Computers are now depreciating slower than cars — the reason is enraging



Computers and their graphics processing units are becoming hot commodities, and prices seem poised to get worse, not better.

Demand for processors is set to affect everything from gaming to run-of-the-mill memory sticks, resulting in a pincer attack on consumer wallets.

'AI demand is driving up the price.'

Blaze Media's Auron MacIntyre noted a recent price spike in handheld gaming device Steam Deck; a product that has been on the market for about four years increased in price by $300 in May.

"Over the entire history of video games, systems went down in price as they get older," MacIntyre wrote on X. "You might say 'who cares it's a child's toy' and fair enough but it's a signal of a wider trend[.] The price is skyrocketing because AI demand is driving up the price on all physical computing hardware from video processors to RAM."

The culprits in the processor gold rush are tech and AI companies buying up GPUs — which are typically used in phones, laptops, and gaming consoles — to power their data centers.

Just a few years ago, a few thousand GPUs in a single facility was considered cutting edge. Now, upwards of 100,000 GPUs are "interconnected through high-speed networking systems designed to operate as a single computational unit" inside one building, according to DataCenters.com.

RELATED: Sick of Microsoft's preinstalled propaganda on your PC? Block it now.

Over the entire history of video games, systems went down in price as they get older

The Steam Deck is 4 years old and getting an eye watering $300 price increase

You might say "who cares it's a child's toy" and fair enough but it's a signal of a wider trend

The price is… https://t.co/XrU0kW7Mnn
— Auron MacIntyre (@AuronMacintyre) May 29, 2026

While the mind may imagine an all-powerful energy source or an unseen technology at the core of these sprawling data centers, photos inside the facilities show they are quite literally thousands of computers plugged into each other, powering the demands of search engines and AI agents.

The writing has been on the wall since at least 2024; Microsoft bought 485,000 Nvidia Hopper GPUs that year. Meta bought an estimated 224,000, ByteDance bought around 230,000, and xAI purchased 200,000 (per Data Center Dynamics).

Last November, Tech Powerup noted official "warnings" from manufacturers that price hikes would be unfolding as memory costs rose over 170% year-over-year.

This prediction turned out to be accurate as GPU pricing has risen between 5% to 20% or more, according to Fusion Worldwide. The outlet notes shortages in high bandwidth memory, graphics double data rate, and dynamic random-access memory.

This has led to product lead times of an additional three to seven months.

RELATED: Out of control: Here's how a company spent $500 million on AI in a single month

Jim WATSON/AFP/Getty Images

Consumers may be well served to hoard some GPUs for themselves too. The demand is not only making prices increase, but depreciation is slower as well.

Business Insider recently cited a company selling refurbished GPUs that said in the second year of a processor's life, it only loses 15 cents on the dollar. In its third year, the same GPU is only losing one additional cent and can be sold for 84 cents on the dollar.

Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!

After fierce debate, Trump opts for federal controls in AI development



AI companies have largely developed their chatbots with very little government regulation, all in an effort to beat China to artificial general intelligence. However, as these services exploit users’ mental health, enable new devastating cybersecurity threats, and arm the U.S. military with advanced capabilities, the Trump administration recently proposed federal regulation to keep the bots in check. Now, President Donald Trump wants to take government oversight a step further by invoking the power to review AI models before they’re released to the public through an executive order that was signed this week.

President Trump has mostly maintained a hands-off approach to AI regulation, bucking attempts at state-level bills to curb development in favor of a centralized federal mandate. There are clear pros and cons to Trump’s National AI Legislative Framework, but it provides a starting point for standardizing an industry where Trump has dragged his feet.

This is why the latest reports of added AI oversight, directly from the U.S. government, come as a surprise, given Trump’s previous stance. If signed, the executive order would mark a sea change within the Trump administration, signaling that AI needs direct government intervention to protect the public from potentially dangerous models.

The question is, why?

This move raises the question: How much AI regulation is too much regulation?

Trump’s decision came after Anthropic — the same company that landed on the military’s supply chain risk list — unveiled a new AI model that was purportedly too dangerous to release to the public. Labeled as Mythos under Project Glasswing, the new model excels at leveraging computer hacking and cybersecurity exploits. In other words, it’s really good at breaking the security measures of critical digital products and services, including operating systems and internet browsers.

If left in the wrong hands, Mythos could pose a huge risk to anything and everything connected to the internet — personal devices, school computers, government systems, banking platforms, and even critical infrastructure like power grids, traffic systems, and more.

Instead of allowing the public to access Mythos outright, Anthropic opted to provide the model strictly to Big Tech companies to help them find security holes in their products before a competing AI platform on the same level as Mythos reaches public status. The goal is to patch these bugs before they are exploitable by hackers using other AI platforms. So far, Mythos has poked holes in Apple’s highly secure MacOS platform and Mozilla’s privacy-focused Firefox browser. Unfortunately, while Mythos is good at finding problems, it’s bad at patching them, with recent reports noting that Mythos can further break software, even when trying to fix it.

