Wix lays off 1,000 workers as the AI jobs apocalypse keeps looking very real

[midian182]

A hot potato: The AI jobs apocalypse that Sam Altman is glad isn't happening continues not to happen. The latest company to lay off a large portion of employees is Wix, which is cutting 20% of its workforce while citing the "fast evolution" of AI.

CEO Avishai Abrahami wrote on X that reducing the Wix team by 20%, or around 1,000 people, was a "hard decision" (four times).

Abrahami said the first reason why the website builder is cutting staff is the exchange rate between the Israeli shekel and the US dollar. He said the shekel has strengthened against the dollar in recent quarters, creating a "structural pressure" on Wix's ability to operate at scale.

Then there's the very familiar second reason: "the fast evolution of AI capabilities." Abrahami writes that "we have witnessed the most significant shift in how companies are built since the invention of modern programming languages in the 1970s."

"This is not just about adopting new tools - it is about rewiring how companies are built, how they think, how they manage and how they operate. Companies that embrace this change will not only build faster; they will build things the previous generation literally could not have imagined."

– Avishai Abrahami (@Avishai_ab) May 28, 2026

The CEO goes on to talk about the need to become a faster, leaner, flatter organization due to AI, with fewer layers between leadership and team members. Then there are the usual lines about needing to evolve in order to compete, we had no choice, etc.

This sort of CEO message has become a depressingly common sight in the AI age. Abrahami's post is very similar to the one Jack Dorsey wrote when Block laid off over 4,000 people earlier this year. The former Twitter boss said it was better to get rid of everyone at once rather than lay people off slowly over the coming months and years as automation took their jobs.

Wix's mass layoffs come just days after OpenAI boss Altman gushed about how excited he was that the AI jobs apocalypse he predicted last year had not come to pass. There have been almost 116,000 layoffs in the tech world so far in 2026, quickly approaching the 124,000 that we saw for the entirety of 2025. A massive percentage of these cuts are tied to AI, either through direct job replacement or companies' resources being redirected toward AI infrastructure.

– Shibetoshi Nakamoto (@BillyM2k) February 26, 2026

Altman's view that things could always be worse might fade away in the coming years. A recent survey of almost 1,000 executives found that 99% said they expect to make some reduction in headcount in the next 24 months.

The irony in all this is that companies spending millions and billions on AI still aren't seeing any meaningful returns. A survey in January found more than half the participating CEOs said AI adoption had not increased revenue or reduced costs, something Uber is all too familiar with.

Permalink to story:

Wix lays off 1,000 workers as the AI jobs apocalypse keeps looking very real

 

[Theinsanegamer]

Once again we have AI hysteria covering for COVID layoffs. I'm sure that some will cry and gnash their teeth over this, but here is the reality:

Wix employees in

2013: 584
2016: 1,284
2019: 2,562
2022: 4,742

So even after these layoffs, wix still employs over 1,000 people MORE then it did pre COVID.

 

[GeoffreyA]

The usual hand wringing and "hard decisions." What's new? It's important to keep one's mind focused on the Emperor's New Clothes.

 

[yRaz]

I swear that this is just marketing to get more funding for AI as the VC money is drying up and the banks have slowed lending significantly. "Look, AI, it's working!". meanwhile, replacing humans with AI is turning out to be 3 to 4 times more expensive with worse results. It would cost them the equivalent of 3-4000 jobs to replace the 1000 humans with AI. The era of cheap tokens is over now that data centers can't keep subsidizing them as customer acquisition costs

 

[GeoffreyA]

I swear that this is just marketing to get more funding for AI as the VC money is drying up and the banks have slowed lending significantly. "Look, AI, it's working!". meanwhile, replacing humans with AI is turning out to be 3 to 4 times more expensive with worse results. It would cost them the equivalent of 3-4000 jobs to replace the 1000 humans with AI. The era of cheap tokens is over now that data centers can't keep subsidizing them as customer acquisition costs

In the end, I would argue, it comes down to performance per watt. Until AI approaches the biological-computing efficiency of Nature (~20W, apparently) to perform task X, Y, Z, it will cost more in terms of the universe's currency and money.

