Latest / Deep Values: Value Investing Deep Dives / Broadcom's Q2 AI Growth and VMware Revolt
Transcript
- 0:00You know it is September 2026, and usually when a massively
- 0:05dominant company reports its quarterly earnings, the market
- 0:08reaction makes intuitive sense. Right, it follows a logic.
- 0:12Exactly. A tech giant post
- 0:14record-breaking numbers. The stock goes up, they miss
- 0:17expectations, it goes down. It's a very comforting logical
- 0:20equation. But today we are looking at
- 0:23something that completely breaks that logic.
- 0:25Completely breaks it. I mean, we are looking at a
- 0:28financial contradiction today. A major company just reported an
- 0:3286.1% revenue expansion for the quarter.
- 0:36Which is just a massive number. Right.
- 0:38They hit almost $30 billion in total revenue.
- 0:41And even Wilder, their artificial intelligence revenue
- 0:44specifically grew by 221% year over year.
- 0:47Yeah, by all traditional metrics, Wall Street should have
- 0:50thrown an absolute parade for this.
- 0:51But they didn't. Instead, the stock immediately
- 0:54dropped by roughly 6%. It really forces a double take.
- 0:57Yeah. I mean, you look at the
- 0:58financial terminals, you see triple digit growth in the
- 1:01single most important technology sector in the world right now.
- 1:04And the market essentially just shrugs and asks, well, what else
- 1:08is hiding under the hood? And that contradiction is
- 1:10exactly what we are dissecting in our deep dive.
- 1:12Today, we're taking a magnifying glass to Broadcom Inc, ticker
- 1:17symbol AVGO. And we have some incredible Rd.
- 1:20maps for this. We really do.
- 1:21Our sources today are two meticulously detailed, freshly
- 1:25published Q3 fiscal 2026 financial and valuation research
- 1:30reports. Yeah, we've got pages of data,
- 1:32DCF models, supply chain analysis, the works.
- 1:36So what's our actual mission for this deep dive?
- 1:38Well, the mission today is to completely strip away the AI
- 1:41hype. You know we're going to ignore
- 1:43the glossy headline numbers for a minute and look at the actual
- 1:46granular mechanics of this business.
- 1:49Because it's not just about the big numbers.
- 1:50Right, exactly. We need to understand the
- 1:52structural risks and the real world valuation of a company
- 1:56that is basically acting as the global toll road for the entire
- 2:00AI build out. The toll road, I like that.
- 2:02Yeah, because when you understand the physical plumbing
- 2:04of this system, you get a much clearer picture of where this
- 2:07technology is actually heading and more importantly, who is
- 2:10profiting from it. OK.
- 2:12Before we get into that plumbing, I have a special
- 2:14message specifically for you listening right now and I want
- 2:17to make sure I deliver this exact text verbatim.
- 2:19Go for it to access more advanced researches, follow
- 2:22Guru's latest movement short and leverage our latest Deep Values
- 2:26agents. Welcome to download Deep Values
- 2:29APP from App Store to track stock market temperature.
- 2:32Welcome to download iOS app Buffett Indicator from Apple App
- 2:36Store or Bubble Watch from Google Play.
- 2:39Links in the description. Disclaimer All analysis is based
- 2:42on publicly available sources and maybe subject to revisions
- 2:45or differing interpretations. All contents from this channel
- 2:49is for research and entertainment purposes only and
- 2:51does not constitute any investment or financial advice
- 2:54or service. Conduct your own due diligence
- 2:56before making investment financial decisions.
- 2:58And that's a vital reminder that we are exploring the mechanics
- 3:01in the research today, not dispensing portfolio
- 3:04instructions. Right, exactly.
- 3:06So OK, let's unpack this setup because we really need to bridge
- 3:09the gap between that weird market reaction in the intro and
- 3:12how Broadcom actually generated this staggering Q3 revenue.
- 3:16Yeah, we have to look at the business model.
- 3:18Let me try an analogy here to help visualize their role,
- 3:21because when you and I hear about the AI boom, our minds
- 3:25usually go straight to. NVIDIA sure, the GPU.
- 3:27'S right. So if NVIDIA builds the
- 3:29hyperfast billion dollar race cars, the graphics processing
- 3:33units that actually do the complex AI thinking, Broadcom is
- 3:37the company building the asphalt, the guardrails and the
- 3:39traffic lights. That is a great way to look at.