Not to be outdone, OpenAI also claims to have a model — GPT-Rosalind — that’s too powerful for public release, this time in the sector of life sciences and molecular biology. Instead of launching Rosalind broadly, the company is offering it to researchers and scientists only.

So far, Anthropic and OpenAI have been socially responsible with their models by self-limiting access, but there’s no mandate to enforce these restrictions. President Trump’s executive order aims to eliminate any leeway and prevent truly dangerous AI models from leaking into the mainstream.

RELATED: The Trump phone is here — and so is the controversy. Is it any good?

BasSlabbers/Getty Images

This move raises the question: How much AI regulation is too much regulation, and what are the ramifications of government overreach on access to the most advanced technology known to mankind? Some view these bills and mandates as a danger to free speech. Others see it as a government power grab meant to control device, internet, and AI access. I’m somewhere in the middle — the government should prevent AI companies from outright harming the people while also keeping the people’s rights and freedoms intact.

Unfortunately, even if the Trump administration has the best intentions with its AI executive order, who’s to say that the next administration will be so benevolent? Direct government intervention over AI models gives the left the precedent they need to overtly regulate and even manipulate AI the next time they take power. Imagine a future where the left blocks AI models on the grounds of “misinformation” and “disinformation” for sharing facts that don’t align with their political views. It’s not like they didn’t try to wipe dissent from the internet before, and if given the chance, they’ll do it again.

Luckily, the left might not get that opportunity. President Trump’s AI executive order was put on hold the day it was meant to be signed, though the unsigned version was leaked online for your viewing pleasure. Still, even with the order paused at the eleventh hour, its albatross looms as a possibility for future AI regulation that could either save the people from certain chatbot destruction or steal away our rights to access “unapproved” versions of these models that don’t comply with the party in power.

It's not the next Jason Bourne flick. The Veldhoven choke point is way bigger than that.



There is a building in Veldhoven, in the Southern Netherlands, where engineers fire droplets of molten tin through a vacuum chamber 50,000 per second. Each droplet is intercepted by a laser, vaporized into plasma, and impelled to emit light at a wavelength of 13.5 nanometers, less than a thousandth the width of a human hair. The plasma is briefly 40 times hotter than the surface of the sun.

The machine that does this work is called an EUV lithography scanner, made by a Dutch company named ASML, the only such manufacturer. Your smartphone only works because of it.

Welcome to the Veldhoven choke point.

The EUV lithography scanner is, by some measures, the most complex manufactured object on earth. In 2025, ASML recorded revenue of 32.7 billion euros, of which it spent €4.7 billion on R&D. The scanner itself weighs more than 150 metric tons. Shipping one requires roughly 250 crates, 40 freight containers, several cargo planes, and 20 trucks.

The ultimate printing press

Lithography is a printing process: a pattern on a mask is optically projected onto a silicon wafer coated with photoresist, and the exposed regions are chemically altered to form circuits. Smaller features require shorter wavelengths of light according to the Rayleigh criterion. For decades, the industry shortened wavelengths incrementally, moving from visible light through ultraviolet to deep ultraviolet at 193 nanometers, squeezing extra performance through immersion fluids and clever tricks of computational correction.

What looks like ubiquitous computation is, underneath, managed scarcity.

The next step, extreme ultraviolet at 13.5 nanometers, required a fundamentally different machine: vacuum chambers, because EUV is absorbed by air; reflective mirrors rather than glass lenses, because EUV is absorbed by glass; mirrors polished to picometer tolerances, because at that wavelength any surface irregularity is an error. Zeiss, in Germany, makes these mirrors. They are roughly a meter across. Each mirror has more than 100 alternating layers of molybdenum and silicon, each layer a few nanometers thick. The largest ones are the smoothest objects humans have ever made.

ASML did not arrive at this position through genius alone. The company began in 1984 as a joint venture between Philips and ASM International, in a shed behind the Philips campus, with a staff that was given little funding and told to figure things out. Its first commercial product failed. The company survived by licensing technology aggressively and co-developing with suppliers. When EUV became the industry’s necessary next step, ASML had already positioned itself at the center of the effort. It acquired Silicon Valley Group in 2001, inheriting proximity to the U.S. research base that had done foundational EUV work. It acquired Cymer in 2013, bringing the light-source development in-house. It launched a co-investment program in 2012 in which Intel, TSMC, and Samsung paid €1.38 billion for the right to help fund EUV’s development and own a piece of the company that would sell them the tools.

