You don't need to know how a car engine works to drive well, and it turns out the same logic applies for AI. McKinsey numbers are saying demand for AI fluency in Europe is up 5x since 2023, mind you, not the people building the models, just the ability to use AI well. Technical AI roles only grew 1.7x over the same period, meaning fluency is growing roughly 3x.

So the people getting ahead rn are the ones who figured out how to work alongside AI.

Entrepreneurs FTW. Other sources are pointing to founders and entrepreneurs as the most AI-optimistic group, and honestly, it tracks, as these people are already comfortable leading. With AI, we all have to step into a manager/leader role, because if you're using it well, you're the one setting direction, quality controlling, and approving or tweaking the work, pretty much like a classical manager. So one takeaway: Learn to lead.

Money flows where attention goes. Automation is estimated to unlock $1.9 trillion in economic value across Europe by 2030. Only 10% of companies report NOT using AI, still less than 40% are seeing any results. The gap doesn't stem from a tech issue. Most companies have bolted AI onto workflows built for 2015 and wonder why it doesn't work. While the companies that are AI-native have rebuilt their whole ecosystem around it, and are the ones reaping benefits.

My guess: results are spotty because many AI initiatives are being led by people who still think AI fluency is optional.

If you believe in AI and want in as an investor, there's more than one way to play it. What matters is knowing which play you're actually making, because each layer of the stack carries different risk and rewards.

This is the compute layer: the chips and hardware every AI model has to run on, whether it's open or closed, wins or loses. Infrastructure gets paid regardless, because someone always needs to rent the hardware.

Inside that layer, there are really three jobs being done. The CPU handles general processing, the everyday thinking. The GPU handles the heavy lifting AI specifically needs, which is why it's become the most fought-over piece of the whole stack. The DPU handles moving data around efficiently as things scale, which matters more than people realize once you're running at serious size. Three different jobs, three different chips, and right now, demand for the GPU is where all the tension sits.

This is the lowest-drama, lowest-surprise layer overall. Steady demand, but also the most crowded trade, which means the easy money here has largely already been made.

Both sit in a higher-risk territory, but they're not the same bet. Model layer risk is about the tech itself losing its edge, while application layer risk is about the business having nothing to stand on once the model underneath is swappable.

Model layer = you're betting on the company building the actual intelligence. The risk here is the tech itself getting commoditize, todays smartest model is next year's free open-source option, and building frontier models is brutally capital-intensive with no guaranteed payoff.

Application layer = you're betting on a company using someone else's model to build a product. The risk here isn't the tech, it's that they don't own their moat, the model underneath can get swapped out, or the lab that built the model can just copy their feature and cut them out entirely. Their edge has to come from something else: distribution, data, brand, workflow lock-in.

Related to infrastructure but distinct: someone has to build and own the actual buildings housing all this compute. This is closer to a real estate bet than a tech bet, location, land, power access, and cooling matter more here than software.It's less flashy, more physical, and moves on a completely different timeline.

AI doesn't just need chips, it needs an obscene amount of electricity to run them. Training and running these models at scale is turning into one of the biggest new sources of power demand in decades. Which means the AI story isn't just a tech story anymore, it's an energy story, and whoever can actually keep the lights on for this thing has leverage most people aren't watching yet. This layer is nuanced because it intersects with things that move slowly and painfully: grids, permitting, politics.

Right now natural gas is filling that gap because it's fast to build, while nuclear is the longer bet, since AI needs power that never dips. Renewables are growing fastest but can't carry the load alone.

So, if you want to invest, Whatever route you choose, or if you spread across a few, treat this like the young, fast-moving field it still is.

Two unicorns of the decade, AI and GLP-1s, in some kind of combo that's supposed to hit a home run, right? But whats worth asking in both cases, who actually profits?

The acceleration is real, AI is compressing the drug discovery timeline from years to weeks. AI platforms can now screen millions of molecular combinations before a single one ever reaches a lab, and Eli Lilly is already building this directly into its metabolic pipeline. A faster pipeline means a much bigger market arriving much sooner than anyone planned for, and Pharma is already building the consumer machinery to meet it.

About consumer; Novo is getting creative reaching the masses

I cannot make this up. Last week, Novo Nordisk flew health creators (paid partnerships) in from 34 countries, for their "Beyond Weight" event at the Royal Danish Opera House in Copenhagen. Nikolaj Coster-Waldau (Jaime Lannister), was up on stage talking nutrition and mental health.

That's a massive (and atypical) step for Pharma in the direction of consumer, celebrity-backed lifestyle brand. Worth keeping an eye on: are they brewing something for an even broader public? I would not be surprised. Also here, it looks like taste is leverage. GLP-1s are getting crowded, so Novo isn't winning on the molecule alone anymore, so they're building a lifestyle brand instead.