Our article on Medium about the gigantic hype of AI capital on the SP500 and the unbeatable opportunity it brings for developers 👇🏻
If you are at least 45, you will probably remember those hectic times of the “dot.com” boom back in the days, when tech concentration became something nobody could ever have predicted.
As it happened then, today we are again living what is without a doubt the next big IT bubble, with just seven companies ( Apple, Microsoft, Nvidia, Amazon, Alphabet, Meta and Tesla) , now accounting for an insane 33,7% of the entire SP500 as of April 2026.
This is not just one more fact, but instead it’s a massive one, when we realise that one third of the US flagship stock index, the benchmark that pension funds, retirement accounts and institutional investors worldwide track religiously, is controlled by fewer companies than fit around a boardroom table.
That number gets even more surreal when you analyse it and see that only by 2016, those same seven companies represented just 12,5% of the same SP500 index, and that the concentration has nearly tripled in just a decade. The top ten companies combined now account for 40% of the index’s total market cap , and that is almost double what the top ten represented in 2015. That’s not something that happens frequently or even every decade or two…In terms used by financial guys, this is an “all time high” in a dataset going back 155 years. Over a century and a half years.
When we look at all this, the obvious question we normally hear from people with a purely financial brain is whether this looks or not like the dot.com bubble, and maybe the proper answer is that while it clearly looks like it on the surface, it doesn’t look like it where it actually counts.
If we look back to the early 2000’s and the specific case of Cisco Systems, for example, we will see that at the peak of the bubble in those years, Cisco was not less than the most valuable company on earth, briefly worth more than 500 billion USD of the time. It traded at 200 times earnings (yes, not 200 times revenue , 200 times earnings) If you bought Cisco stock at its peak in March 2000, you would have waited until 2019 (a whole nineteen year period) just to do break even.
Pets.com, a company many will remember, raised 82 million USD in its IPO of February 2000 and was completely liquidated by November of that same year. It just took nine months, from start to finish.
The dot.com era was built on a very specific kind of magical thinking where people assumed that any business touching the internet deserved an astronomical valuation regardless of whether it made money or even had a coherent path to making money, but most companies actually didn’t need to make a penny.
Website traffic was treated as a proxy for value, revenue was just optional and profits were almost considered unsophisticated and just as a thing small minded companies worried about before they truly understood the internet.
When all this chapter finally came to an end, the wreckage of that collapsed illusion was historic, the Nasdaq fell 78% from it’s peak, and literally trillions in paper wealth were automatically evaporated.
But today we live a completely different reality, at least in our opinion, basically because this time the fundamentals are genuinely different and a careful comparison will help most see the clear difference between the two historical episodes.
The average 2 year forward price to earnings ratio for the biggest AI companies (like Microsoft, Alphabet, Amazon or Meta) currently sit at around 26 times, while at the dot.com peak, that same metric for the largest tech leaders of the time was nearly 70 times, close to three times more. Also, the Nasdaq is currently trading at about 33 times trailing earnings, versus roughly 60 in March 2000, and while no one can deny that these are elevated valuations in both cases, they also exist in a completely different universe from the numbers that defined the bubble era.
To understand this, maybe one of the most evident facts is that the companies driving the current AI “bubble” are, by almost any measure and as simple as that, among the most profitable enterprises in the history of capitalism.
Companies like Apple, Microsoft, Alphabet, Amazon and Meta (alone) generated a combined 350 billion USD in free cash flow and only in their most recent fiscal years, and even if today we are all overwhelmed by great big figures all around us, just think this for a moment: Three hundred and fifty billion dollars in free cash flow.
To name just another big AI monster, Nvidia alone reported a crazy net income of over 120 billion USD for year 2026, and its data center revenue grew 279% year over year in 2024. And the big difference with the dot.com era is that these figures are not speculative future promises but just seriously audited financial independent statements.
If we look back, during the dot.com boom, the big IT players driving the market were mainly just destroying capital, while today the companies driving the market are among the most capital efficient enterprises ever to exist, with just the average top 10 on the SP500 posting a return on capital of 73% versus 18% at the start of the decade.
The infrastructure spending is also structurally and dramatically different from the dot.com era, and here we see again Microsoft planning to invest approx 80 billion just in AI infrastructure alone, or Alphabet raising this figure to target 85 billion, and Meta spending between 115 and 135 billion (nearly double what it spent the year before)
And the most important aspect to notice here is that this massive infrastructure spending is not something speculative or financed by cheap debt or easy to convince investors, but instead it’s backed by a hugely strong balance sheet funded by existing free cash flow, and that is actually a big difference.
