Big Tech is the best way to play AI (and the best way to beat the returns of the S&P 500 and hence the vast majority of professional investors over the next 10 years)

Don't overthink it

The crux of my argument is that BIG TECH is a SYNTHETIC MONOPOLY. I wrote a post about synthetic monopolies which can be found here - but I will re-print the relevant definition below for easy reference:

My definition of a synthetic monopoly (henceforth SM) is any small grouping of companies that are capturing the vast majority of economic benefit from a certain business model/product/service, and which collectively have an extremely low likelihood of being disrupted by companies outside of the group (regulation more likely). Further, an SM is likely to maintain or increase its collective market share while simultaneously maintaining or increasing its collective (and attractive) profit margins.

The difference between an SM and an oligopoly is that many oligopolies do NOT have the attractive characteristics one would expect from a monopoly. For example, Boeing and Airbus are an oligopoly without threat of disruption, but neither company has the profit margins, EPS growth or revenue stability of a monopoly, NOR would they have those characteristics when looked at as if they were a hypothetical single entity. Here’s what I mean.

IF Boeing had great financials and Airbus had terrible financials, and IF when we combined their financials and created pass-through financial statements they looked attractive, THEN they would qualify as an SM. Alas, neither company is attractive so the combined entity would also be unattractive.

Verizon, AT&T and T-Mobile are another example of an oligopoly that wouldn’t qualify as an SM.

There are three main takeaways from this post:

1. The only companies capable of disrupting Big Tech companies are other Big Tech companies

2. IF a Big Tech company is disrupted by another Big Tech company - the economic value derived from said disruption is likely to be approximately equal to or greater than the value it disrupted (i.e. competition within the group as a whole will not result in profit cannibalization)

3. Big Tech is undervalued relative to the S&P 500

#1 The only companies capable of disrupting Big Tech companies are other Big Tech companies

As always I will explain with anecdotes, starting with Google’s Search business (henceforth: “Search”).

There are two big risks to Search:

1. Another company making a competitive search engine. For example, Google pays Apple ~$30B per year to be the search engine on Apple devices. Apple might in-house search.

2. ChatGPT/LLMs provide an alternative mechanism that consumers prefer.

Google…

1. Has 92% market share in Search

2. ~30% margins on earnings before tax

3. 9 products with > 1 billion users

4. >$100B in cash

5. Grew revenues over the past five years at > 18% per year

6. Grew net income over the past 12 months by 27%

Despite all of the above it is trading at approximately the same PE as the S&P 500 and lower than the S&P 500 on a forward basis.

Why?

Companies like Coca-Cola and Waste Management trade at very high multiples relative to their growth and earnings. Both companies are essentially GDP + 2-3% growers. Waste Management has a PE north of 30 even though its 5 year revenue growth was only 6.7%.

People are willing to pay more per dollar of earnings for companies like Waste Management because everyone knows (rightly so) that they will still be collecting our trash 100 years from now. In investing parlance we would say: there is very little terminal value risk in the trash business.

Google on the other hand is perceived to have lots of terminal value risk (as was Meta/Facebook just 18 months ago when everyone though TikTok would destroy them). Investors question whether Google and Meta will still be around in 20 years (at least in any form resembling their current dominance) - much less 100.

I will readily admit that I too am uncertain about the future of Search. I could see it being replaced entirely by the end of this decade. But - I am also near certain that whatever replaces it (if it is replaced) will be a product provided by another Big Tech company.

Google’s biggest threats are Microsoft, Apple, and maybe Meta (all three of them are contenders for launching LLM based search products). What Google is most certainly not threatened by are any companies outside of the world’s giants (I consider OpenAI to be part of Microsoft - which owns nearly 50% of it - and what OpenAI has accomplished is only possible because of its relationship with Microsoft).

What about Nvidia - whose stock has recently gone parabolic?

Microsoft, Amazon, Apple, Meta and Google are all working on developing their own logic chips. Maybe they will contract directly with Taiwan Semiconductor and avoid having to pay Nvidia’s outrageous prices?

AMD might catch up and offer an alternative.