 

[pnntmp]

Once again we have AI hysteria covering for COVID layoffs. I'm sure that some will cry and gnash their teeth over this, but here is the reality:

Wix employees in

2013: 584
2016: 1,284
2019: 2,562
2022: 4,742

So even after these layoffs, wix still employs over 1,000 people MORE then it did pre COVID.

And, on top of that, Wix market cap today is just 1/3 of what it was 1 year ago. With such a collapse, only 20% layoffs is positive news.

Actually this is one of the few cases where AI is really involved, albeit indirectly, but not because the company is replacing developers with AI. It's because the customers are replacing Wix with AI. Today AI is already good enough to make a website, and people don't have to pay subscriptions to Wix and similar companies.

 

[Theinsanegamer]

I swear that this is just marketing to get more funding for AI as the VC money is drying up and the banks have slowed lending significantly. "Look, AI, it's working!". meanwhile, replacing humans with AI is turning out to be 3 to 4 times more expensive with worse results. It would cost them the equivalent of 3-4000 jobs to replace the 1000 humans with AI. The era of cheap tokens is over now that data centers can't keep subsidizing them as customer acquisition costs

Very likely. Much like companies bragging about their new blockchains a few years back. Anything to keep what is left of investment dollars flowing and line going up.

 

[Plutoisaplanet]

In the end, I would argue, it comes down to performance per watt. Until AI approaches the biological-computing efficiency of Nature (~20W, apparently) to perform task X, Y, Z, it will cost more in terms of the universe's currency and money.

20 watts is the energy usage of the human brain, not the whole body. 100 watts is the whole body (at rest only), and we work 8/24 hours so in effect the actual operational efficiency only in energy costs is 300 watts (15x what you suggested). Now you have to look at food costs vs electricity costs. Let's assume a ridiculously high electricity cost of $0.35 per kWh for computers. For food, we'll keep it simple and take an adult male's daily caloric intake of 2400 calories/8 hours of work = 300 calories/hour so watts and calories become equivalent.

Now assume a human ate nothing but the most calorie-cost efficient item on McDonald's menu, the Creamy Ranch Sauce Dip which has a cost of $3.64 per kWh ($2.49 / 460 cal * 1000 calorie/kWh conversion rate). That's more than 10x the cost for equivalent energy input. Of course people don't live off of McDonalds alone, but this is a high calorie item due to its high fat content (in reality humans need to buy tons of low calorie food items).

Of course this is all pointless discussion unless you look at actual work getting done per watt, or really work done per dollar. But 300 watts is similar to what a desktop computer achieves (except food costs are terrible compared to electricity costs). It also highlights how much overhead humans have.

 

[Endymio]

20 watts is the energy usage of the human brain, not the whole body. 100 watts is the whole body (at rest only), and we work 8/24 hours so in effect the actual operational efficiency only in energy costs is 300 watts (15x what you suggested). Now you have to look at food costs vs electricity costs....

An excellent mathematical analysis. The one thing I'd adjust is that modern workers don't work 8 hours/day. The average in the US is a little over 2000 hours per year (22%) whereas in France it's more like 1600 (18%). And that, of course, assumes you're working at 100% efficiency every minute of a work day.

Then you must factor in that the average human doesn't even start working a serious career until age 20 -- sometimes age 30+. A huge expense in training ... and then they retire at around age 60 (often age 50 if they're a gov. worker), and live an average of 15 (or 25) years beyond that. So the total lifetime-adjusted work efficiency of the human brain is closer to 7% for private-sector workers, and likely in in the 2-3% range for government workers.

