Your daily AI briefing. Four stories that matter, explained like we are talking at a bar. June 10, 2026.
AI STOCKS GO INTO FREEFALL: BROADCOM MISSES BY ONE BILLION DOLLARS AND TAKES THE ENTIRE CHIP INDUSTRY DOWN WITH IT
So here is what happened. Broadcom, a company that most people outside the tech industry have never heard of but which is quietly one of the most important AI chip businesses on the planet, released its quarterly earnings last week and told investors its AI chip revenue for the next quarter would be around $16 billion. Analysts expected $17.2 billion. That is roughly a $1.2 billion gap between what the market hoped for and what actually materialized. In a normal industry, in a normal year, you might look at a $16 billion quarter and say that sounds fine. But we are not in a normal industry and we are not in a normal year, and that gap was enough to set the entire stock market on fire.
Broadcom stock fell about 15% in a single day. That is brutal on its own. But the real story is what happened next. The selling spread like a fever across every chip stock in the market. Micron fell 17%. Intel dropped 9%. AMD lost 12.6%. The Nasdaq, the index that basically tracks where the AI economy is betting it will go, fell 4.18% in a single session, its worst day since April of last year. The Philadelphia Semiconductor Index, which is basically a scoreboard for every company that makes chips, plunged 10.3% in one day, which is its worst performance in over six years. This was not a Broadcom story. This was the market asking out loud whether the AI spending boom has already peaked.
And there is a specific reason this particular company miss hit so hard. Broadcom is one of the main manufacturers of custom AI chips for Google, called tensor processing units. An analysis from Macquarie came out shortly after earnings warning that Broadcom’s share of Google’s chip work is expected to drop from around 95% right now to 80% in 2027 and 65% in 2028. That is a slow bleed happening in plain sight. Meanwhile HSBC analysts pointed to falling chip prices and signs of a slowdown in AI spending as their biggest concerns going forward.
None of this means the AI era is over. It means the era of unlimited investor patience with AI companies that promise explosive growth and then deliver merely very good growth might be tightening up. The companies doing the spending, the Googles and Microsofts of the world, have been writing checks at a pace that assumes every dollar going in will come back as ten. When a company like Broadcom suggests maybe the pace is not accelerating anymore, investors start asking questions nobody wants to answer in public. What happens to the AI narrative if the chip demand actually plateaus? We are about to find out.
TAIWAN JOINS THE SIEGE: THE ISLAND THAT MAKES THE WORLD’S CHIPS IS NOW MOVING TO BLOCK THEM FROM REACHING CHINA
Source: Bloomberg
You already know the United States has spent the last few years building a wall around advanced AI chips to keep them out of Chinese hands. Washington restricted Nvidia’s most powerful chips. Then it restricted slightly less powerful chips. Then it got creative about restricting chips that were specifically designed to sneak under the restriction threshold. The whole thing has been a slow, grinding regulatory arms race between American export control officials and Chinese buyers who really, really want the hardware. What changed this week is that Taiwan, the actual island where most of these chips are manufactured in the first place, has now decided it wants in on the blockade.
Taiwanese authorities said they are weighing much stricter export controls on AI chips sold to China, aligning more closely with US rules. The reason is chip smuggling, which has apparently become a growth industry. Earlier this year, Taiwan arrested three people for forging export documents to move Nvidia AI chips into China, the island’s first actual criminal prosecution for semiconductor smuggling. When you are arresting people for forging paperwork to move graphics cards across borders, you have a problem that calls for more than strongly worded guidance.
The irony sitting underneath all of this is thick enough to cut with a knife. The strategy of keeping advanced AI chips away from China was supposed to slow Beijing down in the AI race. The results have been mixed at best. DeepSeek trained a competitive AI model on hardware that was supposed to be restricted and did it at a fraction of the cost American companies spend. Nvidia itself admitted just last month that it has largely conceded the Chinese AI chip market to Huawei. So the export controls have not stopped Chinese AI development. They have rerouted it, improvised it, and in some cases accelerated it by forcing Chinese engineers to optimize harder than they otherwise would have had to.
And still, the coalition around restricting chips is growing instead of shrinking. Taiwan joining the effort is significant not because it adds much practical enforcement power, given that American rules technically already apply to chips manufactured on Taiwanese equipment using American technology. It matters politically. It signals that the governments closest to the supply chain are aligning behind the idea that advanced AI hardware should not flow freely to Beijing. Whether that alignment actually slows the technology transfer, or just adds more paperwork to the people doing the smuggling, is a question that will probably answer itself within the next two years.