Intel might catch up (LOL! just kidding on this one - but Intel might get to make some of Nvidia’s chips one day).

The takeaway is that similarly to Search - IF Nvidia gets disrupted it will be by another member (or members) of Big Tech - not a company outside of the group. Nvidia’s revenue loss is another Big Tech company’s gain via acquiring less costly chips.

Let’s do three more.

Microsoft, Amazon and Google are competing to dominate the cloud. Microsoft and Amazon are able to maintain monopoly like margins while competing - Google is a distant third but still its profitability is improving with time. There is precisely ZERO chance another player comes along and puts a dent in the cloud business of these three companies, however - it is entirely possible that one of them gains share vs. the others.

Amazon has been increasing their share of digital advertising. It is missed on many people that Amazon is Google’s biggest competitor. The reason is that people go directly to Amazon to “search” for a product - bypassing Google’s opportunity to serve an ad. Amazon is increasingly competing with Meta on product discovery to boot. While TikTok is growing quickly - its revenues are still expected to finish 2024 at just ~5% of the revenue of [Amazon, Meta & Google’s] ad businesses.

IF the Synth market (Post here: synthetic humanoid robots) ends up becoming a multi-trillion dollar market over the coming decade - it will either be Tesla or another member of Big Tech who come to dominate or co-dominate that market (Maybe Amazon or Microsoft after buying Figure?)

Collectively, Big Tech holds monopolies in:

1. Computer operating systems

2. Cutting edge logic tech chip design

3. Mobile operating systems

4. Productivity software (read: precursor to “Co-Pilot” type AI assistants in the future)

5. Base-layer of the cloud (explained in this post)

6. Digital advertising

7. Email

8. Social Media

9. Digital content distribution

10. Soon to be synthetic humanoids

11. Maybe someday driverless vehicles?

To summarize: Any disruption experienced by Big Tech will come from other Big Tech. The takeaway is that: as a basket, while individual members of Big Tech may have terminal value risk - the basket as a whole eliminates that risk b/c wherever terminal value is lost by one company it will be re-captured by other members of the same synthetic monopoly.

There is one obvious potential flaw in this logic: one member’s gain may not generate an equivalent amount of revenue and/or profit as the other member’s loss. For example, if Google Search is replaced by LLMs then the business model built around LLMs might generate less revenue and profit than Search does today.

This leads us to the second takeaway of this post.

#2 IF a Big Tech company is disrupted by another Big Tech company - the economic value derived from said disruption across the group is likely to be approximately equal to or greater than the value it disrupted.

Put differently - competition within the group as a whole will not result in profit cannibalization. Let’s walk through a hypothetical.

Assume Search generates $300B in revenue per year and $90B of profit (not far off from reality).

Now let’s assume that Search goes away entirely and is replaced by LLMs that have all scanned the entire internet. If someone wants to buy food, a plane ticket, or get a good deal on a massage they just ask their LLM to find the best deal. Here’s a “worst-case” hypothetical of what the P&L might look like for a Search replacement.

5.5 billion people use Search today. My guess is that:

600 million of those people would be willing to pay $10/month for their LLM (those in the rich world)

1 billion people would be willing to pay $5/month (people in middle income countries)

3.9 billion people would be willing to pay $1/month (people in lower income countries)

Revenue generated would hence be:

600 million X 10 X 12 = $72 billion per year +

1 billion X 5 X 12 = $60 billion per year +

3.9 billion X 1 X 12 = $46.8 billion per year

Total = $178.8 billion.

Let’s further assume that due to compute costs the LLM profit is only 20% instead of 30%. This would give us the following:

178.8 X .2 = 35.76 - > so under our new model the $90B of profit Google makes is replaced by only $35.76B of profit by whoever takes over Search.

This seems like a reasonable “worst case” scenario. Google loses $300B of revenue and $90B of profit and goes out of business but another member of Big Tech gains $178.8 billion of revenue $35.76B of profit.

However, this scenario seems extremely unlikely to me (<1% chance). What is far more likely is that IF something replaces Search it will be better than search and hence the market size and profits derived from it will be even greater.

