Oracle’s head count drops 13% according to latest annual report. Many Risk Warnings
Key Headcount Details
- As of May 31, 2026, Oracle employed approximately 141,000 full-time employees (down from ~162,000 as of May 31, 2025, a ~13% or ~21,000 reduction).
- U.S. Headcount: ~49,000 (down from ~58,000)
- International Headcount: ~92,000 (down from ~104,000)
- Severance/restructuring costs ~$1.84 billion
Oracle’s SEC Filings
While Oracle cut ~21,000 jobs (13% of its workforce) as part of its AI-focused streamlining, the company is simultaneously warning investors about the high costs and competitive risks of its aggressive AI bet.
Excerpts from 10-K Risk Factors
- Our AI products may not operate as anticipated, which could adversely affect our reputation, revenues and profitability. Machine learning and AI, including generative AI, agentic AI and LLMs, are increasingly driving innovations in technology, and AI technology and services are highly competitive and rapidly evolving.
- We are building AI into many of our product offerings and we are also making AI available for our customers to use in solutions that they build. We have made significant investments in AI initiatives, including investments in infrastructure and headcount, and we expect to continue to invest significant resources to build and support our AI products in support of our growth strategy.
- In addition, we rely on partners and suppliers to produce some of our AI products.
- If we are unable to introduce new AI products, or if our AI products fail to operate as anticipated or as well as competing products or otherwise do not meet customer needs, if our competitors’ AI products achieve higher market acceptance than ours, or if we incur costs higher than expected to build and support our AI products, we may fail to recoup our investments in AI and our business and reputation may be harmed.
- Further, if we do not continue to invest significant resources to develop and support our AI products, we may fall behind technological developments and evolving industry standards, which would likewise harm our ability to compete.
Oracle Won’t Be Alone
Oracle is not an outlier. It is part of a broad industry-wide shift.
In 2026, major tech companies are simultaneously slashing headcount and pouring hundreds of billions of dollars into AI infrastructure, data centers, chips, and model development.
The pattern is consistent: cut costs in traditional operations to fund the massive capital expenditures required for AI competitiveness.
Why Other Companies Doing the Same
- Massive AI CapEx Needs: Hyperscalers (Meta, Microsoft, Google/Alphabet, and Amazon) are projecting combined AI-related capital spending of $650–700+ billion in 2026. This funds enormous data center builds, GPU clusters, custom silicon, and power infrastructure. To protect margins and satisfy investors, companies are offsetting these huge investments with workforce reductions and efficiency gains.
- Recent Examples Across Big Tech:
- Meta: Announced ~8,000–10,000 job cuts (about 10% of workforce) explicitly to free up resources for AI spending (including $115–135 billion in projected 2026 capex).
- Amazon: Cut ~16,000 corporate roles in early 2026 (on top of prior reductions), citing streamlining and AI-driven efficiency.
- Microsoft: Offered voluntary buyouts to thousands of employees while ramping up heavy AI investments.
- Others: Atlassian, Salesforce, Snap, Cloudflare, GitLab, and Coinbase have also announced significant percentage-based cuts tied to AI restructuring.
- Broader Trend: Over 150,000 tech jobs have been cut in 2026 so far, with AI/automation cited as a factor in a large portion of them. Companies are flattening management layers, automating repetitive work, and redirecting headcount toward AI-related roles where possible — but overall headcount is declining as AI tools allow smaller teams to do more.
- Investor Pressure + Competitive Race: Wall Street rewards companies that show discipline in non-AI spending while demonstrating aggressive AI bets. Those that fall behind on AI infrastructure risk losing market position in cloud, search, advertising, enterprise software, and more.
Three Key Conclusions
Oracle’s 13% workforce reduction, $1.84 billion in restructuring costs, and simultaneous AI risk warnings reflect a playbook that most major tech firms are following.
The winners in the coming years will likely be the ones that most effectively balance aggressive AI investment with leaner, more productive organizations.
This wave of “AI streamlining” is still in its early stages.
AI was one of the reasons for over employment a few years ago, so this is expected.
Makes sense as Microsoft is down about 30% because of its capital expenditures with AI.
“This is the end, my only friend the end.” — Jim Morrisson
This can be great. An outsider can never figure out an Oracle database structure, let alone what data is in it.
Unlike text search, the database is intimidatingly closed.
Converting the database into searchable indexable text could be useful if that is the goal.