APPLE CLIMBS INTO BED WITH GOOGLE AND NVIDIA AT THE SAME TIME AND CALLS IT PRIVATE CLOUD COMPUTING
Source: CNBC | TechCrunch
Apple spent years building a reputation as the company that does not trust anyone else with your data. Do not use Google. Do not use Microsoft. Do not let your information leave the device. Privacy as a brand value, privacy as a product feature, privacy as a reason you should buy an iPhone over an Android. And now Apple has announced that its most advanced AI model, the one doing the heavy lifting in the revamped Siri, will run on Nvidia GPUs housed inside Google’s cloud infrastructure. Same Apple. Same privacy promises. Different architecture than the one those promises were built on.
Here is what Apple announced at WWDC this week. The company has a new model it is calling AFM Cloud Pro, which is their frontier-grade AI that handles the most complex requests Siri needs to process in the cloud rather than on your device. According to Apple, this model is comparable in quality to Google’s own Gemini frontier models, which is a bold claim they dropped somewhat casually. And this model runs on Nvidia hardware sitting in Google’s data centers. Apple says it maintains its full privacy guarantees through a system it calls Private Cloud Compute, which is designed to ensure that your data is processed but not stored or accessed by Google or Nvidia.
Whether you believe that is partly a technical question and partly a question of how much you trust three of the largest technology companies on earth to have agreed not to look at your stuff. Apple has staked its cloud privacy architecture on a model where the cloud node processes your request in a sealed environment and then forgets it happened. Security researchers have examined earlier versions of this system and generally found it to be more rigorous than typical cloud setups. But putting your most private AI requests through Google’s infrastructure, running on Nvidia hardware, is a different proposition than Apple’s previous story of keeping everything on the device or in Apple’s own data centers.
This is also Tim Cook’s last WWDC. He announced he is handing the CEO role to John Ternus on September 1, and this conference is his final big stage appearance as the man running Apple. The legacy he is leaving includes a company that came late to AI, fell behind embarrassingly on Siri while ChatGPT ate the cultural moment, and is now spending billions and partnering with its own competitors just to get back into the game. Whether AFM Cloud Pro actually closes the gap on Gemini and GPT-4 in real-world use will become clear when iOS 27 ships in the fall. But the fact that Apple had to call Google and ask to borrow their data centers to make it happen is a detail that the old Apple would never have allowed anyone to report.
THE STUDY YOUR BOSS NEVER WANTED YOU TO SEE: COMPANIES FIRING PEOPLE FOR AI ARE NOT ACTUALLY MAKING MORE MONEY
Source: Fortune
Gartner surveyed 350 large companies, all of them with annual revenues of at least a billion dollars, and asked them whether their AI-driven automation had actually paid off. Eighty percent of the companies that had piloted AI and then cut workers said they had reduced their workforce. That part was not surprising. That is what everyone has been reporting, and it matches what you have seen in your industry, your city, or your LinkedIn feed over the last eighteen months. The part that should make every CFO who signed off on an AI-justified layoff uncomfortable is what the study found next: the cuts were not generating the returns that were supposed to justify them.
The companies that actually got better results from AI were the ones using it to amplify what their employees could do, not the ones using it as a reason to have fewer employees. Helen Poitevin, the Gartner VP analyst who led the research, put it plainly: chasing value only through headcount reduction is likely to lead most organizations down a path of limited returns. That is analyst-speak for: you fired your writers and your customer service team and your junior analysts and replaced them with AI subscriptions and the numbers are not adding up the way the pitch deck said they would.
This has been a sneaking suspicion building for a while. A CFO survey published earlier this year found that finance chiefs were privately expecting AI-driven job cuts to be nine times higher than what they were publicly announcing, which tells you two things: first, the real scale of what is coming is being deliberately obscured, and second, there is enough uncertainty about whether any of it will work that executives are not exactly eager to make promises in writing. You do not hide your projections if you are confident in the outcome.
The deeper problem the Gartner study identifies is the gap between perceived and actual productivity gains. Companies believe AI is making them significantly more productive. The measured reality is more modest. Part of this is timing: the benefits of AI adoption tend to take longer to show up in revenue than the costs show up in the headcount numbers. But part of it is also that replacing a human with an AI system is not always a straight swap. You lose institutional knowledge, client relationships, the kind of judgment that does not appear in a job description. Some companies are discovering this the hard way after the fact, which is not a great way to run a business that cost-cut its way into a worse competitive position.
The companies winning right now are using AI to help their people do more, move faster, and work on higher-value problems. The companies losing are the ones that convinced themselves the AI subscription fee was cheaper than a salary and then found out the math only works if the AI actually does the full job, which most of the time it still does not. The study is not saying AI does not work. It is saying the strategy of treating AI as a way to get rid of people is a losing bet, and the scoreboard is starting to prove it.
Quantum Beat runs daily. More at aiuntmedia.com