As per the most recent estimates, Google Search accounts for ~90% of all online searches worldwide and ~27.7% of total digital advertising revenues, followed closely in second place by Meta with ~26.8%. The global advertising revenue market surpassed 1 trillion USD for the first time in 2024 (Global ad revenue to top $1 trillion, dominated by Google and Meta). What about the global subscriptions economy? According to recent data, it is a ~600 billion USD business worldwide as of 2024, set to grow to 1.5 trillion USD by 2028 (source). What I just described here is the total pot of revenues available to all companies now developing AI LLM-based tools.
What about the costs? So far, little more than 1 trillion USD worth of capital has been invested in developing AI, the bulk of it going into buying Nvidia GPUs. McKinsey estimated that 7 trillion USD will be required to scale up AI data centers by 2030 to keep pace with the current growing demand (The cost of compute: A $7 trillion race to scale data centers) between new infrastructure, power, and operating costs needed to support it.
If we put costs and revenues together, assuming AI grabs 100% of the market shares of global advertising and subscriptions market (which is totally impossible), companies investing in AI will mathematically never be able to break even, right? Suppose this is the case in the very best-case scenario, unless Ads and Subscription businesses start growing massively because people will prefer to pay more for AI and cut back on food and living expenses. In that case, the numbers are already saying that what’s going on right now is totally unsustainable. Some of you might be willing to argue that AI will unlock massive savings for companies that will be able to cut costs and jobs made redundant by AI. If this is going to be the case, aren’t people going to have less and less money available to spend because of high unemployment? The logical consequence is that companies will eventually have to scale back advertising costs to maintain the current Ads ROI levels, eventually biting into the global advertising market revenues pot. Second-degree effects will be, of course, on their top line with much less demand, especially for discretionary consumer goods and services. So while subscription services might rise because of corporate demand for AI, the economic crisis sparked by high unemployment, especially among white-collar higher-salary jobs, will eventually shrink the advertising market.
If these are the projections, why then are companies still pouring hundreds of billions to develop and scale up AI as fast as possible? Because in the same way Google managed to control 90% of the internet search market, the very few (if not the last one) able to survive this CAPEX carnage will effectively have a huge revenue market available to split among themselves. Today there are already 44 companies out there offering AI LLM-based model services, and it should not come as a surprise that access to AI today is given for free to those users, most of them, who do not have constant significant demand for it. The remaining ones, a small cut, are paying for “pro” subscription services. Today AI revenues are a tiny percentage of the investments, which is why it is granted that sooner rather than later, AI LLM models will become a platform to sell advertising in the same way search engines evolved (post).
Clearly, a big stream of capital is required by companies in the sector to remain in the race and not file for Chapter 11, and how are these companies convincing investors to give them all this money? Very simple: fabricating revenues and portraying healthier operations than what they are in reality and promising huge profits in the future (that are mathematically impossible). Those of you reading my articles for a long time know I have been warning about this non-stop for years, and now more and more high-profile chickens are coming to roost after the implosion of the former unicorn Builder.Ai (Builder.ai Collapse Exposes AI Deception in Tech Industry).
At this point, the game is very clear: put your competitors out of business to grab their share of the revenues (eventually). In this game, mega-cap companies with virtually endless amounts of capital like Google or Meta, and already a strong presence in the advertising market clearly have a strong advantage. Hold on, doesn’t this mean that these two companies I mentioned are effectively pouring huge amounts of CAPEX just to defend their current revenue pot? Bingo.
Now, tell me one thing: if a company is investing a huge amount of resources when the expected revenue growth is expected to be marginal, what’s going to be the ROI on this deployed capital? Much lower than the current one. If companies like Amazon or Walmart invest a ton in AI to cut their operational costs and lay off thousands of workers, with ripple effects throughout their supply chain and the economy, what’s going to be the impact on these companies’ revenue line? Unless you assume robots will go to the supermarket in the future or buy widgets online for their pleasure, then the outcome here is pretty logical, isn’t it?
Clearly, this whole AI push right now is a zero-sum game because these companies will eventually be constrained by the size of the economies they operate in. Sure, if a lot of smaller companies go out of business, the remaining ones will benefit greatly and their shareholders will hit the jackpot like those who believed in Google in the very early days did, while those who backed AOL, Yahoo, or other providers are still licking their wounds. Obviously, there will be a massive glut of data center capacity down the road because the survivors won’t need to take over losers’ operations to run theirs. Isn’t this exactly what happened in the aftermath of the DotCom bubble? Paradoxically, the glut of computer capacity at that time drove computing costs to the bottom, giving access to the system to companies with less capital and set up in a garage that later became today’s behemoths like Google.
Here is the paradox: those that will survive the carnage I described will have to watch their backs from newcomers, although I won’t be surprised if those successful ones are acquired for a big price tag like what happened to YouTube, Instagram, or WhatsApp, giving behemoths the opportunity to grow even bigger and capture a bigger chunk of revenues made available in the economy. An economy that, let’s not forget, is just growing because it’s relentlessly inflated through government deficit spending supported by central banks’ debt monetization. Once upon a time, companies that successfully became monopolies like Standard Oil were forcefully split up by governments because everyone knew how these are bad for the economic system; however, this hasn’t happened for many decades now, and I wouldn’t hold my breath on it happening because there are too many economic interests at stake, and with indexes like the S&P500 becoming more and more concentrated on a handful of companies, such a move would surely trigger a market crash no politician can stomach anymore nowadays.
JustDario on X | JustDario on Instagram | JustDario on YouTube