Here’s a more plausible scenario of how things could play out.

Search still disappears and Google loses the $300B of revenue and the $90B of profit.

The top line figure above still holds true ($178.8B goes to the company that replaces search).

However, in addition to the $178.8B of revenue, the LLM provider offers services to the businesses who are still trying to reach consumers and charges them a fee for doing so. Essentially - companies still serve ads, but instead of serving them on the front page of a Search Engine they are paying the LLM provider to make sure that when the AI assistant scans the internet for the best place to eat nearby their restaurant comes up on the list of recommendations. There might be a laundry list of additional services that companies spend money on (to the LLM provider), here are two more I came up with off the top of my head:

1. LLMs will have far more intimate knowledge of what consumers want - so companies will probably pay the LLM provider to help them optimize what products they sell, what price they sell them at, how they create loyalty programs, how they describe their products, etc. Alternatively, each of these features could be provided by new specialized upstart companies built on top of the LLM “Base-Layer”. Either way the LLM is making more money.

2. Companies could pay the LLM provider to push a consumer to purchase something that they weren’t even pro-actively looking for. For example, the LLM might know that a person’s anniversary is coming up and suggest the perfect vacation based on that person’s tastes, personality, budget, and stage of life. If the LLM has Agency (SEE THIS POST) - then it could go so far as to plan out a few different potential trips, reach out to the airline companies, hotels and rental car agencies pro-actively - and get the companies to “Bid” on what they would be willing to pay in order to land a near-certain sale. This would be Meta/Facebook style discovery on steroids.

Example #2 brings up an interesting side-point about price discovery and driving productivity. If an LLM is negotiating on behalf of a consumer to always get the best price (something particularly impactful on big ticket items) - this would be an example of an “add-on” that the LLM could charge to the consumer. Maybe the base-plan is $10 per month, but then there is an added “Best Price Guarantee” package that costs $100 per month. For $10 you can ask the LLM where to buy a car. For $100 you can get an LLM that will call every dealership in the country and negotiate the best possible price and package.

Companies might hate this - but the net result will be huge gains in productivity and lower prices for consumers.

Interestingly, this example also illustrates why I call Big Tech companies “Vortexes” (SEE THIS POST) - they keep taking an ever larger share (as a percentage) of economic value.

Consider how easy it might be to justify a $100/month subscription.

$100 per month is $1200 per year. $1200 is 5% of $24,000. Most people spend $24,000+ per year on things in the rich world. If the LLM could save you 5% on $24,000 worth of purchases then it has justified the $100 per month subscription. Anything beyond that is profit to the consumer. Consider the potential savings just on groceries. The difference in the cost of groceries at Publix vs. Kroger vs. Aldi’s vs. Whole Foods is enormous. People don’t go to three different grocery stores each time they go shopping - their time is too valuable. So - each time they shop they are making a decision to prioritize one factor over another (convenience, price, want a high quality meat that only one store has and just buy the other things because they’re already there, etc). But, if your LLM is doing all of the ordering it could make sure to get the best price on each individual item - likely covering the cost of the delivery fee and then some.

You could easily save $1,200 on the purchase of a vehicle by negotiating with 10 different dealers…

Undoubtedly at $100/mo you will also be getting a private tutor, personal assistant, etc…

If 600 million people subscribe to the premium tier of LLM we’re talking about $720,000,000,000 of revenue per year ($720 billion).

And as for the margin compression I mentioned above - I don’t see any reason why the margins for an LLM subscription would be less than Search in the long run. Margins are a function of competitive pressure and value delivered. If the value is higher and if only 2 companies are capable of delivering the service - they (margins) are likely to remain exorbitant.

To summarize, we might turn $300B of revenue and $90B of Profit (earned by Google) into $800B+ ($720B for the premium tier + all the other tiers) in revenue and $240B+ in profit (assuming similar margins to Search) earned by some other company within the synthetic monopoly.

This brings us to the final takeaway of the post.

