I spent a week at CES: all just bullshit dressed up in AI new clothes.

Well-known tech media personality Edward Zitron wrote an article this week revealing his observations at CES, exposing how the tech industry is filled with meaningless AI product demonstrations, from robots pretending to fold clothes to various chatbots, highlighting that the current AI investment bubble is even more severe than the dot-com bubble of the past… The main points are summarized and translated below.
(Background: CES 2026 features a new “Bitcoin mining water heater,” claiming to earn $1,000 annually)
(Additional context: The 25 craziest ideas at CES 2026 are all here)

I just spent a week at CES, and one word kept echoing in my mind: bullshit.

LG, a company known for appliances and TVs, showcased a robot( inexplicably named “CLOiD”), claiming it could “fold clothes”( at a very slow speed, only under limited conditions, and sometimes it would fail) or cook(—I mean, put things into an automatic oven) or find your keys( in a video demo), but they have no intention of actually releasing this product.

The media generally gave lenient reviews; one reporter even considered this barely functioning tech demo a “milestone,” because LG is now “entering the robot space,” with a product they never planned to sell.

So why did LG showcase this robot? Of course, to deceive the media and investors! And hundreds of other companies also displayed robots you can’t buy, even if reports might say otherwise, what we’re seeing isn’t “the future of robotics” in any meaningful sense. What we see is what happens when companies lack creativity and can only copy each other. 2026’s CES is the “Year of Robots,” just like someone sitting in a cardboard box wearing a captain’s hat claiming to be a sailor.

However, compared to the absurd wave driven by large language models, robot companies are surprisingly more ethical: from anonymous startups in the Venice Convention Center basement to companies like Lenovo endlessly talking about their “AI super agents.” In fact, screw it, let’s talk about this.

“AI is evolving and gaining new capabilities, perceiving our three-dimensional world, understanding how things move and connect,” said Lenovo CEO Yang Yuanqing, then introduced Lenovo Qira’s demo, claiming it “redefines the meaning of technology built around you.”

People might expect the next demo to be a stunning showcase of future tech. But instead, a spokesperson took the stage, asking Qira to show what it could see(—namely, multimodal functions available in many models for years), receiving summarized notifications(, which are available in almost any large language model integration and are very prone to hallucinations), and asked “what to buy the kids when you have time,” at which point Qira told her, quoting, “Las Vegas Fashion Mall has some kids’ Labubus that will drive them crazy,” referring to the tool-based web search available since 2024.

The host pointed out that Qira can also add reminders: this has been available on most iOS or Android devices for years, along with document search, then showcased a conceptual wearable device that can record and transcribe meetings—something I saw at CES at least seven times.

Lenovo rented out the entire Las Vegas Sphere to showcase a damn chatbot powered by OpenAI’s model on Microsoft Azure, and everyone acted as if this was something new. No, Qira isn’t a “big gamble” on AI—it’s just a chatbot imposed on anyone buying a Lenovo computer, full of “summarize this” or “transcribe this” or “tell me what’s on my calendar” features, sold by business idiots with little real-world experience, marketing it as something people should care about.

Want better-looking videos or audio from your TV? Screw that! What you get is Google’s Nano Banana image generation and Samsung’s other large language model features.

You can now generate images on your TV using Google’s Nano Banana model—an useless idea sold by a company that doesn’t know what consumers really want, packaged as making your TV assistant “more helpful and visually appealing.” As David Katzmaier correctly said, no one asked to install large language models in their TVs to “click and search” for things—that’s something normal people wouldn’t do.

In fact, most trade shows felt like companies playing fill-in-the-blank with startup pitches, trying to fool people into thinking they did something, rather than just slapping a frontend on a large language model. The most obvious example: a bunch of useless AI-powered “smart” glasses, all claiming to do translation, dictation, or run “applications” with bulky, ugly, hard-to-use interfaces, all using the same large language model, doing basically the same things.

These products exist because Meta decided to pour billions into launching “AI glasses,” with a bunch of followers described as “part of a new category,” rather than “a bunch of companies making useless bullshit no one wants or needs.”

This isn’t about companies genuinely afraid of making mistakes, let alone media, analysts, or investors judging them. It’s the behavior of the tech industry, hiding behind “giving them a chance” or “being open to new ideas,” avoiding meaningful criticism of their core businesses or new products—much less regulation!—even though these ideas are always the same as what the tech industry has said before, even if they’re meaningless.