The first commercial electronics enabled by EUV appeared in 2019. The research had begun in the 1980s. Nikon and Canon, both serious competitors in earlier generations of lithography, fell behind because they lacked the network. They did not have the suppliers, the customer co-investment, the acquired capabilities, or the tolerance for 30 years of deferred returns. Dominance in hard technology can look like the patient assembly of dependencies.

Printing money

The scanner imposes a disciplined way of seeing matter at scales that have no analogy in ordinary experience. A human hair is approximately 70,000 nanometers wide. ASML’S next generation of scanner, the High-NA EUV, has a resolution of roughly 8 nanometers. This required a redesigned optical system called anamorphic optics, in which the image is scaled differently in horizontal and vertical directions.

It would be a mistake to think of ASML’s dominance as residing in the scanner. The actual dominance is in the installed base of machines already running in fabs in Taiwan, South Korea, and the United States; in the €8.2 billion in annual service revenue that grew 26% in 2025; in the field engineers who operate on a 24-hour global rotation, resolving roughly 95% of issues locally, constituting a permanent guild of expertise that no competitor can easily replicate.

RELATED: The Trump phone is here — and so is the controversy. Is it any good?

BasSlabbers/Getty Images

The Dutch government has restricted exports of ASML’s advanced tools since 2023, with additional restrictions added in 2024 and 2025. The United States has pressed allies toward wider controls. China, which generated roughly a third of ASML’s 2025 sales, has been cut off from the most advanced systems and is expected to account for only about 20% of revenue in 2026. In May 2026, the Dutch government publicly objected to proposed U.S. legislation that would extend restrictions further. A private firm in Veldhoven has become a standing item in diplomatic correspondence between sovereign states.

ASML employs more than 44,000 people of 143 nationalities across more than 60 locations, and approximately 80% of its components come from a global supplier network. The machine that prints the world’s smallest features is assembled from a wide collaboration: German optics, German lasers, American light-source expertise, Belgian research infrastructure, Taiwanese and Korean and American customers. What looks like a Dutch company is a Dutch-coordinated actor network that has been stabilized, over decades, into something that behaves like a single artifact.

We speak of the digital world as if it were weightless, as if computation were a condition of the atmosphere rather than a product of factories in specific places run by specific people under specific export licenses. EUV lithography makes the concealment harder to maintain. The allegedly frictionless economy runs on tin plasma, picometer-smooth mirrors, and the continued willingness of the Dutch government to issue the right permits. What looks like ubiquitous computation is, underneath, managed scarcity: a single network managed from Veldhoven.

Hackers easily fool Instagram's new AI identity verification, humiliating Meta once again



There’s plenty of talk about how AI chatbots will take everyone’s jobs by the end of the decade, but so far, few companies have gone all in on the supposed wave of the future. One such company that did take the plunge, however, is Meta. After recently laying off 8,000 employees in favor of an AI workforce, Zuckerberg is already reaping the repercussions of little-to-no human oversight as AI just caused one of the most devastating Instagram account breaches in its history.

The Instagram bungle of the decade

Over the weekend, news broke that Instagram accounts were being hacked en masse with no clear reason why. Many of the targeted accounts included rare handles of high value, well-established accounts with more than a decade of ownership, and high-profile accounts belonging to elite users. Some of the most notable victims included the Obama White House and the Chief Master Sergeant of the U.S. Space Force John Bentivegna. Stranger still, account owners claimed that their passwords were changed without their knowledge or consent.

Platforms and apps will eventually require Real IDs for both age and identity verification.

At least in the case of Obama’s account, the hackers posted propaganda depicting former Iranian Major General Qasem Soleimani holding hands with former Iraqi Deputy Commander Abu Mahdi al-Muhandis, both eliminated at the hands of the U.S. military in 2020. The images included Arabic messages claiming that the White House is under Shiite control.

The exploit wasn’t due to a Meta system breach or password leak posted to the internet. As internet sleuths started to pore through the information, they discovered that Meta’s AI-powered identity verification system was easily duped into transferring ownership to thieves.

How hackers secretly stole Instagram accounts overnight

An X account that goes by Dark Web Informer published a video to their feed detailing exactly how hackers gained access to these high-profile Instagram accounts. The process was so simple that practically anyone could pull it off.

  1. The hacker enabled a VPN on his device and set it to the location of the account owner to trick Instagram into believing that he currently resided in the area.
  2. The hacker went to the “forgot password” section of Instagram and typed in the target account handle.
  3. On the next page, instead of choosing to send a recovery email that would alert the owner of the hacker, they selected “Get Support,” which opened a chatbox with the Meta AI support assistant. Note that this is an AI-powered chatbot, not a real person.
  4. From here, the hacker could tell the chatbot that the recovery email addresses on file were no longer valid and that any security codes should be sent to a new inbox, which pointed directly to an email address owned by the hacker.
  5. Meta AI support assistant sent an email to the new inbox with no additional verification, prompting the hacker to paste that code into the Meta AI chat to prove he was the person submitting the request.
  6. The Meta AI support assistant accepted the verification code and allowed the hacker to proceed with changing the password on the account.
  7. The hacker confirmed the new passcode, taking ownership of the handle and essentially locking the original owner out of the account entirely.