- 3:42It because when you have a million race cars trying to
- 3:44communicate at the exact same time, Broadcom builds the
- 3:48networking switches and optical connections that prevent a
- 3:50massive pile up. And what's fascinating here is
- 3:53that processing speed is completely irrelevant if the
- 3:56data cannot travel from 1:00 server to another instantly.
- 3:59Exactly. And Broadcom's dual pillar
- 4:01business model is engineered to capture that exact bottleneck.
- 4:05According to the Q3 reports, roughly 70.4% of their revenue
- 4:09comes from Semiconductor Solutions.
- 4:12So that's the hardware side. Right.
- 4:13That includes custom networking chips, the Tomahawk 5, the
- 4:17Jericho 3X switches, and then the remaining 29.7% comes from
- 4:22infrastructure software, which is heavily driven by their
- 4:25recent acquisitions. But the reports highlight the
- 4:28semiconductor side is the main growth engine, right?
- 4:30Like what do the specific numbers look like for their AI
- 4:33hardware division? Oh, the figures are monumental.
- 4:36I mean, AI specific semiconductor revenue alone hit
- 4:39$16.7 billion this quarter. Wow, yeah, which represents that
- 4:45221% year over year growth we mentioned and management's
- 4:49guidance projects Q4 will be even larger targeting 34.8
- 4:53billion in total consolidated revenue.
- 4:55It's. Just for 1/4, yes.
- 4:57But the projections that truly jump off the page are the future
- 4:59teasers. Management anticipates AI
- 5:02revenues reaching 115 billion in 2027 and 230 billion by 2028.
- 5:08230 billion. That's insane.
- 5:10It is, and a significant portion of this involves companies like
- 5:13Anthropic scaling up as their largest XPU customer.
- 5:16OK, let's pause there, because that term pops up a lot in the
- 5:18research XPU. What exactly does that
- 5:21encompass? It's essentially an umbrella
- 5:23term. As AI architecture evolves, we
- 5:26aren't just using central processing units or CPU's
- 5:29anymore. Right, the traditional computer
- 5:31brains. Exactly.
- 5:32Now we have GPU's for graphics, TPU's for tensor processing, all
- 5:36these customized chips. So XPU is just the industry
- 5:39shorthand for any specialized processing unit used in these
- 5:42massive data centers. Got it.
- 5:44OK. So how does a single company
- 5:47capture so much of that XPU ecosystem?
- 5:49We're talking about hundreds of billions of dollars in projected
- 5:52revenue. There must be a technological
- 5:54reason why competitors aren't just undercutting them.
- 5:56Right, and the answer lies in a physical engineering mode,
- 5:59specifically a technology called Co Packaged Optics or CPO.
- 6:03CPO. Yeah, to grasp why this matters,
- 6:06you have to picture the physical reality of a modern AI data
- 6:09center. We are no longer talking about a
- 6:11single closet with a few servers, right?
- 6:13Hyper scalers, You know, companies like Google, Meta,
- 6:16Microsoft, they are building sprawling AI clusters containing
- 6:19hundreds of thousands of XP us. I mean these are facilities that
- 6:22consume as much electricity is a medium sized city.
- 6:25Which is exactly the core problem.
- 6:27When you scale to that size, moving data over traditional
- 6:31copper electrical lines becomes a literal physics nightmare.
- 6:35Because of the heat. Because of the resistance, yeah.
- 6:37Copper has electrical resistance, so when you push
- 6:40massive amounts of data through it at high speeds, that
- 6:43resistance generates heat. Oh, so you are literally
- 6:46burdened electricity just to move information a few feet?
- 6:49Exactly. Even standard pluggable optical
- 6:52cables require too much power to convert the electrical signals
- 6:55into light and back again. And power is the ultimate
- 6:58currency right now. I mean, data centers are already
- 7:01maxing out local electrical grids all over the country.
- 7:04Which is why Broadcom's Co packaged optics is such a
- 7:07massive game changer. They figured out how to
- 7:10integrate photonic circuits literally using lasers to
- 7:13transmit data via light directly alongside the silicon chip on
- 7:17the exact same package. Wait, so the laser is right next
- 7:20to the chip? Yes, by removing the distance
- 7:23the electrical signal has to travel before becoming light,
- 7:26they reduce the power consumption of data transfer by
- 7:28up to 50% compared to traditional transceivers. 50%
- 7:32power savings in a facility using a GW of electricity.
