Reflection on AI

I’ve started to wonder what “AI” really means.

It’s probably one of the most repeated words online and offline these days. On stage, in headlines, at conferences, in small talk, even in WhatsApp groups. But behind the buzzword, what are we truly referring to? What does it mean when people say “AI is the future,” or “AI is disrupting everything”?

Everyone seems to have their own version of the story. For some, AI is ChatGPT. For others, it’s about chips, robotics, or the promise of autonomous driving. AI is a shapeshifter depending on who you ask.

So these past few months, I’ve tried to look past the noise. I spoke with experts and “experts.” Sat through startup pitches and global conferences. Compared corporate vision decks with backroom conversations. What emerged is both funny and a bit unsettling. Here are a few things that stood out.

1- Everybody is talking about it …

“AI” is everywhere. You hear it in startup pitches where founders promise to “revolutionize” industries they barely understand. You read it in annual letters from CEOs who, until recently, couldn’t tell a transformer from a Tesla. And you feel it in conversations that begin with “So… what do you think about ChatGPT?”

It’s hard to escape it. In 2020, there were only ten major AI events globally. In 2025, we’re looking at thirty. Triple the noise. In parallel, the number of AI companies has grown from 50,000 to 70,000 over the same five-year stretch, an average of 11 new companies every single day, including weekends. It feels as if everyone is building something “AI-powered” even if that means dressing up glorified keyword search in a neural net costume.

AI-related jobs have followed the same upward curve: from 3 million globally in 2020 to over 7 million expected by the end of this year. The U.S. alone employs 2.2 million people in the field. China isn’t far behind with 2 million, and Europe, will hit 1 million soon. Patent filings are also skyrocketing: China is moving from 35,000 to 60,000 per year between 2020 and 2025, and the U.S. has doubled from 10,000 to 20,000 over the same period.

Note the difference: China has 33 workers per AI patent filed; the U.S. has 110. 

2- Nobody is talking about the same thing

But here’s the thing: for all the words written, posted, and debated about AI, very few people seem to be talking about the same thing.

Some mean the tools (ChatGPT, Claude, Midjourney, Perplexity, Gemini) what you and I might call the interface layer. Others refer to the underlying methods: natural language processing, machine learning, reinforcement learning, symbolic logic, or the use of large language models. And then there’s the hardware crowd, those who speak about AI but mean chips, cooling systems, and Nvidia’s trillion-dollar ascent. For them, it’s all about compute power and who controls it.

To some, AI is a tech stack; to others, it’s a geopolitical race. One founder I met referred to it as “the electricity of the 21st century.” A hedge fund analyst called it “the next bubble.” A French philosopher called it “a mirror for our own confusion.” All three might be right.

What fascinates me is how much semantic chaos hides under the word “AI.” A single term, yet a thousand meanings. It can mean a bot that generates memes… or a military-grade autonomous drone. A predictive model that helps detect tumors… or an algorithm that serves you cat videos. You could argue AI is everything but it is nothing until it’s applied.

When nobody’s talking about the same thing, it’s easy to pretend we’re all aligned. But when the dust settles, it might turn out that “AI” was just a placeholder, something we projected meaning onto, depending on what we needed it to be.

3- Most of the time, it’s used to inflate the price of an underlying

A few weeks ago, I attended an AI Forum in Hong Kong. I showed up full of enthusiasm, half expecting to see robots navigating hallways, drones mapping the ceiling, and investors whispering about breakthroughs in autonomous intelligence. The reality was, let’s say… slightly different.

As I walked through the booths and started asking companies what they actually did, the answers got blurrier by the minute. A fintech startup explained it used a chatbot API. Another company in education said they “helped people using AI”. A consulting firm was presenting their cloud infrastructure as “AI readiness.” Even some traditional industrial players managed to squeeze in a reference to their data warehouse and call it a day.

AI is mostly a buzzword.

One particularly memorable moment came during a pitch session for VCs. A founder took the stage to present his AI-driven proptech platform. It sounded sleek. I was almost on board, until a jury member asked, “You’ve been around for a few years now, but still haven’t monetized anything significant. Why the pivot to AI?”