Other reports claim that the Meta AI support assistant may also prompt users to submit a selfie or video of themselves to verify their identity before proceeding with the password change. However, these safety measures were also easily thwarted by creating an AI-generated photo or video of the owner’s face that was good enough to trick the Meta AI support assistant into permitting access.

RELATED: Companies are tracking you based on the ads you see. Here's how to stop them.

BestForBest/Getty Images

As you can see, this exploit could be used on any account, anywhere, anytime, and the Meta AI support assistant gave up the ghost every time. The exploit has since been patched.

Zuckerberg made a mistake, but users lose in the end

If there were ever a way to prove that AI chatbots weren’t ready to replace humans in the workplace — especially in high-profile positions like account security — this is it. Meta has abused its workforce over the last several months, tracking their keystrokes to train AI and even sending 10% of its staff to the unemployment line, all to usher in the AI revolution where bots do all the work and humans are left to muddle through the mediocrity of degrading technology, fewer employment opportunities, and AI overload.

Unfortunately, these types of security breaches are just another reason that platforms and apps will eventually require Real IDs for both age and identity verification, and we’ll have no choice but to play along or lose access to the apps and services we use every day.

In the end, the only way for executives like Zuckerberg to learn that it’s better to keep real people on the payroll in lieu of AI is for the consequences to hit fast and hard. Unfortunately, this time, users were hit the hardest, but at least Zuckerberg’s decision to turn account security over to AI is a mistake he gets to own and atone for in whatever litigation is sure to come his way for his careless decisions.

Japan is close to finding cure for rare disorder that devastates children



A rare defect that can be devastating to children is getting a first-of-its-kind medicine from Japanese researchers.

A new treatment is now five years in the making, and after being used for medical applications, it's likely the product will be available for use by the general population as well.

'We feel that people's expectations ... are high.'

Since 2021, Japanese researchers have been hoping to find a solution for anodontia, the medical term for the complete absence of teeth, according to the Cleveland Clinic.

The initial study for this project states that anodontia and congenital tooth agenesis are common tooth anomalies affecting 1% of the worldwide population, resulting in a high rate of missing teeth.

The solution, according to lead researcher Katsu Takahashi, is to counteract a protein called USAG-1, which inhibits the growth of teeth.

"We want to do something to help those who are suffering from tooth loss or absence," he told the Mainichi in 2024. "While there has been no treatment to date providing a permanent cure, we feel that people's expectations for tooth growth are high."

The goal of the project is to give young children who have no teeth the joy of a real smile.

RELATED: Japan’s beautiful love affair with America

EyesWideOpen/Getty Images

The research was moved into human trials in October 2024 and lasted until October 2025. The controlled trial involved 30 males between 30 and 65 who were missing one or more molars, and the medicine was administered through one single intravenous dose.

The study has since been marked as completed, but little public information has been released. However, the Economic Times reported that as of April, preliminary analyses showed positive results with no significant side effects.

The next phase of the research is reportedly to test the medicine on children between 2 and 7 who suffer from congenital anodontia.

RELATED: That customer service rep with the American accent might still be an Indian guy — here's how

- YouTube

The teams at Kitano Hospital and Kyoto University Hospital believe that it may be soon possible to grow teeth not only in people with the aforementioned conditions, but also for common conditions like tooth loss from cavities or injuries.

According to Popular Mechanics, if the latest trials are successful, the researchers believe the medicine will become available to the public for all forms of tooth loss around 2030.

Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!

Google is about to overhaul the Android. You'll either love it or hate it.



Every year around mid-spring, Google hosts its annual IO event where it shows off the latest and greatest features coming to its products, services, and platforms. While the keynote address is now mostly reserved for AI and Gemini, we still got a sneak peek at the cool new things coming to Android 17. Here’s what you can expect when the update lands later this summer.

Gemini Intelligence

AI is everything these days, with Google even admitting it’s become an “AI-first” company under CEO Sundar Pichai, so it only makes sense that Android would receive a fresh new AI-powered update.

Those who despise it may do the unthinkable: consider switching to iPhone.

Perhaps as a cheeky nod to the flailing Apple Intelligence, Google introduced Gemini Intelligence, the first native AI agent for Android. Its main mission is to automate tasks on your device to free up your time to do other things. Just like you, Gemini Intelligence can interact with the apps installed on your device. If you don’t have an app required to complete a task, it can shift its workflow to the web via Chrome to navigate webpages.