- 7:36I mean, that isn't just a minor upgrade.
- 7:37That is billions of dollars saved, plus less strain on the
- 7:41public power grid. It's an engineering hurdle that
- 7:43takes years of research and billions in capital to clear.
- 7:47Competitors cannot simply replicate that overnight.
- 7:50It makes Broadcom basically the only viable solution to the
- 7:54power crisis facing these tech giants.
- 7:56OK. That definitely explains the
- 7:58cash flow numbers we saw in the reports.
- 8:00Yeah. What were the exact margins on
- 8:02that 30 billion in revenue? Broadcom generated $13.665
- 8:06billion in free cash flow just in Q3 alone.
- 8:09Just cash. Pure cash.
- 8:11That equates to a 46.2% margin. For every dollar in revenue that
- 8:16comes in the door, nearly $0.50 drops straight to the bottom
- 8:18line is unencumbered cash. Now, how do they physically
- 8:22achieve believe that? Because manufacturing advanced
- 8:25silicon is famously expensive. I mean, building a single modern
- 8:28microchip factory costs upward of $20 billion.
- 8:31Yeah, they achieve it by operating as a fabulous
- 8:33semiconductor company. Fabulous.
- 8:35Meaning no fabrication plans, right?
- 8:37They do not own or operate any factories themselves.
- 8:41They design the intricate blueprints and they hand those
- 8:44designs over to TSMC, the Taiwan Semiconductor Manufacturing
- 8:48Company, to physically print the silicon wafers and perform the
- 8:53advanced packaging. So they completely outsource the
- 8:56most capital intensive part of the entire industry.
- 8:59They let TSMC spend the 20 billion on the factory while
- 9:03they keep the 46% profit margins on the intellectual property.
- 9:07It's a brilliantly efficient model, but the true engine
- 9:10behind this cash generation, the thing that keeps the margins so
- 9:14consistently high across all their divisions, is CEO Hawk
- 9:18Tan's corporate integration playbook.
- 9:21Right, the Hawk Tan playbook. It's legendary on Wall Street,
- 9:24and it is notoriously ruthless. I want to push back on this
- 9:27playbook heavily actually, because reading the details and
- 9:30the sources felt alarming to me. How so?
- 9:32Well, when Broadcom acquires a mature technology franchise, the
- 9:35playbook is essentially gut the corporate overhead, fire the
- 9:39vast majority of the sales and marketing teams and drastically
- 9:42consolidate operations. That's the playbook, yeah.
- 9:44Now, I know the reports note they spent nearly $11 billion on
- 9:48R&D for 2025, but if you are slashing all the operational
- 9:53support and squeezing the acquired company to the bone,
- 9:56aren't you just eating the seed corn?
- 9:58You mean sacrificing the long term product for short term
- 10:01gains? Exactly.
- 10:03It feels like aggressive financial engineering designed
- 10:06to juice the stock price for executive bonuses rather than
- 10:10sustainable innovation. Doesn't the actual product
- 10:12eventually suffer? It's a totally valid skepticism,
- 10:15but if we connect this to the bigger picture, you have to look
- 10:18at their target audience to understand why it works without
- 10:21killing the product. Broadcom does not sell consumer
- 10:24electronics. Right, no retail stores.
- 10:26They aren't trying to convince small businesses or everyday
- 10:28people to buy their networking switches.
- 10:31Their entire direct sales model targets a hyper concentrated
- 10:34base, the world's largest cloud providers and the Fortune 500.
- 10:39So they basically don't need a multimillion dollar global
- 10:42advertising campaign. They don't need marketing at all
- 10:45when you are selling essential physically constrained
- 10:48infrastructure to Google or Amazon.
- 10:51You just need a direct phone line between your lead engineer
- 10:54and their head of data center architecture.
- 10:56Wow, OK. Yeah, by eliminating the
- 10:58traditional sales funnel, Hoktan can redirect those billions into
- 11:02pure research and development. They maintain the technological
- 11:05Moat while completely shutting the corporate bloat that
- 11:08normally plagues massive conglomerates.
- 11:10And that structural reality clearly resonates with top
- 11:13institutional shareholders because their reports mentioned
- 11:16giants like Vanguard and BlackRock overwhelmingly
- 11:19supporting this strategy. They support it because the
- 11:22returns are undeniable. Over a five year period,
- 11:24Broadcom has delivered a 1082% total stockholder return.