When Facebook announced it would rebrand as Meta, as a way to pursue “the successor to mobile internet,” it offered no real evidence besides a series of terrible VR apps, but don’t worry, Casey Newton at Platformer told us Facebook would “strive to build a maximized, interconnected experience directly from science fiction: a world called the metaverse,” adding that the metaverse “is all the rage.” Similarly, Dan Newman of Futurum Group said in April 2022 that “the metaverse is coming,” and “may continue to be one of the biggest trends in the coming years.”

But after three years and $70 billion, the metaverse is dead, and everyone acts as if it never happened.

Oh! In a rational society, investors, analysts, and media would never believe a word Mark Zuckerberg says again. Instead, they happily reported on his mid-2025 blog post about “personal superintelligence,” where he promised everyone would have a “personal superintelligence” to “help you achieve your goals.” Can large language models do that? No. Will they? No. That’s okay! That’s the tech industry.

No punishment, no consequences, no criticism, no skepticism, no retribution—only celebration and reflection, only growth.

Meanwhile, the biggest tech companies continue to grow, always finding new ways( mainly through aggressive monopolies and massive sales teams) to boost numbers, to the point that media, analysts, and investors have stopped asking challenging questions, and naturally assume they—and the financiers supporting them—will never do anything truly stupid.

The tech, business, and financial media are now well-trained to understand that progress is always the main story, and failure somehow “necessary for innovation,” regardless of whether anything is actually innovative.

Over time, this creates an evolutionary problem. The success of companies like Uber—reaching profitability after burning billions over a decade—convinces reporters that startups must burn a lot of cash to grow. To persuade some media members that something is a good idea, all it takes is $50 million or more in funding, and larger funding rounds make criticizing a company less appealing, because of the fear of “betting on the wrong winner,” assuming that the company will go public or be acquired, and no one wants to be wrong, right?

This naturally creates a new world of startup investing and innovation, driven by a corrupt economy obsessed with growth at all costs. Startups are rewarded not for creating real businesses or good ideas, or even new categories, but for playing the “brainwash venture capitalists” game—either by becoming “founders worth betting on,” or by attracting the next multi-billion dollar potential market scam.

Maybe they’ll find some product-market fit, or grow a large audience by offering services at unsustainable costs, but all of this is done with the knowledge that they will soon be rescued via IPO or acquisition.

The stagnation of venture capital

For years, venture capital has been rewarded for funding “big ideas,” and most of the time, those ideas have paid off. Eventually, those “big ideas” stopped being “big ideas for necessary companies,” and became “big ideas to grow as fast as possible and dump into the public markets or other companies afraid of being left behind.”

Going public used to be easy [from 2015 to 2019, over 100 IPOs each year, with a continuous flow of acquisitions giving startups a place to sell], until the overhyped bubble burst in 2021(, which saw $643 billion in venture investments that year), leading to 3,110 IPOs before October 2023 that lost 60% of their value. Years of foolish bets based on the assumption that markets or big tech would buy any company that scared them.

This has caused the current liquidity crisis in venture capital, with funds raised after 2018 struggling to return capital to investors, making VC less profitable, which in turn makes raising funds from limited partners harder, reducing available startup funding, and causing startups to pay higher rates, as SaaS companies—some of which are startups—extract more from customers each year at higher rates.

All these problems boil down to one simple thing: growth. Limited partners invest in venture capitalists who can show growth; venture capital invests in companies that show growth; this increases their valuation, allowing them to sell at higher prices. Media reports focus not on what companies do, but on their potential value, which is mainly determined by the company’s vibe and how much funding they’ve raised from investors.

And all of this only makes sense in a liquid environment. According to the overall TVPI( of funds raised after 2018—how much money is made per dollar invested—most venture capital firms have failed to generate more than their principal for investors over many years.

Why? Because they invested in bullshit. That’s it. They invested in companies that will never go public or be sold to others—garbage. While many think venture capital is about early, high-risk bets on startups, most of the money actually goes into late-stage bets. A more honest way to describe this is “doubling down on established companies,” but those of us living in reality see it as an inherently different culture—more akin to stock investing than understanding any business fundamentals.

![])https://img-cdn.gateio.im/social/moments-78cdeef6a1-540653a351-8b7abd-e2c905(

Maybe I’m naive, but I see venture capital as about discovering emerging technologies and giving them the means to realize ideas. The risk is that these companies are early and might fail, but those that don’t fail will grow rapidly. Conversely, Silicon Valley waits for angel and seed investors to take risks first, or spends all day browsing Twitter, looking for the next big thing.