It’s literally like having a personal assistant inside your phone. Tell it what to do, and it will go out and do it on your behalf. Theoretically, that means Gemini Intelligence can do your online shopping, reserve tickets to an upcoming concert, and even book your vacation, complete with airplane tickets and a hotel stay.

In many ways, Gemini Intelligence is the ultimate personal assistant that Google Assistant was always meant to be when it launched more than a decade ago, and with generative AI, we might finally be at the cusp of having useful AI agents in our mobile devices. That may excite some users who are bullish on the AI rush, while those who despise it may do the unthinkable: consider switching to iPhone.

While the Gemini Intelligence demo looked promising, we’ll have to test Google’s new AI agent firsthand to see if it lives up to the hype, which we’ll get to do soon. Gemini Intelligence is coming to Google Pixel 10 phones and Samsung Galaxy S26 devices later this summer.

Gemini Intelligence/The Android Show/I/O Edition 2026

Pause Point

Phones are addictive, and even with the Digital Wellbeing settings baked into Android that let you set app timers and silence notifications, sometimes your favorite app still pulls you in and wastes hours of your time before you realize what’s happened. If you’re trying to kick certain addictive apps to the curb and setting time limits isn’t enough, Pause Point might be what you need.

This new feature acts as a check point between you and your apps. When activated, a pause screen will show up on your device the next time you open an addictive app. The screen stays up for 10 seconds, giving you time to consider if you really want to click through and waste hours of time doomscrolling on social media or watching cascades of short-form videos. During the waiting period, you’ll be prompted to take a couple of meditative breaths or browse photos of people you should spend time with instead of scrolling. You can even tell it to suggest different apps to open instead of that one addictive thorn in your side.

RELATED: Sick of Microsoft's preinstalled propaganda on your PC? Block it now.

DigitalVision/Getty Images

Either let the time go by and open the app anyway, or tap “Don’t Open” to close the addictive app and reclaim the free time you almost lost.

The idea of Pause Point is to get you to think twice before you spend too much time looking at a screen, and even though it might sound a little silly for some, it’s a great tool for other users who want to break their app addiction without completely throwing out their smartphone.

Pause Point/The Android Show/I/O Edition 2026

Other interesting announcements

Google had a few more interesting updates before the close of the event. Here are the highlights:

Expanded Quick Share support: It’s easy to share a photo from one iPhone to another with AirDrop, or between Android phones with Quick Share, but it’s nearly impossible to do the same across OS platforms. That’s changing now that Android 17 includes interoperability with AirDrop, letting iPhone owners and Android users exchange photos, videos, and other files over the air with just a tap. The list of supported devices is small right now, but it’s a step in the right direction to knocking down the walled garden that historically made it harder for Apple and Google devices to communicate with each other.

Quick Share + AirDrop supported devices/The Android Show/I/O Edition 2026

New iOS transfer tools: Again, it’s easy to upgrade from an iPhone to an iPhone and Android to an Android, but it’s historically been difficult to switch platforms entirely. With Google’s new iOS transfer tools, more of your personal data can be pulled from an old iPhone to a new Android phone, including apps, app data, home screen layout, calls, contacts, and more.

New iOS transfer tools/The Android Show/I/O Edition 2026

Android Auto optimizations: Widgets you can easily glance at are coming to Android Auto to provide more contextual information about your apps while on the road. Immersive navigation offers a new 3D view of Google Maps when driving, complete with lanes, stop signs, stop lights, and other markers. You can also watch videos on the internal display when your car is stopped or listen to the audio of the video only when in drive.

Android Auto updates/The Android Show/I/O Edition 2026

Googlebooks

Serving as its “one more thing” moment, Google had a final surprise to tease before the end of the event. It’s called Googlebook.

Not to be confused with Google Books, the books archival service established in 2004, or Google Play Books, the company’s little-known competitor to Amazon Kindle and Apple Books, Googlebooks are a new breed of laptop that run on a combined version of Android and ChromeOS.

Just like mobile Android, this laptop-ready version of Android is centered on AI. Gemini lives in the cursor. Simply wiggle the mouse to summon it to the forefront to read your screen and complete different tasks based on what it sees — schedule a calendar event it spotted in your email, automatically write a reply, or prompt Gemini with your own queries. The goal is to make laptops more useful with AI, though I’m not sold on the initial demo. It feels like Googlebooks are still searching for a problem in need of a solution to make them stick.

Nevertheless, Googlebooks are meant to replace Chromebooks as Google’s flagship laptop platform, though Google claims that Chromebooks are also sticking around, at least for schools and other institutions.

