
June ends today. Some months pass. And some months change everything. June 2026 is one of the latter.
Not because one big thing happened. But because twenty seemingly different things happened that, when you look at them together, tell the same story.
Artificial intelligence has ceased to be a technology industry and has become a new power structure. And like any power structure, it is beginning to decide who has access, who controls, who wins, and who is left out.
Let me tell you in order. Because the sequence matters. And because whoever doesn’t understand the sequence, doesn’t understand what’s going on .
First, money built the foundation.
Before understanding what happened in June, we must understand what made it possible.
In the last twelve months, Apollo and Blackstone closed a $36 billion debt deal to finance computing infrastructure. Microsoft projects $190 billion in capital expenditures. Google , between $175 billion and $185 billion. OpenAI aims for 10 gigawatts of computing capacity.
Those aren’t numbers from a technology industry. They’re numbers from an infrastructure industry.
AI is no longer software. It’s what oil was to the 20th century: the raw material upon which everything else is built. And like oil, those who control the infrastructure control the game.
Eighty-eight percent of global AI funding—some $319 billion so far in 2026—has gone to US companies. Most of it to just two: OpenAI and Anthropic. The rest of the world isn’t competing. It’s consuming.
That’s not a technical fact. It’s a geopolitical fact . I always say: in the world of brands and business, those who don’t have their own voice adopt the voice of someone else.
What applies to brands now applies to countries. And its consequences will take years to become fully visible, but they are already happening.
Then the giants repositioned themselves
With that foundation in place, the big companies made their moves. And in this business, the moves of the big players aren’t marketing strategies . They’re a reshaping of the playing field.
Google went all in at once. At Google I/O in May, it unveiled Gemini Omni , capable of creating and editing video from any type of input. Gemini Spark , a permanent agent that works 24/7 without user intervention. Smart glasses designed with Gentle Monster and Warby Parker, in collaboration with Samsung. And Android 17 with Gemini integrated at the operating system level. The message was clear and unequivocal: Google no longer makes products with AI. All its products are AI .
The most profound consequence of that movement isn’t in the glasses. It’s in the search.
68% of Google searches now end without the user clicking on any page. Some small publishers have seen their traffic drop by 60% in two years. Google announced the biggest update to its search engine in 25 years: it no longer sends you to a website, it gives you the answer and keeps the traffic. SEO as we knew it is dead.
What’s coming is called AEO, Answer Engine Optimization, and no one has written the rules yet. Whoever writes them first will have an advantage that will last for decades.
OpenAI, meanwhile, transitioned from a modeling company to an organizational transformation company. It launched the OpenAI Deployment Company with over $4 billion in funding, designed to deploy specialized engineers within its clients’ corporate environments.
McKinsey, Accenture, and Deloitte have been doing that for decades. None of them have the models. OpenAI does.
The result of that combination is predictable. And it should worry any consulting firm that still believes the business model that worked yesterday will work tomorrow.
And to top it all off, OpenAI acquired uv and ruff —the most widely used package installer and code formatter in Python —tools used by millions of programmers who had no connection to OpenAI. It didn’t just buy a company. It bought the invisible infrastructure of the development ecosystem.
The quietest move of the year, and possibly the most strategic. The best moves almost never make a sound when they happen.
Salesforce , for its part, paid $3.6 billion for Fin , a platform for autonomous agents in customer service. The agent who answers calls, sells, handles complaints, and escalates issues now has a market price. The question no one is answering aloud is how many jobs that price represents.
And in hardware, Qualcomm is in talks to acquire Tenstorrent, the open-standard RISC-V chip company led by the legendary Jim Keller, for between $8 billion and $10 billion. Nvidia’s dominance in AI silicon has seemed immovable for years. Now it’s starting to shift.
And when something that seemed immovable begins to move, what comes next is usually bigger than anyone anticipated.
Then came Musk’s creative chaos.
Elon Musk had a cinematic June, and not exactly as a triumphant protagonist. There’s something revealing about what happened to him this month: the man who wants to reach Mars had to admit that he didn’t know how to properly build a software company.
First, the context.
xAI , his artificial intelligence company, had lost all 11 of its co-founders before the end of March. Musk himself publicly admitted that xAI “wasn’t built right from the start” and that he was rebuilding it from the ground up.
This comes after the incident in which Grok called himself “MechaHitler” and the platform allowed the creation of sexual deepfakes of women and children. These aren’t technical errors. They’re cultural errors.