 

[yRaz]

20 watts is the energy usage of the human brain, not the whole body. 100 watts is the whole body (at rest only), and we work 8/24 hours so in effect the actual operational efficiency only in energy costs is 300 watts (15x what you suggested). Now you have to look at food costs vs electricity costs. Let's assume a ridiculously high electricity cost of $0.35 per kWh for computers. For food, we'll keep it simple and take an adult male's daily caloric intake of 2400 calories/8 hours of work = 300 calories/hour so watts and calories become equivalent.

Now assume a human ate nothing but the most calorie-cost efficient item on McDonald's menu, the Creamy Ranch Sauce Dip which has a cost of $3.64 per kWh ($2.49 / 460 cal * 1000 calorie/kWh conversion rate). That's more than 10x the cost for equivalent energy input. Of course people don't live off of McDonalds alone, but this is a high calorie item due to its high fat content (in reality humans need to buy tons of low calorie food items).

Of course this is all pointless discussion unless you look at actual work getting done per watt, or really work done per dollar. But 300 watts is similar to what a desktop computer achieves (except food costs are terrible compared to electricity costs). It also highlights how much overhead humans have.

So I absolutely lost it at the creamy ranch sauce comment.

 

[Plutoisaplanet]

So I absolutely lost it at the creamy ranch sauce comment.

I was gonna originally go with sausage biscuits but I wrote a little JavaScript for kicks to compare price to calories and that one was the best value lol. I thought it would be amusing!

 

[GeoffreyA]

20 watts is the energy usage of the human brain, not the whole body. 100 watts is the whole body (at rest only), and we work 8/24 hours so in effect the actual operational efficiency only in energy costs is 300 watts (15x what you suggested). Now you have to look at food costs vs electricity costs. Let's assume a ridiculously high electricity cost of $0.35 per kWh for computers. For food, we'll keep it simple and take an adult male's daily caloric intake of 2400 calories/8 hours of work = 300 calories/hour so watts and calories become equivalent.

Now assume a human ate nothing but the most calorie-cost efficient item on McDonald's menu, the Creamy Ranch Sauce Dip which has a cost of $3.64 per kWh ($2.49 / 460 cal * 1000 calorie/kWh conversion rate). That's more than 10x the cost for equivalent energy input. Of course people don't live off of McDonalds alone, but this is a high calorie item due to its high fat content (in reality humans need to buy tons of low calorie food items).

Of course this is all pointless discussion unless you look at actual work getting done per watt, or really work done per dollar. But 300 watts is similar to what a desktop computer achieves (except food costs are terrible compared to electricity costs). It also highlights how much overhead humans have.

Bravo. Thanks for taking the time. Now, compare the energy use of brain-in-a-vat vs. the best LLM model : -)

 

[yRaz]

Swas gonna originally go with sausage biscuits but I wrote a little JavaScript for kicks to compare price to calories and that one was the best value lol. I thought it would be amusing!

Sausage biscuit wouldn't have been as funny as the ranch sauce. But the sausage biscuits are FIRE, they are a blue collar staple. I will order like 4 or 5 of them and eat them in the truck in the way to work. And I have no idea how I'm not fat, but I do masonry for a living and it is one of the most calorie intensive trades. I also often skip lunch and sometimes passout before dinner

 

[airbornex]

Sam Altman going on record to say he's glad the AI jobs apocalypse isn't happening, followed within days by 1,000 Wix layoffs explicitly citing AI, is the kind of timing that would get cut from a screenplay for being too on the nose.

 

[GeoffreyA]

20 watts is the energy usage of the human brain, not the whole body. 100 watts is the whole body (at rest only), and we work 8/24 hours so in effect the actual operational efficiency only in energy costs is 300 watts (15x what you suggested). Now you have to look at food costs vs electricity costs. Let's assume a ridiculously high electricity cost of $0.35 per kWh for computers. For food, we'll keep it simple and take an adult male's daily caloric intake of 2400 calories/8 hours of work = 300 calories/hour so watts and calories become equivalent.