- 11:29Over 1000%. Yes.
- 11:31And a huge part of that is executive alignment CEO talk.
- 11:34Tan's 2025 compensation was reported at $205 million.
- 11:39Which I have to say, sounds completely absurd on the
- 11:41surface. It does until you look at the
- 11:43structure of it. 99.4% of that compensation is locked in
- 11:47performance stock units or PS US.
- 11:49Meaning he doesn't just get the cash.
- 11:51Right, these aren't just guaranteed shares.
- 11:53He only receives that value if massive multi year stock price
- 11:58hurdles are achieved. If the company's valuation
- 12:01stagnates, his compensation essentially evaporates.
- 12:04It forces management to think strictly as long term owners.
- 12:08IT owners love returning capital.
- 12:10Management has a disciplined history of aggressive stock
- 12:13buybacks here, which shrinks in the amount of shares available
- 12:16and makes the remaining ones more valuable and massively
- 12:19raising the dividend. Oh, the dividend growth is
- 12:20crazy. The sources show they bumped the
- 12:22dividend from a mere 19 cents a share in 2016 all the way up to
- 12:26$2.60 a share by 2026. Which is an incredible wealth
- 12:31compounding machine for early investors.
- 12:32But this brings us to the friction point.
- 12:35The structural crack. Exactly, if the cash flow is
- 12:37bulletproof, the technological mode is deep, and the M and A
- 12:40playbook is this successful, why did the Stark drop on those
- 12:45flawless Q3 earnings? Here is where it gets really
- 12:48interesting because the dark side of that ruthless
- 12:53optimization finally meets reality.
- 12:56The numbers look perfect on the spreadsheet, but when you look
- 12:58at the human element, how they are actually treating the
- 13:01customers, giving them that cash, a massive structural crack
- 13:05starts to show. And this centers around their
- 13:07$69 billion acquisition of VM Ware.
- 13:10Yes, let's talk about VM Ware. For those unfamiliar, VM Ware is
- 13:14the bedrock of modern enterprise IT.
- 13:17It is software that allows companies to run virtualization.
- 13:20Which basically means doing more with less hardware.
- 13:23Right. Instead of buying 10 different
- 13:24physical servers to run 10 different applications, VM Ware
- 13:27lets you run 10 virtual servers on a single physical machine.
- 13:30It maximizes hardware efficiency.
- 13:32Almost every major corporation uses it.
- 13:34It is deeply entangled in their infrastructure.
- 13:36Right, so Broadcom buys VM Ware and immediately implements the
- 13:40Hoktan playbook. They killed perpetual licenses.
- 13:44You can no longer just buy the software once and own it
- 13:46forever. It is now entirely subscription
- 13:48based. Which is standard industry
- 13:49practice now, to be fair, but they paired it with aggressive
- 13:52bundling. Aggressive is putting it mildly.
- 13:55They in first a 72 core minimum purchase per CPU.
- 13:59Let's make this tangible for you listening.
- 14:01Imagine you're an IT manager for a midsized hospital.
- 14:05You only need to run software on a 16 core or 32 core processor.
- 14:09Which is very common. Very common, but Broadcom comes
- 14:12in and says we don't care, you have to pay for a 72 core
- 14:15license anyway. Yeah, that hurts.
- 14:18The sources detail enterprise clients experiencing price
- 14:21increases ranging from 85% to over 500% virtually overnight.
- 14:26Pay the new ransom or your hospital system shut down.
- 14:29It really highlights the extreme double edged sword of high
- 14:32switching costs. Virtualization software is
- 14:34embedded into the very foundation of an enterprise's
- 14:37digital architecture. You can't just RIP it out.
- 14:39Disentangling your company from VM Ware takes years of planning,
- 14:42retraining, staff, accepting significant operational
- 14:45downtime. Broadcom knows this.
- 14:47They're leveraging that friction to exercise monopolistic pricing
- 14:50power today. But the customers are not just
- 14:52quietly accepting it, they are literally going to war over
- 14:55this. The Dutch Ministry of
- 14:57Infrastructure and Water Management actually took
- 15:00Broadcom to a Hague District Court to force them to support
- 15:04legacy licenses. And the government won.
- 15:06Wow. Yeah.