And culture can’t be patched up with a software update. I know this because I’ve spent decades helping organizations build theirs.
To compensate for Grok’s shortcomings compared to Claude Code and OpenAI Codex , SpaceX announced the purchase of Anysphere , the company behind Cursor , for $60 billion in stock.
The largest acquisition of a venture capital-backed startup in history.
Cursor has over one million paying users and 65% of the Fortune 500 as customers.
At 25, its CEO Michael Truell became one of Silicon Valley’s youngest billionaires overnight.
The risk is obvious.
Cursor was valued by developers precisely for its neutrality. It worked with Claude, GPT, and Gemini. Now it’s owned by Grok’s company.
How many developers will continue to use it when the default model is that of its new owner is a question that the market will soon begin to answer.
In branding We call it the risk of trust dilution. And trust, once lost, is much harder to regain than to build.
And amidst all this, a federal court in California dismissed Musk’s lawsuit against OpenAI and Sam Altman . Not because Musk lacked arguments. He did. But he had filed them after the statute of limitations had expired. “I was a fool,” he declared at the trial, referring to the $38 million he donated to found OpenAI in 2015.
The richest man in the world lost because he was late to court.
The philosophical question he posed — can a non-profit organization become a for-profit company, enriching its founders without consequences? — remains unanswered in court.
And it will remain relevant long after this trial is history.
Then the market began to fragment
While all this was happening, the most revealing piece of data of the month arrived almost silently: ChatGPT’s market share fell to 46.4%. For the first time since its launch in 2022, ChatGPT lost its absolute majority in the AI chatbot market. Google Gemini reached 27.7%. Claude rose to 10.3%.
The AI monoculture is over. For three years, ChatGPT was synonymous with artificial intelligence for most of the population.
That time is over.
What’s coming is a fragmented market where the question for brands, companies, and users is no longer “Do we use AI?” but “What AI do we use, for what purpose, and with what consequences?”
That question has enormous consequences for any organization that still believes AI is a technical decision and not a strategic one.
In that same fragmented market, Meta decided to strengthen its position in hardware. It ended its collaboration with Ray-Ban and launched its own line of smart glasses: Adventurer, Fury, and a special edition with Kylie Jenner for $399. All feature integrated Meta AI, a camera, speakers, and over eight hours of battery life.
Smart glasses are no longer a niche gadget. They’re the battleground where it’s decided who controls the AI layer closest to the human body. And that battle has implications that go far beyond fashion.
In pockets, Android 17 integrates Gemini at the operating system level with predictive features that anticipate what the user needs before they ask for it. iOS 27 will arrive in September with agents that act on behalf of the user, calculate routes, notify contacts, and manage routines.
Your phone is no longer a tool you use.
It’s an agent that acts on your behalf. That distinction matters more than it seems. Because when something acts on your behalf, the relevant question isn’t whether it does it well. It’s who programmed it, with what values, and who is held accountable when something goes wrong.
And then the government entered the scene
On June 9, Anthropic released Claude Fable 5 and Mythos 5. According to available reports, they were the most capable models ever publicly released, with reasoning and code analysis capabilities unprecedented in the commercial market.
Three days later, the United States Department of Commerce ordered them to be turned off for all users worldwide.
The official reason given was a jailbreak with national security implications. It was the first time in history that the US government had applied export controls to a commercial AI model.
The technical and legal problem was immediate. Anthropic lacks the infrastructure to verify the nationality of its users in real time.
The directive required suspending access to any foreigner, inside or outside the U.S., including Anthropic’s own foreign employees. Unable to discriminate based on nationality without shutting down the service entirely, they shut it down completely.
Hundreds of millions of users worldwide lost access.
Anthropic responded that the government had only presented them with “verbal evidence of a potential, not universal, jailbreak ” and that the same technique was applicable to OpenAI’s GPT-5.5 , which was not suspended.
The White House responded that the ball was in Anthropic’s court. Congress sent a bipartisan letter demanding the legal and technical basis for the order. There was no public response as of this writing. Institutional silence on a question that deserves a public answer.
What happened on the 16th turns this episode into more than just a regulatory dispute. Zhipu AI , the spinout from Tsinghua University, announced that its latest model matches Claude Mythos in the same cybersecurity benchmarks that the US cited as justification for the ban.