Now assume a human ate nothing but the most calorie-cost efficient item on McDonald's menu, the Creamy Ranch Sauce Dip which has a cost of $3.64 per kWh ($2.49 / 460 cal * 1000 calorie/kWh conversion rate). That's more than 10x the cost for equivalent energy input. Of course people don't live off of McDonalds alone, but this is a high calorie item due to its high fat content (in reality humans need to buy tons of low calorie food items).

Of course this is all pointless discussion unless you look at actual work getting done per watt, or really work done per dollar. But 300 watts is similar to what a desktop computer achieves (except food costs are terrible compared to electricity costs). It also highlights how much overhead humans have.

I mentioned money, yes, and there the lines are blurry—at least to me, whose maths is bad. But the gist of my argument was that the brain, taken in itself, uses drastically less energy (the universe's currency) to perform a task (calculations) than a contemporary LLM model running across X amount of GPUs. I don't doubt that getting round the energy cost will require biological computing, in emulation of Nature. A bit like comparing ENIAC to a modern low-power CPU.

 

[SirRonWhite]

A lot of it is also investment capital. Ability for a corporation to get loans at certain %s.
If you are not laying off and embracing AI, financial institutions act like you do not exist.
The billionaires shaping ... sadly our future... are hyper focused on AI, even if there is no current profit results to be seen.
Time to take up a trade and be ready for AI to tell you what to do, how to live, and when to take a ****.

 

[Endymio]

But the gist of my argument was that the brain, taken in itself, uses drastically less energy (the universe's currency) to perform a task (calculations)

And the gist of our argument is that employers don't utilize racks of naked human brains in jars on a shelf, and thus your argument is irrelevant. What matters is "total cost of ownership", which for human employees is extraordinarily high.

You've also forgotten the tremendous speed advantage of ML models. When you need an analysis of blood-work and imaging to diagnose an obscure acute illness, a routing plan to direct tomorrow's fleet of thousands of delivery trucks, or a split-second targeting decision during wartime -- what good is a human decision several days later?

I don't doubt that getting round the energy cost will require biological computing, in emulation of Nature.

No, but it will require circuitry much closer to how the human brain operates:

https://www.lanl.gov/media/publications/1663/1269-neuromorphic-computing

 

[GeoffreyA]

And the gist of our argument is that employers don't utilize racks of naked human brains in jars on a shelf, and thus your argument is irrelevant. What matters is "total cost of ownership", which for human employees is extraordinarily high.

And what is the TCO of using today's ML models at scale? (I don't know.)

You've also forgotten the tremendous speed advantage of ML models. When you need an analysis of blood-work and imaging to diagnose an obscure acute illness, a routing plan to direct tomorrow's fleet of thousands of delivery trucks, or a split-second targeting decision during wartime -- what good is a human decision several days later?

No doubt, it's faster in many tasks.

No, but it will require circuitry much closer to how the human brain operates:

https://www.lanl.gov/media/publications/1663/1269-neuromorphic-computing

Yes. I have noted, in the course of my comment history, that today's AI models lack many features of the brain: real-time training and persistent state, the faculties other than language, the sense modalities, and so on. That's one, key side of the coin. I also believe that, to match and surpass the brain's efficiency, the system will have to be "fabbed" in a biological medium.

In principle, it's the architecture that counts, which, during testing, could be virtualised, running on conventional C/GPUs, and later fabbed/printed/grown/generated in the biological medium for deployment.

 

[bviktor]

Self-fulfilling prophecy at its finest

- aimlessly spend $ billions because muh AI
- it doesn't solve anything but you can't afford paychecks anymore either
- oooh we told you, AI WILL crush the job market!

 

[Northy]

Wix market cap drops sharply over 12 months, CEO claims it's not a problem as AI is saving them as he lays off employees.

Suuure.