Googlebooks/The Android Show/I/O Edition 2026

Android 17 is coming

All the features covered today are expected to roll out in Android 17 to Google Pixel phones first this summer. After that, it will make its way to other Android phones as OEMs like Samsung, Motorola, and OnePlus integrate their user interfaces into the final Android codebase for their own devices.

Top companies admit humans cost less than AI — but still want more bots



The cost of doing business today may be higher than ever, even if it involves fewer humans.

While some major U.S. companies are starting to see the vast costs of their robotic colleagues as prices soar for AI-driven operations, companies are still pushing employees to use more and more AI.

According to executives at computing companies, the cost of AI has now exceeded the typical employee salary totals.

The mantra is that even more AI usage needs to happen.

"For my team, the cost of compute is far beyond the costs of the employees," Bryan Catanzaro, vice president of applied deep learning at chip maker Nvidia, said in early May.

The cost of AI computing, especially when it comes to coding, has come as a surprise to some companies once they start integrating it into their teams and spreading access to their engineers.

Most of the major corporations have been using Anthropic's Claude, which is seemingly cheap when it comes to image generation, but dollar signs pile up when generating documents or computer code.

As Forbes reported, Uber ran through its entire 2026 AI budget in just four months. Chief technology officer at the company, Praveen Neppalli Naga, even admitted to spending $1,200 by using AI for a personal demo, with the company's engineer cost ranging from upwards of $250 per month in usage, all the way up to $2,000 per month.

RELATED: DOJ asked to probe whether Biden officials let Microsoft off easy in exchange for cushy jobs

Huiying Ore/Bloomberg/Getty Images

Between December and March, Uber achieved a 95% usage rate among its engineers to implement AI tools and use Claude for coding.

Over at Microsoft, thousands of its developers were invited to use Claude for coding, but so were project managers, designers, and other employees.

The Verge reported that after starting in just December, the usage has become so popular that the company is making a switch and adopting Microsoft's own Copilot model into its workflow.

The mantra shared by all of these companies is that even more AI usage needs to happen. Amazon, Uber, Microsoft, Nvidia, and Meta are pushing employees to keep spending tokens.

RELATED: Self-driving trucks are about controlling the roads — not making them safer

Idrees MOHAMMED/AFP/Getty Images

Uber ranked its engineers on internal leaderboards based on Claude code usage. A Meta employee reportedly made a leaderboard titled "Claudenomics" to track which workers were using Claude the most.

Fortune reported that Amazon is pushing employees to "tokenmaxx" and use as many tokens as possible.

As icing on the cake, Nvidia CEO Jensen Huang recently said he believes eventually every employee at his company will work alongside 100 AI agents.

Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!

Out of control: Here's how a company spent $500 million on AI in a single month



Artificial intelligence usage at a single company spiraled out of control and led to a half-billion-dollar bill, according to a recent report.

A consultant is sounding the alarm on what could become the norm for companies in the near future: paying for AI integration may not be all it is made out to be.

One token is equal to approximately four written characters in English text.

An AI consultant recently provided Axios with a stunning revelation that has sparked intrigue across the globe. According to the unnamed insider, one of the consultant's clients spent approximately $500 million in a single month on AI usage.

These costs reportedly piled up because the company failed to put usage limits on its employees who have AI licenses for the large language model Claude.

Anthropic, Claude's operator, has different pricing structures that go up to $25 per million tokens. This may seem low, but one token is equal to approximately four written characters in English text or "0.75 words," Anthropic says on its website.

This of course includes punctuation marks.

Token consumption can be quite heavy when it comes to documents. For example, a PDF costs ~125,000 tokens, a large document is ~25,000 tokens, and a webpage is listed at ~2,500 tokens.

RELATED: The Trump phone is here — and so is the controversy. Is it any good?

Chris Ratcliffe/Bloomberg/Getty Images

Images through Claude Vision are calculated using a formula based on picture size in pixels. The formula is width * height / 750.

For example, a YouTube thumbnail is 1280 × 720 pixels and would therefore cost about 1,229 tokens. While it might end up costing just under $5 to produce around 1,000 average-sized images, the high costs are believed to stem from the scale of employee usage as well as when Claude is used to code.

The unnamed company — which was described in a LinkedIn post as a U.S. corporation — reportedly gave employees unfettered access with zero spending caps or usage limits.

RELATED: Google's AI overhaul of Search will overfish the internet to extinction

Chris Ratcliffe/Bloomberg/Getty Images

Without any guardrails in place, a company that creates documents or webpages or even performs coding with AI could be passively spending tens of thousands of dollars.

One chief technology officer told Axios that employees had been using AI for some of the most trivial tasks, which included checking the weather. Token plans may not be as they seem and are not "all you can eat" buffets, the CTO said.

Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!