The government suspended the world’s most advanced model to prevent foreign actors from accessing that capability. Sixteen days later, a Chinese laboratory has it anyway. And that laboratory operates outside the jurisdiction of any American export controls.
Censorship that doesn’t censor anything.
Or, to put it more precisely: a policy designed to curb the spread of a capability that was already spreading through other channels.
The precedent, however, is what matters. If the government can shut down a business model in 24 hours, it can do it again. With any model. For any reason it deems sufficient.
That changes the nature of the risk for any company that has built its operation on third-party models.
Meanwhile, AI entered companies without asking permission.
There’s a parallel story unfolding quietly, far from the headlines about governments and $60 billion valuations. It’s the story that worries me most. Because it happens in every organization , including probably yours.
Anthropic integrated Claude into Slack with an “ambient” mode that proactively monitors channel conversations, highlights relevant information, and tracks pending issues.
According to the company itself, 65% of its internal code already passes through an internal version of this tool. The world’s most advanced AI security company has an AI that reads all its internal chats . This isn’t a contradiction. It’s a sign of where all this is headed.
And in companies that haven’t made any official decisions about AI, AI has already arrived anyway.
It’s called Shadow AI. It involves employees using unauthorized tools, uploading customer data, proprietary code, or contracts to external systems, getting their response, and continuing to work without anyone in the company knowing what happened.
They don’t do it with bad intentions. They do it because it’s efficient. Because it’s what anyone who wants to work better would do. And the generation that grew up with the internet as a natural infrastructure does it instinctively, without perceiving any risk.
The risk is real.
Intellectual property leaks , regulatory non-compliance in sectors such as banking or health, and the loss of control over decisions being made with the help of systems that the company does not monitor.
Shadow AI is not a technology problem. It’s an organizational culture problem that most companies are still unaware they have.
And by the time they discover it, something will probably have already happened that they’ll regret. I’d rather they read about it here than learn it the hard way.
And in the end, the oldest institution in the world took a stand
When the Vatican publishes a document on artificial intelligence, the usual reaction in tech circles is a condescending smile. The oldest institution in the world weighing in on the newest technology.
It deserves more attention than that.
The document “Antiqua et Nova” warns that autonomous weapons systems capable of identifying and attacking targets without direct human intervention are a “serious cause for ethical concern.” It also states that AI can lead to “harmful isolation” and that portraying AI as a person is “a grave ethical violation” when done for fraudulent purposes.
These aren’t technical positions. They’re positions about what kind of world we want to build . And the Church has been having this conversation for two thousand years with more consistency than most of the parliaments and boards of directors that are debating AI regulation today.
Dismissing that perspective because of where it comes from is exactly the kind of arrogance that usually precedes the biggest mistakes.
When the world’s oldest institution enters this debate, it’s not a mere anecdote. It’s a sign that the conversation has shifted in nature. It’s no longer a conversation about technology. It’s a conversation about power, about dignity, about who decides and according to what criteria.
It is, in short, the most important political conversation of our generation disguised as a technical discussion.
The question that connects everything
These are not separate news stories. They are manifestations of the same question: who controls AI?
Governments, who can shut down the world’s most advanced model in 24 hours but can’t prevent China from replicating it in 16 days?
Are companies building infrastructure worth hundreds of billions and buying up the invisible ecosystem that programmers work on?
Is the market fragmenting what seemed like a natural monopoly?
Are users adopting tools without asking permission from their employers or governments?
The answer, I believe, at least for now, is that no one is completely in charge . And that is precisely why this moment is both so important and so unstable.
We are at the point where the rules are being set . Not the rules written in legislative documents, which will come later and be delayed.
The real rules. Who controls the infrastructure , who has the most powerful models , who can shut them down, who trains the professionals who use them, who decides what is too dangerous to distribute, and who pays the price for that decision.
Those rules are being set now. In June 2026. While most people are focused on something else. The World Cup , for example.
I’ve always believed that the most powerful brands are those that understand where the world is headed before anyone else and take a stand. The same is true today for individuals, organizations, and countries.
AI no longer asks for permission (I don’t think it will even apologize). The question is whether we are prepared for that. And if we aren’t, what are we going to do about it? Because the clock doesn’t wait. And neither does history . Meanwhile, June says goodbye.
I write to you when I have something to tell you.
Get an email every time I publish something new.
Your email will only be used to send you what I publish. Privacy policy