- 15:08And now you have corporate giants like AT&T, Siemens United
- 15:12Healthcare, They are all filing lawsuits over this coercive
- 15:16bundling. This is the risk of squeezing
- 15:18the lemon until the rind breaks. While it juices the immediate
- 15:22cash flow, it creates profound ill will.
- 15:25Absolutely. Enterprise IT managers have very
- 15:27long memories. When they feel held hostage,
- 15:30they will spend the next five years designing their
- 15:32architecture specifically to eliminate you from their stack.
- 15:35So. They're building an off ramp.
- 15:36Exactly. It opens the door for
- 15:38competitors like Microsoft Hyper V Nutanix or open source
- 15:41solutions like Proxmox to steal market share over the next
- 15:44decade. OK.
- 15:45So add that software friction to the short term hardware risks
- 15:49detailed in the reports, because we talked about how brilliant
- 15:52the fabulous model is relying on TSMC for manufacturing, but
- 15:56advanced chips require a specific type of packaging
- 15:59called call wise chip on wafer on substrate.
- 16:01Yeah, call was. It is the physical process of
- 16:04stacking the silicon dyes very closely together so they can
- 16:07communicate faster. Right.
- 16:09And TSMC has a finite global capacity to perform that call
- 16:13Wise packaging. Building new facilities to do it
- 16:15takes years. Broadcom is in a literal knife
- 16:18fight with NVIDIA and AMD for every single slot on TSMC's
- 16:22assembly line. Yeah, the supply chain
- 16:24bottleneck is real. If they can't physically get the
- 16:26chips packaged, they cannot realize that $230 billion
- 16:30revenue projection, regardless of how much demand Anthropic or
- 16:33Google has. And this tension is what makes
- 16:36valuing the company incredibly difficult.
- 16:38You have the hyper growth of AI hardware colliding with finite
- 16:41supply chains and a massive customer revolt on the software
- 16:44side. So what?
- 16:44Does this all mean for the numbers?
- 16:46How do we actually value value this?
- 16:48The stock is hovering around $369 a share.
- 16:52The research reports ran Broadcom through several
- 16:54different value investing frameworks and the results are
- 16:57fascinating because it completely fails traditional
- 17:00Benjamin Graham deep value screens.
- 17:02Yeah, for listeners who follow traditional value metrics,
- 17:05Broadcom's numbers are Nosebleed high.
- 17:08They have a price to earnings or PE ratio of 53.1.
- 17:12Just to set a baseline for that, a classic value investor might
- 17:15look for APE ratio around 15. You are paying 53 times the
- 17:19company's current earnings. And the price to book ratio is
- 17:2120.7, meaning the stock price is value in the company at over 20
- 17:25times what its actual net assets are worth on paper.
- 17:28Graham would look at those multiples and run the other way.
- 17:30He absolutely would. But if you apply Charlie
- 17:33Munger's framework, which looks less for deep bargains and more
- 17:36for high quality compounder businesses, it tells a different
- 17:39story. Broadcom boasts a return on
- 17:41invested capital of 23.2% and a gross margin of 66.5%.
- 17:46Those are fantastic metrics. Yeah, the overall scorecard in
- 17:49the report gave it 10 passes, 7 failures, and two partials.
- 17:54It is an undeniably exceptional business, but you are paying a
- 17:57massive premium for the privilege of owning it.
- 18:00To figure out what that premium actually translates to the
- 18:02reports, utilize discounted cash flow or DCF models.
- 18:06Right, which basically estimates all the cash a company will
- 18:09generate in the future and discounts it back to what that
- 18:12cash is worth today. And the spread of the results in
- 18:15these three DCF scenarios is just hilarious to me.
- 18:18Let's walk through them. What happens in case one?
- 18:20The no growth scenario if we assume the AI super cycle stops
- 18:24tomorrow and their revenue never grows another inch.
- 18:26Well, even in that absolute doomsday scenario, Broadcom's
- 18:30current margins are so robust that the massive cash they
- 18:33already generate puts a hard floor on the stock's intrinsic
- 18:36value at about $76 a share. OK, so 76 dollars is the
- 18:39absolute basement. What about case 2, the
- 18:41conservative scenario? This model assumes they only
- 18:43grow at 7.39% a year for the next decade.
- 18:47We use that specific number because it was their absolute
- 18:49lowest historical growth rate over the last 10 years.