Earlier, the same founder had told me privately that they were now moving to a subscription model, giving clients access to an “AI-powered platform for real estate contract maker.” I’m not saying it’s a bad idea. But I’ve seen this movie before: promising startups repackaging themselves as AI ventures, replicating existing solutions, and asking for seven-figure funding because they spent three months in a bootcamp learning Python.

At this rate, I should call Influidence an AI company. I do use AI tools every day. Maybe I should raise money that way.

Between 2020 and 2025, global funding into AI companies went from $22 billion to $110 billion: a fivefold increase. $73 billion was already deployed by Q1 2025 alone. We are clearly not short of capital. But what we often lack is clarity.

As investors, we’re trained to be wary of hype, of the widening gap between price and value. But sometimes, price isn’t about value at all. Sometimes, it’s about looking cool. Investing in AI today feels, at times, like buying a front-row ticket to the future… even if the show hasn’t been written yet.

I’m not denying there are great projects out there, some are truly game-changing. But many are not. They’re just chasing the trend because the label opens doors. It flatters the ego, secures media coverage, and keeps pitch decks exciting. It’s a shortcut to relevance. And like any shortcut, it usually comes with a cost.

4- AI might change the world, but it hasn’t made much money yet

If you strip away the buzzwords, the valuations, and the breathless press releases, and look at the numbers, you find that American AI companies are not exactly fountains of cash.

While S&P 500 energy companies generated over 20% free cash flow every year since 2020, and even tech companies sit at a healthy 15-10% rate, AI is… well, lagging behind. Badly.

In 2020, the average free cash flow of American AI companies was -10%. By 2025, it’s expected to reach just 12%. It’s not a disaster. But it’s also not what you’d expect from the most hyped, most capitalized, most mythologized sector of our time.

Average Free Cash Flow (FCF) of American AI Companies vs. Other Industries (2020–2025)

But, the multiples say otherwise. LLM vendors are trading at over 40x revenue. Even fledgling AI infrastructure players, some with little more than a beta and a roadmap, are securing nine-digit rounds at nosebleed valuations. Meanwhile, consumer discretionary or healthcare businesses with real traction and real margins get a fraction of that attention.

So what gives?

Source: https://www.finrofca.com/news/ai-startup-valuations-q1-2025-edition

It’s the same story we’ve seen before. The narrative outpaces the numbers. People don’t invest in AI right now because of its cash flows, they invest because they’re afraid to miss what comes next. It’s a fear-driven premium. A bet on inevitability. An act of faith.

And faith has its place. But it shouldn’t replace discipline.

Not every AI company is worthless, obviously. Some are quietly building tools that will shape the next decade. But a surprising number are selling futures with no underlying. As investors, we have to remember: not all revolutions are profitable, and not all profitable things look like revolutions. So unless you’re okay with a $1 turning into $1.09, it might be time to reassess where the real value lies.

5- The big guys (Google, Facebook, Microsoft, Amazon, Nvidia, DeepSeek, etc.) are gonna win

Fairly, they’re just better equipped. They have the money, the brains, the infrastructure, the data and the platform to go to market quickly. They’ve been working on this long before it became fashionable. Some of them are running hundreds of internal models already and have entire divisions dedicated to AI safety, deployment, or commercialization. They understand what they’re doing.

Even when a start-up manages to stand out, it usually doesn’t last long. Either it gets acquired or sidelined. The acquisition machine is well-oiled and aggressive. That’s what they’ve done for years, and it works. They pick what’s interesting, and they leave the rest behind.

A lot of investors should reconsider the idea that throwing money at AI ventures is the best way to capture the upside. There might be more interesting plays elsewhere. In many cases, you’re financing a learning curve that Big Tech has already completed. And even if your AI start-up manages to survive, the minute it gets real traction, you’re competing against firms that have already solved distribution and scale.

That doesn’t mean you shouldn’t invest in AI. It just means that not all AI bets are equal, and most of the value, by design, will be captured by the ones who are already ahead.