Either they are making immense gains in productivity at these firms or they are cannibalizing themselves ahead of an anticipated big meal.
Or you don’t understand what is actually happening, which is my bet.
What is “actually happening” that isn’t understood?????
of course they are cannibalizing themselves in anticipation because what is passing for “artificial intelligence” are actually just large language models that are not intelligent in the least.
I’m shocked that Mish doesn’t see the mother of all malinvestment here, but alas.
AI is great. I’ve been using AI stuff (neural networks) for many years. The frontier-level AI is astonishingly amazing. But here’s a thing to keep it mind. It can be wrong. It can be really, really wrong, in curious and sometimes comical ways. I used TensorFlow to built networks, and then built one just with Python (not even using PyTorch – just the NumPy math stuff) – and then I had Gemini help me expand it’s depth, so that the neural-net had more layers. Gemini is Google’s AI thing. It wrote some very fine code for me. The stuff worked good – I could get indications of when to put on these mean-reversion, simple trades. Did many in a row. Ran my account up from significant losses (because I am not a very good trader), back to a nice big number. After a sharp up-move in the stock I had been specializing in, the AI forecast a downturn. Seemed a reasonable call. I closed out, with a nice 4-figure profit, and thought I was a clever boy. Except the AI was wrong, and I was wrong, because I am kind of stupid. I was wrong – and the AI was even more wrong. The AI can be *really* wrong – just like a human can be. And I have a sneaky feeling, that as the frontier-level AI’s are enhanced, their ability to be *really wrong* (like a human can be, for similar reasons), is probably going to be amplified.
We are going to be using them (the big AI’s) for everything, I fear. And when they are wrong, their perfect wrongness will get us all killed – since we will be lulled into doing dumb things that we might not have done, because we (wisely) felt we could not know the outcome. The AI’s will make us dangerously over-confident. The big position I sold, has continued to zoom up, and up and up. If I had not been so damn confident, I would not have traded out. Losing my good (and very big) position, has been a soul-crushing experience. I don’t blame the AI stuff. I just did a really dumb thing. But what’s interesting, is that the AI power (and there is real power in a good AI), is going to make us all do crazy sh/t that we would not normally do. And as knowledge workers are let go, and big orgs get lean and mean, and dependent on their AI’s, that “really wrong” feature of AI is still gonna be there. Just like happens with human wrong-decisions, the AI wrong decision will be implemented, and there will not be enough mavericks left in the Big Org, to actually stop the launch sequence. The AI will effectively be in charge. And it will make a mistake, and whack everything.
This will happen. I just can’t say when and where. 🙂
The question I have, is if you say AI was correct in getting you to buy a stock, but wrong when it told you to sell, how do you know that AI was correct in getting you to buy in the first place? Sounds like a 50-50 coin toss to me.
We only fix the things they get wrong that we actually catch. The more subtle hallucinations are where the permanent danger is.
When everything is bid up by AI bots, the bots cannot be wrong. Only the sky is the limit. Is it what’s behind the recent stock highs?
AI will learn from its mistakes, unlike most humans, and self-improve until it doesn’t make any mistakes!
How will it know it made a mistake ?
Ah fair JoJo, life’s a mystery to ye
This was an excellent read. TY.
As everyone else has already pointed out, this smacks more and more of dot com by the minute. It’s practically written on the walls! Unlimited free samples for all, no actual path to profits, investor money being spent galore, regulation looming behind, and circular financing out the wazoo.
Because most corporate customers right now aren’t paying the actual cost of it. It is operating on the all-you-can eat buffet model in hopes people don’t eat more than it cost to put the food out. The minute things get billed per-token…
https://arstechnica.com/ai/2026/06/anthropic-pauses-token-based-billing-for-its-claude-agent-sdk/
https://arstechnica.com/ai/2026/04/github-will-start-charging-copilot-users-based-on-their-actual-ai-usage/
People cancel their subscriptions fast. The gains go away. And the future looks like small models hosted locally.
Mak’s comment, quoted again below, hits the nail on the head:
I can run Qwen 3 on my computer for free, Qwen does exactly what I need out of an AI tool running on my own hardware with a fraction of the resource waste. So, why would I need a subscription to Claude or CoPilot? Not to mention the censorship from those models that I don’t get with my own.
No, what you see here is a collection of the greediest companies on earth trying to turn a society enhancing technology into a monopoly with increasing desperation.