#3 Big Tech is undervalued relative to the S&P 500

Asset “A” is undervalued relative to asset “B” if asset A out-performs the risk adjusted returns of asset B over time. By this definition (the only correct definition, imho) Big Tech is undervalued relative to the S&P 500.

Let’s look at some Big Tech valuations based on the next 12 month sales and earnings estimates to get us started:

EBT is “Earnings Before Tax” - so if you want to convert the EBT multiple into a PE you can multiply the EBT $ X .82 (implying an 18% tax rate) and then divide the market cap figure into the result to get a reasonable approximation.

As a basket, the EBT multiple of the above companies is approximately 25 and their PE ratio (again, as a basket) is around 30.

For reference, the PE ratio of the S&P 500 is around 21 using financials over the same timeframe.

A lay person might take the 30 and divide it by the 21 [30/21 = 1.428] and conclude that our Big Tech basket is 42.8% overvalued [1.428-1 = .428 X 100% = 42.8%]. However, a more savvy individual would notice that the list of companies above have margins that are on an entirely different stratosphere than the average company in the S&P 500. That same savvy individual would realize that Big Tech is exceedingly likely to grow earnings at least 5% greater per annum as a basket than the rest of the S&P 500.

The average net profit margin for the S&P 500 hovers between 11% and 12%. The net profit margin of our basket above is just north of 22%, and would be far higher if Amazon didn’t skew the figures by having enormous revenue and low margins.

Ex Amazon, Tesla and Intel the net profit margin of our basket rises to just over 30%.

Margins represent a company’s power in the market. Higher margins are a result of less competition and hence more pricing power. All else equal this would lead to a higher multiple.

The pricing power of the group of companies above is far beyond anything the world has seen since the first corporations were established many hundreds of years ago.

To re-iterate from above, companies like Waste Management (and utilities, consumer staples, etc) trade at premium multiples because they are perceived as having lower terminal value risk. I agree that they have lower terminal value risk on an individual basis. But the terminal value risk of the basket of Big Tech companies taken as a whole is negligible.

Since 1970 the S&P 500 has grown profits at approximately 7% per year. That figure is closer to 6% since 2006.

Meanwhile, the Nasdaq 100 has grown profits around 14.5% over the same period.

If the earnings multiple doesn’t change then:

earnings growth + [dividends/repurchases] = total return.

I think it would be very conservative to assume that Big Tech grows earnings at a rate 5% greater than the S&P 500.

If that happens there is room to outperform long-term (10 years) even with meaningful multiple contraction.

How my argument falls apart

If I’m wrong it will be for one three reasons:

1. Profits within the group are cannibalized via intra-group competition

2. An upstart not currently in the group obtains meaningful market share at the expense of a group member

3. Multiples contract far more than I perceive as likely

Fortunately, because the starting multiple of the basket is 30 and not 40 or 50 - the risk of meaningfully underperforming the basket of shitcos making up the S&P500 (by my estimate north of 300 of them are fundamentally unattractive businesses) is low even if we see cannibalization, multiple contraction and competition from upstarts. I really can’t envision a scenario where Big Tech as a basket underperforms over 10 years.

What seems more likely to me is that the industries of the future (humanoid robots, driverless cars, AGI, etc) are dominated by Big Tech and lead to earnings growth that is faster relative to the S&P 500 than it has been historically.

Because Tech has rocketed over the past year there is obviously a risk that it takes a dive in the near-term or under performs the S&P 500 over a period of 2-3 years (unlikely, but possible).

While the risk/reward for overweighting Big Tech is not nearly what it was a year ago -it’s still worth betting on if you have a long-term perspective.

The easiest way to do this is to go long the $QQQ. Alternatively, market cap weighting the top 20-30 tech companies and allocating a portion of your portfolio to the basket would also work.

If I weren’t already so tech heavy and were starting from scratch I would probably market cap weight the top 30 companies but cut the allocations to Nvidia, Google and Apple in half (Apple due to growth concerns, Google because it is run by morons, and Nvidia because it recently went parabolic).

That said - I would probably end up being wrong to make those adjustments. Doing anything other than market cap weighting skews the underlying logic of buying the basket. But doing so would make me sleep better at night nonetheless.