that is just necessary for taxes or
- 28:58diversification. After decades, TAN still holds
- 29:01over 1,000,000 shares, but the consistent selling at record
- 29:04highs does suggest they're prudently taking some chips off
- 29:07the table, recognizing that the valuation is at historic peaks.
- 29:12So beyond the valuation models, what are the big structural
- 29:15existential risks that could just derail this whole thing, no
- 29:18matter how well Hock Tan executes?
- 29:20I see three big ones. First is the Apple concentration
- 29:23risk Project Everest. Despite the extension, Apple's
- 29:27long term goal to in house those chips is still the single
- 29:30biggest specific threat. It could wipe out eight to $10
- 29:32billion in high margin revenue overnight.
- 29:35OK, what's #2? Pure geopolitical vulnerability.
- 29:38Broadcom is 100% fabulous. They are existentially reliant
- 29:41on TSMC in Taiwan for all their advanced AI chips.
- 29:45The three nanometer nose, the COAS packaging, all of it.
- 29:48All of it. Any conflict in the Taiwan
- 29:51Strait would instantaneously halt their ability to ship
- 29:53product. There is no alternative foundry
- 29:56capacity anywhere in the world that can handle their needs
- 29:59right now. It is a massive single point of
- 30:01failure. A risk they have no direct
- 30:04control over. None.
- 30:05And finally, regulatory headwinds.
- 30:08The aggressive bundling and pricing of VM Ware has already
- 30:11drawn intense scrutiny from the EU and the FTC.
- 30:15So while the deal is done, they could face huge fines or be
- 30:18forced to change their behavior in the future.
- 30:20And critically, that regulatory scrutiny probably means they're
- 30:23effectively barred from anymore massive M&A.
- 30:26The Hoktan playbook of growth by acquisition is likely over.
- 30:30They have to grow with what they have.
- 30:32So what does this all mean for you, the listener?
- 30:34Broadcom is this masterpiece of strategic and financial
- 30:37engineering. They are perfectly positioned to
- 30:39capture value from both the AI CapEx cycle and this captive
- 30:43enterprise software base. They're benefiting from a true
- 30:46perfect storm of tailwinds, and the cash flow generation is just
- 30:50phenomenal. It's a financial truth backed by
- 30:52real contracts and tech leadership.
- 30:55That $50 billion AI revenue target isn't just a dream.
- 30:58It's based on concrete design wins.
- 31:01However, the take away here is that the easy money, the money
- 31:05from that initial valuation RE rating has probably already been
- 31:08made. The market has priced in the
- 31:11near flawless execution of that bull case.
- 31:13So investors today are making 2 core bets, first on the long
- 31:17term durability of this AI super cycle, and 2nd on Broadcom's
- 31:21ability to maintain its duopoly status with NVIDIA while
- 31:25navigating these huge geopolitical risks.
- 31:27You're basically betting that high switching costs can
- 31:29continue to override customer anger, a financial calculation
- 31:33that has worked perfectly so far.
- 31:34Exactly. And for listeners who like to
- 31:36monitor the total market temperature, you know,
- 31:38evaluating if companies are hitting the growth rates needed
- 31:40to justify these valuations, tracking that gap between cash
- 31:43flow and valuation multiples is absolutely key.
- 31:46Which brings us to our final provocative thought for today,
- 31:49Broadcom build a software empire on proprietary lock in, where
- 31:53the risk of leaving is higher than the price hike.
- 31:56But as the world and even Broadcom's own hardware division
- 31:59champions open, interoperable standards like the Ultra
- 32:02Ethernet Consortium, how long can their reliance on captive
- 32:06pricing models really last before a technically feasible,
- 32:10economically superior alternative finally emerges for
- 32:14their biggest, most profitable customers?
- 32:16Something definitely worth pondering is that battle between
- 32:18proprietary systems and open standards continues to shape the
- 32:21future of the cloud. Thanks for joining us for this
- 32:24deep dive. We'll see you next time.