The problem with this system is that it naturally rewards scams, and the emergence of certain technologies is inevitable—they will fight against a system that has already expelled good judgment and independent thinking.

Generative AI lowers the barrier for anyone to cobble together a startup that can tell all the right things to venture capitalists. Atmosphere coding can create a “working prototype” of a product that can’t scale) but can raise funds!( The fuzzy issues of large language models—such as their data greed, huge data security problems—give founders the chance to create many “observable” and “data-credible” companies, while the heavy costs of running anything near large language models mean that VCs can make huge bets on inflated valuations, allowing them to arbitrarily increase their net asset value as other desperate investors pour into subsequent rounds.

As a result, AI startups accounted for 65% of all venture capital funding in Q4 2025. The disconnect between venture capital and value creation) or reality( has led to hundreds of billions flowing into AI startups that are already unprofitable, and as their customer base grows, margins worsen, and inference costs) to generate output( increase, it’s clear that it’s impossible to create a profitable foundational lab or large language model-driven service, and even renting GPU for AI services seems unprofitable.

I also need to be clear: this is far worse than the dot-com bubble.

U.S. venture capital invested $11.49 billion) in 1997 (equivalent to $23.08 billion today(), $14.27 billion) in 1998 (equivalent to $28.21 billion today(), $48.3 billion) in 1999 (equivalent to $95.5 billion today(), and over $100 billion) in 2000 (equivalent to $177.1 billion(), totaling $344.49 billion) in today’s dollars(.

This is just $6.174 billion more than the $338.3 billion raised in 2025 alone, with about 40-50%) roughly $16.8 billion( invested in AI, and in 2024, North American AI startups raised about $106 billion.

According to the New York Times, “48% of internet companies founded since 1996 still existed at the end of 2004.” The 2000 bubble mainly burst with suspicious and clearly unsustainable e-commerce stores like WebVan) ($393 million in venture capital(), Pets.com) ($15 million(), and Kozmo) ($233 million(), all of which filed for IPOs, though Kozmo failed to sell itself to the market in time.

But in a very real sense, the “internet bubble” experienced by everyone has little to do with actual technology. Public market investors blindly poured money into any company that even smelled like a computer, causing most major tech or telecom stocks to trade at absurd multiples of earnings) Microsoft’s case: 60x(.

When the bullshit internet stocks collapsed, the world realized that the magic of the internet wasn’t a cure-all for every business model, and the bubble burst, with no magical moment turning WebVan or Pets.com into real businesses.

Similarly, companies like Lucent no longer got rewarded for suspicious circular transactions with firms like Winstar, which led to the telecom bubble bursting. Millions of miles of dark fiber were sold cheaply in 2002. The oversupply of dark fiber was eventually seen as positive, leading to demand surging as hundreds of millions of people went online in the late 2000s.

Now, I know what you’re thinking. Ed, isn’t this exactly what’s happening here? We have overvalued startups, multiple unprofitable, unsustainable AI companies promising IPOs, overhyped tech stocks, and one of the largest infrastructure builds ever. Tech companies are trading at absurd multiples of earnings, but the multiples aren’t that high. That’s good, right?

No. Not at all. AI boosters and well-meaning folks love to make this comparison because saying “things got better after the dot-com bubble” helps them rationalize doing stupid, destructive, reckless things.

Even if it’s like the dot-com bubble, the situation will be absolutely fucking catastrophic: Nasdaq fell 78% from its March 2000 peak, but due to the incredible ignorance of tech industry power brokers, I expect the consequences to range from disastrous to catastrophic, almost entirely depending on how long the bubble takes to burst and how willing the SEC is to approve IPOs.

The AI bubble will be even worse because the investment scale is larger, the contagion broader, and the underlying asset—“GPUs”—completely different in cost, utility, and fundamental value from dark fiber. Moreover, the fundamental unit economics of AI—whether infrastructure or the AI companies themselves—are far more terrifying than anything seen in the dot-com era.

In simple terms, I am genuinely very worried, and I am tired of hearing people make this comparison (referring to 2000’s internet and today’s AI bubble).

![])https://img-cdn.gateio.im/social/moments-7530f7007c-5e16a1d6a6-8b7abd-e2c905(

)##