Sick of Microsoft's preinstalled propaganda on your PC? Block it now.



Microsoft’s leftist roots run deep, from its problematic co-founder Bill Gates and his “philanthropic” ventures to its partnership that gave birth to MSNBC — recently rebranded to MS NOW — which regularly spews left-wing talking points dressed up as actual facts. It even went out of its way to stuff leftist news stories into the widgets bar of every Windows computer via its MSN news aggregator, a service that prioritizes left-wing content over the right. That’s all about to change, though, as a string of complaints over Windows’ waning security and stability lead Microsoft to make its operating system a little more user-friendly, and the MSN feed is one of the first bad ideas on the chopping block.

A brief history of MSN

Microsoft launched the Microsoft Network all the way back in 1995, and the company has crammed it into the desktop operating system ever since. Debuting on Windows 95, it started as an online dial-up service meant to compete directly with the early internet juggernaut that was AOL.

One year later, Microsoft secured a lucrative joint venture with NBC News. In an effort to consolidate power across established cable TV and the new internet machine, the companies formed MSNBC. The nexus of their partnership saw that NBC News continued to provide 24-hour news coverage through conventional channels while Microsoft delivered those stories to its online users, boosting viewership ratings for both brands on the way to the top.

By 1998, Microsoft spun the MSN brand out into the news aggregation service known today as MSN.com. For maximum visibility, Microsoft set MSN as the homepage of its Internet Explorer browser, which, at the time, dominated the web with a 90% market share over second-place competitor Netscape Navigator. You can even see an early version of the original MSN.com thanks to web archives.

Microsoft ultimately walked away from MSNBC in 2012, selling its stake in the partnership to pursue its own venture. No longer in need of NBC’s reporting alone, Microsoft became an independent news distributor under MSN.com. This move would give Microsoft full control over which news outlets it chose to feature, as well as the right to hoard the spoils of its ad revenue from these stories. To make MSN virtually unavoidable, Microsoft injected its media influence into the Windows task bar in late versions of Windows 10 and all of Windows 11 under the name Microsoft Start. Now, any time you glance down at the task bar on your Windows PC, you’ll see messages from MSN that cover weather, finances, “breaking news,” and other topics, all begging for you to click and read.

Today, the MSN feed is one giant leftist propaganda billboard meant to promote any outlet that doesn’t espouse right-leaning ideas or values. What’s worse is that Microsoft can use this baked-in “feature” to serve left-wing content by default to Windows machines at workplaces, schools, and homes around the nation. Everywhere.

Screenshot by Zach Laidlaw/Windows 11

Microsoft’s mission to win back user trust

Unfortunately for Microsoft, a series of blunders have left users unhappy with the company’s portfolio of platforms and services. Here are just a few of its recent mistakes.

  • Windows has suffered from several critical bugs since the start of the year, all chipping away at OS security and eroding user trust.
  • Microsoft’s aggressive push to inject its AI platform, Copilot, into every app and service has been poorly received, with users complaining about its ubiquitous integration that has only complicated usability.
  • Xbox has suffered from major annual losses, with significant drops in hardware earnings and a small drop in software revenue.
  • Where Windows once dominated the low-tier and mid-tier PC market, Apple’s new affordable Macbook Neo poses a significant threat to a computer segment that Microsoft historically kept mostly to itself.

In short, Microsoft is feeling the heat of competition, self-inflicted failures, and customer dissatisfaction on multiple fronts, and the only way to earn back user trust is to fix some of its more egregious mistakes. One of these is Windows’ user experience.

A new ‘Start’ for Windows

Microsoft Start is divided into two sections — the “Discover” view is powered by MSN propaganda, and the “Widgets” view shows only pertinent information without political commentary. By default, Microsoft Start opens to the Discover feed, filled with news stories designed to capture your attention. It’s clickbait. Meanwhile, the more useful Widgets are hidden behind an extra click; the default widgets include useful information, like weather, sports, finances, events near you, and a couple of other stragglers.

In an upcoming Windows update, Microsoft Start will show the Widgets view first, with the Discover feed as optional. Don’t get too excited, though. Microsoft will continue to pack left-wing stories into Microsoft Start, but on the upside, users can soon ignore it wholesale, making PCs a little less politically intrusive.

When asked about the decision, Microsoft said, “We're working to make Widgets feel less distracting and overwhelming by making the experience quiet by default. To do this, we're testing a new set of default settings designed to reduce unexpected alerts and visual interruptions.”

Get rid of MSN now

The refreshed Microsoft Start is currently only available in preview builds of Windows for developers. However, you don’t have to wait to banish the MSN feed from your taskbar. If you really want to kick it to the curb now, open Microsoft Start, click on the Settings gear at the bottom, uncheck the green toggle beside “Discover,” and that’s it! Microsoft Start will now default to the widgets view, which you can customize to your liking.