The Next 5 Years: A Supersonic Tsunami
Peter H. Diamandis
Jun 21, 2026
https://metatrends.substack.com/p/the-next-5-years-a-supersonic-tsunami
The smoke test will be the quality of the product produced.
The AI boom, the modern day version of the gold speculating 49ers, is like the WSOP. Very few will make the final table and along the way the ones that made the right moves or were luckiest will take up the chips (pun intended!) from the less fortunate, leaving only a handful left playing. The Oracle report basically says “if we aren’t sitting on a sufficiently large stack of chips, if we made a misstep along the way or if our competition makes more rewarded steps, we’ll be out of the WSOP, and by out we mean all the way”. It seems they are all playing for all or nothing; not sure if that’s the only play in IT but that’s what it appears.
The only winner in the AI space is looking to be China. They have the best cost LLMs and they just released GLM 5.2 which everyone is talking about. It’s claimed to be as good as Anthropic’s Opus but only cost 20% of what Anthropic charges.
China also has the energy (wind, solar, hydro) to power real datacenters while US folks pull a NIMBY on datacenters, windmills or solar farms. Doesn’t help the clown in charge is anti-green energy.
https://www.youtube.com/watch?v=iJw3NxTqP5g
you forgot coal, the fuel whose energy output they quadrupled, to get to their “green” place. what if the U.S. could just quadruple its coal use, care zippity squat for the environment while they did so, and then used that to build the green power you mentioned? you for that? cuz that’s what they did. don’t pretend they just developed it in a vacuum.
Maybe China can be your exit strategy.
I won’t be moving to China but I will be leveraging every cheap thing they got from EVs, AI, pharmaceuticals, to manufactured goods. Only chumps and fools pay more and get less.
I pay more for quality car tires and for quality leather shoes. I am still using one pair of Italian leather shoes I bought in 1984. The Vibram soles were replaced four times. Think how many car crashes I avoided by paying more for tires. Chump???
Where were those tires made? I’m sure you’ll say USA but not understand that rubber doesn’t come from the USA nor most of the steel used in the belts, nor the chemicals used to make them.
Your 80% China component tire made in the USA is definitely worth all the extra money you pay! /s
Not AI streamlining, but AI washing.
“The winners in the coming years will likely be the ones that most effectively balance aggressive AI investment with leaner, more productive organizations.”
I find this to be unclear. I assume that the “winners”, in this case, are the ones that make the most profit/money. At this point. for me, it is unclear whether AI will generate more revenue that human labor – but that is the “balance” I believe is implied. I wonder if the producers of the “best” AI (define best as you wish) will “win” in a meaningful way. Is an organization of 100% AI “bots” going to generate more money? Is one that is 100% human labor going to generate more profit? Is 50/50 (whatever that means) going to “win”? Is the “balance” going to different for each sector? for each organization? for each dept? for each individual (whatever that means)?
There are a lot of questions – and the best I can do now is guess about the future and operate in the reality that exists right now.
The stark reality that no one wants to face is that economic models the world has been using up to now are going to be killed off in a future of unlimited abundance, where AI controls the manufacture and distribution of all necessities to humans and its robot workers toil tirelessly 24 x 7 for the benefit of humans, without complaint, and for no wages or benefits.
In this post-scarcity world, there will be no meaning or need for concepts like money, profit, entrepreneurship, assets, ownership or human labor. There won’t be poor people nor will there be billionaires. Housing will be free. Healthcare will be free. Food will be free. Trips around the world will be free.
EVERYTHING WILL BE FREE, FREE, FREE AND EVERYONE WILL BE EQUAL
Or… hear me out… we waste most of our resources and fight each other over the dwindling remnants.
It’s difficult to know if Jojo is being facetious or simply repeating the AI pitch?
Paradise on earth? The New Jerusalem cometh?
NOT facetious at all! This IS the future. You and many others lack the imagination needed to envision a world different form what you have been indoctrinated into.
This vision arrives to me as a fantasy. The powerful (currently the rich) will not easily or willingly cede their advantage.
What is more likely is a continuance of false scarcity controlled by the upper class elite, while the poor and downtrodden sink lower (which is hard to imagine but, for me, likely).
As opposed to everything being free – air, water, sunlight – etc etc will become things that are “charged” for.
Try to focus! You cannot control that which is not scarce, that which if you can’t get it here, you can easily get it over there at the same or less cost.
Many studies have shown that the best return on investment in IT are firms that follow after innovations have become mature and are true and tested.
