I donât really buy it, I think these businesses are just struggling. If AI were the real deal, weâd see decreasing prices and phenomenal output of material goods and services. Thatâs just not happening.
Why would you think this initially?
Its completely possible for AI productivity improvements to increase profit margins (vs have lower prices) initially.
Once enough companies utilise the tech you would see price wars (and prices would come down) via competition.
Thats probably coming next.
Because price equals marginal cost and AI (or insert other cost reducing thingy here) would reduce the marginal cost and hence prices. A company wouldnât get away with trying to charge more than that for long due to competitive pressures.
If Block had any good ideas, or money to spare, they wouldnât be laying off 40% of their workforce. Thatâs a ridiculously high percentage. They would be moving their staff hardcore to the next AI-fueled product idea and quadruple (or insert some AI-hype multiplier) their product lines or else their competitors (also using AI) would.
Since they are not redirecting their remaining staff towards new product lines and innovation, my conclusion is that the layoffs are due to the business not performing well.
They over-hired during Covid but they already layed off people based on that logic.
Boosting profit margins is what makes the most sense.
The company is doing well (AI summary)
Key figures for full-year 2025
Revenue: about 6.25 billion dollars in Q4;
Gross profit: 10.36 billion dollars for 2025, up 17% year over year;
Q4 2025 gross profit: 2.87 billion dollars, +24% year over year, showing acceleration versus earlier quarters.
Adjusted operating income: 2.08 billion dollars for 2025, with margin around 20% and 30% year-over-year growth;
Q4 adjusted operating income 588 million dollars, +46% year over year.
Adjusted diluted EPS: 2.37 dollars for 2025; Q4 EPS 0.65 dollars, +38% year over year.
Iâm with you on this. AI will absolutely start replacing jobs, but I canât see how, today, a company could suddenly drop 40% of staff on this alone. If it is a major restructure of the operations leveraging AI, can they simply say âeveryone thatâs left, go use AI to get it all doneâ?
I have found AI to be useful at work creating structured text/documentation from fragmented discussions (emails, document, etc). Itâs much better than it was a year ago, but after fighting with it a bit the last month or so, it feels like its just a more powerful version of the same algorithm, like it made 10 loops of logic instead of 6. I can see from copilot that it is including additional âtoolsâ within these loops.
Ultimately, it looks like it is creating a decision tree from the prompt, then generating loops of text that it will pass back through a final review before publishing a response. It is interesting as this is similar to a general problem solving process from an actual human, but in order for it to produce an increasingly complex response, you need an increasingly complex prompt.
This feels like it is going to become a constraint. If we use AI to prompt AI and iterate, even with all the data within an organization, itâs going to converge on the same answer. We see examples of this in the military use case - it keeps picking nuclear weapons or keeps killing the people in charge of the AI. Businesses that rely heavily on AI are going to exchange innovation for market convergence. Innovation is new solutions to existing problems. You might say it can solve a new physics or math problem, but all it is doing is producing a result that is the convergence of all the existing research. Thatâs a data exercise, not innovation.
So increasingly complex prompts are needed to produce a more valuable result. This becomes a skill on its own, and also a limitation on how AI can be used. Itâs important for us to learn that skill, so we can use AI effectively, and understand itâs limitations. Companies need to use AI wisely so they donât kill their innovation and competitive advantage. Those that are racing to go all in on AI are making a big bet, its not a strategy at this time, not when we are talking about 40% of staff..
Iâm using AI quite a bit at the moment in my job. It did a software conversion task for me in about an hour which likely would have taken me 2-3 weeks. On the other hand, itâs helping me on a model selection process and I think the savings is maybe half a day on a day to a day and a halfâs work.
Iâve been half listening to these talks on AI How can we prevent AI from becoming a menace? | CBC Radio
One of the things I heard mentioned is that a lot of AI stuff struggles with autonomy and doing longer projects. Unless someone is there to prompt it, it doesnât do anything. Once you do prompt it, you have to figure out if it gave you useful information or not. If your user is naive, theyâre likely going to get led astray. I find with complex tasks I often get given a lot of rabbit holes to wander into or the AI is left with a lot of unanswered questions and itâs hard to deal with them as answering itâs prompts leads to a series of cascading tangents and the thread gets lost.
To answer this question, the answer is yes. But only for Block and their type of business (tech companies are very exposed to automation).
Automating workflows and using Agentic AI is now viable in financial services as well. Goldman Sachs has been doing this quietly for the last 6 months.
I am also skeptical of anything a CEO says about why they did anything. CEOs are expected to spin everything, especially mass layoffs.
But, I also donât think weâll see notable price changes until millions of people are laid off.
Just for a random concrete example, I think we can count the number of gig workers that Waymo is displacing.
The average FT Uber/Lyft drives 100? rides per week. Waymo is doing 500k or so. So Iâm guessing here Waymo has displaced around 5,000 workers. But Waymoâs prices are set to roughly match the market right now.
Waymo is growing around 100%/year. So this year they will probably displaced another 5k workers. And next year another 10k next year, and 20k the year after that, and 40k after. Presumably at some point it would be really big economic news.
But at what point will it be profitable at current prices. And when would they drop prices? Presumably when they have a competitor and/or theyâve outgrown their markets. But thatâs probably still quite some time-- 5+ years?
And then finally the prices will drop, but only by so much.
Which doesnât look like it will be Tesla anytime soon. Perhaps Zoox?
I am reminded of this phrase regarding bankruptcy:
It happens graduallyâŚand then suddenly.
I think we are there at the US-level in tech-heavy sectors. I do agree though that many companies will use AI as a cover to let go off employees during the coming US economic downturn.
I have never really understood how these tech companies ever grow to the employee base that they have. It seems like at any point in time a significant % of them are chasing after new markets that have a low chance of becoming profitable ever. All companies need to have some sort of R&D area, but Iâm guessing Block really had too much of that going on and their cuts are part AI driven but mostly just a result of looking at their project portfolio and realizing that they could get out of a lot of it with a headline on AI. Itâs a good way to get the stock to pop overnight on any company announcement.
I donât pretend to be the least expert in startups, but I assume part of the problem is that they are building stuff, and once itâs built itâs built.
Howâs AI gonna help with producing physical goods?
Agent Smith will motivate the humans to work harder.
The hype machine envisions autonomous unmanned factories.