Screenshots by Zach Laidlaw/Windows 11

You’re now free from leftist propaganda, at least in this section of your Windows PC.

Data brokers can learn all about you just from what online ads you see. Here's how to stop them.



Digital ads are a commonality across the internet. You see them in Google Search, social media feeds, and even on your favorite websites. If you spend enough time online, you might’ve grown accustomed to ignoring them.

Unfortunately, a new study reveals that what has become a necessary annoyance to the modern web might also have the power to reveal personal and private information about your interests, beliefs, and more. Even worse, these personal details about you can be gathered without clicking on a single ad, thanks to AI.

Websites can’t stop any company from collecting and using this data.

The study

In a study published by UNSW Sydney in early May, researchers revealed an alarming new trend about online ads: These seemingly innocuous bits of marketing materials on sites all around the web can be used to reveal and track a person’s most privately held values and beliefs – including political affiliation, degree of education, and employment status – simply by monitoring the advertisements users see online.

To be clear, it’s not the ads themselves that can gather specific data about you, but it’s the collective presence of the ads displayed that reveal personal traits. Here’s how it works:

Using Facebook as the catalyst for the study, researchers reviewed 435,000 ads distributed to a relatively small subset of 891 users. After monitoring which ads were served to each user, they ran the correlated data through a large language model and discovered four main points about the results:

  • Researchers could infer users’ personal traits without accessing their browsing history or personal data on their devices. All they needed was a log of their ad history.
  • User profiles could be created after a short browsing session (though they didn’t outline how long a session needed to be to make it work).
  • AI-based personal trait matching rivaled and even exceeded human capabilities.
  • The AI-powered process was both 200 times more affordable and 50 times quicker than relying on human analysis alone.

The thing that makes this study so startling is that users don’t have to actively share any information about themselves, no cybersecurity loopholes or zero-day exploits are required, and platform holders behind today’s operating systems, web browsers, and websites can’t stop any company from collecting and using this data.

RELATED: A secret bot army is phishing, scamming, and sabotaging our lives

gremlin/Getty Images

This isn’t the first time LLMs have been used to reveal extremely private information about online users. Earlier this year, we reported how AI can reveal the real identities of anonymous accounts simply by comparing writing styles.

Should you be worried?

At this point in the story, you might be wondering if you have anything to worry about. The answer is “maybe,” depending on how your smart devices are configured.

The bright spot of the study is that your personal interests can’t be measured if the data is never recorded. UNSW Sydney noted that extensions, like those you’ll find in the Chrome Web Store, Safari Extensions, and Microsoft Edge Add-Ons are the likely avenue for data collection. The more extensions you have installed on your devices, the higher your chances are that your ad history could be abused. If you don’t have any extensions installed, your chances of ad data collection drop precipitously.

That’s not to say that all extensions are bad. However, even the innocent ones have the power to view which webpages you visit and even the content on those pages.

Ironically, another possible method for ad data collection are ad blockers. While blockers can effectively prevent websites from showing you ads, some may still access served ad data and gather it for user profiling. You especially want to watch out for ad blockers that claim to be free. Remember, if you’re not paying for the product, you likely are the product.

Even without extensions in the mix, data brokers can still collect plenty of information about you, and you still don’t have to click on an ad to hand it over. The sites you browse on the internet are filled with cookies — little crumbs of data — that track where you go and which pages you click from site to site. Even if you don’t click on an ad, simply visiting a product page or website is enough to leave a cookie in your browser that tells brokers the things you like and the things you don’t. These can then be used to build profiles on your browsing habits to target you with other ads you might actually click, which you should never do, as evidenced by the stark rise in social media scams.

Ways to protect yourself from ad data collection

Staying safe and anonymous online is an increasingly difficult task. However, if you want to give yourself the best shot at nullifying this ad data collection “exploit,” try out these tips:

  • Remove all extensions from your preferred web browser. This is probably the top way to block bad actors from recording your ad data.
  • Install a VPN. Many VPNs come with built-in ad-blocking technology. If you choose to add a VPN to your device, make sure it’s a RAM-based option with a no-log system that actively prevents the VPN provider from recording or saving user data. Some popular RAM-based VPNs include ExpressVPN, NordVPN, and CyberGhost VPN.
  • Block cookies entirely. Some browsers will let you block all third-party cookies. Unfortunately, this may break some websites, so your mileage may vary.
  • Clear out your cookies often. Set a reminder to delete your browser history and cached data every week or month. This can make it harder for sites to monitor your activity over time.
  • Browse in private mode. While “private” or “incognito” mode won’t obscure your web traffic from your ISP, many browsers come with extra tools to reduce or block cookies and other tracking methods that brokers use.