Organizations that lead are usually doing so to prove they are “ahead”, but apart from marketing this image, it is unclear whether they benefit. Don’t forget lots of organizations fail as the end result of failed IT projects — not working “solutions” are often fatal.
The part about the AI “Boom” I’m having trouble processing is where will the scarcity come from to generate the huge profits and revenue needed to pay off all this these massive investments in AI Cap-Ex?
Agreed. AI or more precisely LLM are already delivering productive gains to end users. However, how big tech will make a profit out of it is hard to see. Most LLMs are various flavours of the same concept. Some are better than others at certain tasks but most companies don’t have a clear pathway to being clear ‘winner’ of monopoly profits. It will be an oligopoly at best.
With extreme capital expenditure and unlike software development, the marginal cost of supply isn’t anywhere near close to zero. So more customers mean more costs just like regular goods and services.
The winners will be society and the losers will be investors.
You just saw it in the article…permanently reducing expenditures on salaries for people that do nothing all day but sit around and gossip become a perpetual annuity from layoff day = 1 and those savings flow right to the bottom line (i.e, net income, ie, profit) once the severance break-even is reached.
You could use AI to replace every HR department (pure cost center, generate zero income) in every company in the country.
Moronic.
What happens when the HR department has to make a judgement call about a human being?
I assume you want bots to be your jury as well
What’s happening in AI now is not much different from what happened to businesses in the early 90s when computers loaded with Microsoft products (word, excel, project management software etc) started appearing en-mass on every employees desktop to increase productivity via internal corporate networks.
I remember in my first couple jobs after graduating (88) that most employees (secretaries, receptionists, warehouse guys etc) didn’t have a PC on their desk and had no clue how to use one (I still remember interoffice mail being delivered by a person). By 92 or so everyone had one and lots of jobs were disappearing then as well as companies were embracing internal corporate networks. That same thing would play out again in the late 90s early 00s with the internet.
Lots of mistakes will be made (adopt too early, adopt too late, pick wrong technology) and the losers will probably go out of business just like they did then.
VP: “secretary get in here”
Secretary: “yes sir”
VP: “take a memo, please”
Secretary: “go ahead sir”
VP: “blah, blah blah, yada yada yada”, now read that back
Secretary: “blah, blah blah, yada yada yada”
VP: ” now take that down to the typing pool”
Secretary: “yes sir”
……
…..
…..four hours later memo returns, and God forbid there’s a typo
+++++++++++++
Today: VP Chud opens up Outlook, types up a memo, clicks send (10 minutes) — no typing pool, no secretary, etc etc etc
You should have gone back further in your timeline to when the secretary sat on the bosses lap to take a memo!
Any your current take labeled “Today” is incorrect also.
Today, boss enters a sentence or two prompt to an AI instructing it to create a memo on a subject, gets finished memo 20 seconds later, copies memo into email, clicks send. Takes a longer lunch after such hard work
If you use GMail, you see a query whenever you type an email to the effect of “hit the tab key to improve” or similar.
It will be interesting if the bubble pop is actually caused by the companies themselves due to AI over investment
as opposed to the usual hype driven malinvestment of the investment community
warsh might be onto something
We don’t why are we doing it, but everybody else is doing it, so we have to do it, and maybe it somehow makes sense.
History rhymes once again:
“The Era of Growth Over Profit:The shift toward valuing companies based on their cash burn rate rather than immediate profitability became prominent during the late 2000s through the 2010s, often referred to as the “growth at any cost” era. This period saw a paradigm shift in investor and entrepreneur mindset, moving from the traditional belief that “cash is king” to prioritizing rapid user base expansion and market dominance over short-term profits.
This approach was heavily influenced by the massive success of companies like FB and YT, which deliberately operated at a loss for extended periods to secure a dominant market position before monetizing their platforms. Investors began viewing a high burn rate positively if it correlated with significant user acquisition and potential long-term market control, particularly in internet-based and platform businesses where network effects are crucial.
Prominent Examples of Investor Losses:While the “burn rate” strategy led to successes like FB and Uber, it also resulted in significant losses for investors when companies failed to achieve profitability or market dominance. Notable examples include:
Definition of Oracle ? A Deity possibly revealing hidden Knowledge of the Future?
Oracle and other A.I. companies are cutting staff as they need more money redirected to funding those ever increasing A.I. development costs. But does Oracle also expose the final parts of an A.I. bubble nearing a pop?