- 18:52OK, so sluggish pace. Right.
- 18:54Even projecting that sluggish pace, the intrinsic value comes
- 18:57out to roughly $162.00 a share. Which is still less than half of
- 19:01what the market is currently pricing it at.
- 19:04So what did the math look like for case 3, the nominal
- 19:07scenario? If we assume they can maintain
- 19:10their historical average growth rate of 34.2% every single year
- 19:14for the next decade, the model suggests the equity is worth
- 19:17$1400 a share. 1400 Yeah. So how is a regular learner
- 19:22supposed to interpret a valuation gap that swings wildly
- 19:26between a 76 dollar floor and a $1400 ceiling?
- 19:30Why is the stock sitting at 369? This raises an important
- 19:33question about current market psychology.
- 19:36By pricing the stock near $369, the market is assuming that the
- 19:40nominal perfection case will unfold without any meaningful
- 19:43macroeconomic interruption. They're assuming no bumps in the
- 19:46road. Exactly.
- 19:48Investors are assuming Broadcom will seamlessly compound at
- 19:50massive double digit rates, that TSMC will miraculously build
- 19:54enough Kawas packaging capacity to meet demand, and that the VM
- 19:57Ware lawsuits won't fundamentally damage the
- 19:59software division's terminal value.
- 20:01They are pricing in a perfectly paved toll road with 0 traffic
- 20:05jams, no construction delays, and no one ever taking a detour.
- 20:09And a strict value practitioner operates completely differently.
- 20:13They demand a margin of safety, usually insisting on a 30%
- 20:17discount to a conservative valuation to protect against
- 20:20unforeseen disasters. Right.
- 20:22To get that margin of safety on Broadcom based on these reports,
- 20:26an investor would essentially be forced to sit on their hands and
- 20:28wait for a severe macroeconomic contraction, a massive market
- 20:33crash, because right now you're paying top dollar for flawless
- 20:36future execution that has not yet occurred.
- 20:39It's like an athlete jumping over a bar set at 10 feet, but
- 20:42the crowd boos because they assumed the athlete was going to
- 20:45jump 12. The numbers were flawless, but
- 20:47they weren't more flawless than the perfection already priced
- 20:50into the stock. That perfectly explains the post
- 20:53earnings drop. It really is a delicate balance
- 20:55of incredible technological supremacy and highly aggressive
- 20:59corporate monetization. So to quickly recap our journey
- 21:02today, Broadcom is without a doubt an absolute cash
- 21:06generating machine. They have built a deep,
- 21:08physically constrained technological Moat in AI through
- 21:12Co packaged optics and networking switches that lower
- 21:15data center power consumption. But they are trading at an
- 21:17extreme premium. Right, heavily constrained by
- 21:20TSMC's manufacturing limits and they are actively fighting
- 21:24bruising public wars with their own massive enterprise software
- 21:28customers over VM Ware licensing.
- 21:31You now know who is building the roads and more importantly, who
- 21:34is charging the tolls. Understanding this specific
- 21:37plumbing of the AI boom gives you a much smarter, more nuanced
- 21:41perspective than just watching the headline numbers of the
- 21:43latest consumer AI chat bots. I couldn't agree more.
- 21:46Before we wrap up, is there one lingering question we should be
- 21:49taking away from all this data? Something to chew on.
- 21:51There is a massive variable that isn't cleanly modeled in any DCF
- 21:55spreadsheet, but it could change the entire landscape.
- 21:59Broadcom's current semiconductor supremacy relies heavily on
- 22:02hyperstalers like Google and Meta, utilizing them as
- 22:06engineering partners to build application specific integrated
- 22:09circuits, or ASICS. The custom chips.
- 22:12Right, chips designed for one hyper specific task rather than
- 22:15general computing. The toll road works incredibly
- 22:18well because Google pays the toll.
- 22:20But what happens to Broadcom's massive 46% cash flow margins if
- 22:25over the next decade these trillion dollar hyper scalers
- 22:28decide they have finally acquired the in house
- 22:31engineering expertise to design 100% of their custom silicon
- 22:35completely by themselves, cutting out the toll road
- 22:37entirely? Wow, what happens when the cars
- 22:40learn to build their own roads? That is the billion dollar
- 22:43question. Thank you for joining us on this
- 22:45deep dive. Keep questioning the headlines.
- 22:47Look for the plumbing. We'll see you next time.