The companies building now AI capacity have among the highest free cash flow and strongest balance sheets in the entire equity market. They are also, and this matters, experts in compute logistics. They have managed enormous data center operations for a decade and they are definitely not guessing or experimenting by basing their decisions on hypothesis.
On the other hand, looking aside all these magnificent numbers and maybe with more conservative glasses, we also think that none of this means the bulls are entirely right, and to honour the truth we should also acknowledge the parts of the argument that still remain unresolved, because if we are to be cautious and look for example at the famous Shiller CAPE ratio (a measure that has predicted past bubbles with reasonable accuracy) we would see that it currently sits around 38 to 40, and the scary thing is maybe that the only time it’s been higher in 155 years of data was precisely at the dot.com peak of 44,19.
Also, SP500 top ten concentration now exceeds dot.com levels by nearly 50%, with a recent MIT study founding that 95% of enterprise AI pilot projects have produced zero meaningful return.
Those are also facts we should not forget, and maybe the most correct way of taking in all this is not just saying "everything is fine”, but understanding that the companies at the center of this new story are truly making gigantic amounts of money, that AI use is only starting to reach enterprise scale, and that the bubble pops not when some ratios are high but when the earnings stop growing, and whether the earnings stop growing is the only question that maybe actually matters.
With regards to us in the developing industry, what is interesting from a software development perspective is probably that the AI race happening right now is, in historical terms, the equivalent of when cloud computing went from an interesting concept to infrastructure that every software team was building on top of.
The companies at the center of that moment (Azure, Google Cloud, etc) didn’t just benefit from the transition, but instead they created an entire ecosystem of opportunity that made millions of other companies and developers wealthier and more productive simultaneously, and that’s the pattern that we think will be repeated and that is more interesting.
The numbers today suggest the developer opportunity is truly enormous in ways that haven’t fully shown up yet, with a global software market that reached 823.92 billion USD in 2025 and is projected to reach the stunning figure of 2.24 trillion by 2034.
The custom software development segment, one that for us is maybe highest in the podium, is expanding at a 22,71% annual growth rate (nearly double the general software market) Also, global IT spending on software is projected to exceed 6 trillion in 2026, the largest spending category increase across all segments.
Many also project that around 40% of enterprise applications will be integrated with task specific AI agents by the end of 2026, up from less than 5% in 2025, with the application software market looking at growing to approx 780 billion by 2030 on that trajectory alone.
The developer adoption numbers are also striking, and if we listen to the JetBrains 2025 Developer Ecosystem Survey (with over 24.000 professionals), 85% of developers now regularly use AI tools for coding and software design. Over 51% of all code committed just to GitHub in early 2026 was either generated or substantially assisted by an AI tool. The AI coding tools market alone reached 12.8 billion in 2026, with a 150% increase with respect to the previous year.
The productivity implication is what actually changes all, and this is what is dictating what a small team can actually build. AI coding assistants can now save tiny teams 3 to 5 hours per developer and per week, according to many researchers, and in practical terms that means a team of five developers today has the productive output of a team that would have required seven or eight people just two years ago.
For small startups, for freelancers or for boutique software agencies, the above is actually a structural competitive advantage that compounds.
So to brief, the valuations of the main AI monsters are indeed elevated, but the critical thing to understand about that statistics is what it implies rather than what it states. It clearly does not mean AI doesn’t work , but instead it means that most enterprises are still figuring out how to deploy it effectively. They are just in an already very profitable experimental phase, and the companies, developers and consultants who can bridge that gap are sitting in front of a demand curve that is only beginning to ramp in order to change this world forever.
We think that this huge AI market concentration is ultimately a reflection of where the people and the market believes the next decade of economic value creation will be concentrated. That belief may be somewhat overstated in valuation terms, but the adoption curve with AI doesn’t look similar under any circumstances to what we saw during the dot.com era.
Cisco at 200 times earnings was a story about hype meeting accounting. Nvidia at 41 times earnings, generating 120 billion USD in net income in a single year is an absolutely different story.
Developers must understand that distinction and simply build the layer of software that sits on top of these platforms . They should not just act as spectators to the biggest capital allocation event of the history of technology. They must be participants.
Top comments (0)