This week, The Wall Street Journal published an alarming illustration of the effects of AI potential share of GDP which, I would argue, did more to muddy the waters than tell anyone anything, mainly because its measure, for some reason, covered the period 2025 to 2032, meaning that six of the eight years in the analysis were spending estimates.
To be fair, this analysis came from Stijn van Nieuwerburgh of the Brookings Institution rather than the Journal itself, but I cannot express how deeply it is unhelpful it’s about discussing things years into the future.
The Journal itself recognized this, giving us a much more useful figure, my emphasis:
Projecting investments is tricky and total spending could well be considerably lower. Yet the money invested this year in data centers already represents an unprecedented investment in recent history. Investments in AI in the United States are expected to reach 1.9% of GDP in 2026, according to new estimates from Goldman Sachs.. The railroad boom of the late 19th century marked the last time when the construction of a new industry was a larger part of the economy.
However, it is important to point out that this is almost entirely due to AI data center construction and GPU sales rather than anything to do with companies actually renting AI compute. In other words, The AI itself does not help drive U.S. economic growth, but rather provides the infrastructure on which it can theoretically operate.
This is extremely problematic, because it means that at some point data center construction will slow down or stop (as discussed in this week’s free newsletter) through a combination of moratoriums and ever-increasing debt, removing any contribution to GDP and leaving AI – either through revenues generated by the sale of services or through improved productivity – to fill the deficit.
The problem of calculating exact The contribution of the technology industry is that many different elements of these companies come together in different “industries” (there are seventeen in total) around the world. the BEAs data based on specific economic contribution.
For example, Apple’s vertical services (like iCloud) would enter the “US Information/ICT” indices of the GDP calculation, but its iPhone, Mac, and iPad sales would go into manufacturing, the same where NVIDIA’s GPU sales would go. ICT also doesn’t include revenue from consulting services or IT services from companies like Accenture, but that’s not really relevant to the analysis.
Regardless, the ICT sector’s contribution to GDP is actually very, very useful for this calculation, because it specifically includes AI software sales And AI GPU rentals. There are two numbers to look at here. On the part of nominal GDP – strictly the amount of its contribution to GDP – technology’s contribution has remained stable over the past two years. In other words, all these supposed GPU rentals and AI software sales in 2024 and 2025 haven’t really done much on an economic basis.
I can already hear someone shouting that we need to measure “real GDP” – which takes into account improvements in software and hardware that would, in theory, increase the ICT sector’s contribution to real GDP.
The problem I have with this analysis is that the BLS Producer Price Index for Software Companies — a measure of software package price changes over time that the BEA uses to calculate real GDP — is currently on hold lower than that of 1997, suggesting that software prices have abandoned with time in a period where general prices have roughly doubled.
This is not at all accurate based on the actual experience of people purchasing software. As I covered in the Haters’ Guide to the SaaSpocalypsemore than half of SaaS companies have raised prices every year since 2022, customers are paying more every year for the same features, and overall SaaS inflation was nine percentage points higher than consumer inflation every month of 2025. This problem started around 2022, when Microsoft has increased its pricesinspiring industry-wide inflation.
In other words, the BLS’s “quality” adjustments appear to treat many of these price increases as if customers were getting better software for their money, rather than paying more money for the same software, with the BEA in turn viewing this as businesses are buying more software.
The BLS estimates that software is effectively the same price as it was in 1997, largely because it gives customers “more value” and allows them to “do more,” which doesn’t make sense if you’ve used a Microsoft product recently.
The BLS’s preferred method for these adjustments is based on cost of switching (IE: how much more expensive to provide), meaning that any price increases related to AI services, which require expensive tokens to provide, could potentially be considered. the same Or even cheaper in the eyes of the BLS.
The BLS data underestimates the actual increase in the cost of software in recent years, and as a result, the BEA could – accidentally – overestimate its contribution to GDP. And AI services are only making things worse.
With this in mind, I calculated the gross value (a company’s sales minus costs purchased from other companies, hence excluding salaries) added by the ICT sector relative to real GDP, and found that while technology’s share has grown steadily, this growth has not changed dramatically in the AI era, even using the flattering BEA figures.
Side note: I want to be perfectly clear about the data I’m discussing here. The ICT industry part of what you are about to read is including GPU rentalbut BLS data around software does not include them.
All of this is to say that while various statistical agencies have a fairly distorted view of the tech industry – at least when it comes to software sales and infrastructure rentals – the contribution of the AI era seems a bit poor.
In preparing this newsletter, I realized something a little worrying: Fundamentally, any economic analysis of “AI’s contribution to the economy” is based on either flawed data or shaky assumptions about AI and the tech industry itself. Economists have focused on trying to rationalize the astonishing amounts of money invested in AI, and in doing so they have failed to ensure that even their the simplest assumptions – like how much the tech industry itself contributes – meaningfully capture what is happening.
Today’s newsletter is a frank assessment of the true effect of AI on the economy, specifically focused on really measuring what’s happening. Today rather than the endless what-if analyzes you find everywhere else.
You see, everyone is obsessed with metrics that don’t matter – cost per token, teraflops, and vague analyzes of employment data – all to avoid a much darker point: when you remove investments, AI has had a negligible effect on GDP.
And beneath the surface, I found evidence that software sales’ contribution to GDP may have been significantly off since 2022.
Gn bussni
