Hey everyone,

Welcome back to Stocks To Space, where I curate the best ideas, tools and resources I’ve found each week as I explore my curiosities.

Today, I'm sharing some of the most compelling technology predictions that will continue shaping our world in 2025.

For today’s piece, I’ve gotten help from my favourite geeky YouTube channel, Fireship.

If you find this thought-provoking, forward it to a friend.

Today at a Glance

• AI continues evolving at breakneck speed, Big Tech faces unprecedented regulatory pressure, and consumer preferences are shifting as emerging technologies become increasingly embedded in daily life.

• The 11 tech trends defining 2025: Reasoning models, AI agents, humanoid robots, brain chips, AR/VR disillusionment, quantum computing, AI-driven development, crypto, Trumponomics, monopoly crackdowns, and the anti-cloud movement.

• The most significant market shifts often come from the unexpected. Whether it's AI dominance, regulatory upheaval, or disruptive innovation, 2025 will be a year of volatility, transformation, and opportunity.

11 Tech Trends for 2025

Predictions reveal our deepest hopes and fears about the future.

In 2024, AI continued evolving at a breakneck pace. Neuralink successfully installed a brain chip in real humans. The pressure on tech monopolies was relentless.

Yet despite all the doom-sayers, artificial superintelligence (ASI) didn't take over the world.

As we enter 2025, these themes will continue to mature.

This year’s predictions are less about isolated breakthroughs and more about reshaping our fundamental relationship with technology.

Consumer preferences have changed since ChatGPT launched in November 2022. The novelty of this breakthrough faded quickly.

People on the hedonic treadmill will ask: “What’s the next thing?” There might not be a new ‘thing’, but the existing thing will become increasingly embedded in our lives.

So, let's explore the technology trends shaping our future in 2025—they might make you rich, they might make you cry, or both.

Artificial Intelligence

The first three trends are all related, in one way or another, to AI.

This is unsurprising since AI's tailwinds have contributed to most of the Magnificent 7’s stock performance and have driven record funding to AI startups.

On the ground:

• AI models are becoming more intelligent, faster, and cheaper.

• Infrastructure is more reliable.

• Tooling is getting better.

Anyone focussing on AI expects this to continue in 2025.

What happens in reality—anyone’s guess.

Reasoning Models

Right now, the elephant in the room for knowledge workers is OpenAI's o3 model, released in the grand finale of OpenAI’s “12 Days of OpenAI” event in December 2024.

o3 crushed every benchmark test it was put through.

It reached ‘International Grandmaster’ level in coding competitions and aced the Software Engineering benchmark (SWE-bench) test.

But what really blew everyone’s brains was its 88% score on the ARC-AGI evaluation, a test designed to measure human-like reasoning.

This evaluation does what it says on the tin: it tests AI systems' ability to generalise and solve novel tasks.

This is a key component of general intelligence, so we’re acutely testing the model for any indicators of AGI.

Now, AGI is a big and scary term that people throw around too nonchalantly. It’s worth defining for this conversation.

When we talk about AGI (Artificial General Intelligence), we're talking about AI that can match or exceed human-level performance across most cognitive tasks.

Not just coding or writing but the full spectrum of human intelligence—problem-solving, creative thinking, and adaptive reasoning.

This is where things get complicated for o3. While its outperformance of other models is undeniably mind-blowing, there are some caveats:

• Each task costs over $2000 in compute, though this will likely improve.

• It still stumbles on questions that most humans would find trivial.

• OpenAI's demo only showed it building a simple Python UI—something I’ve done regularly with much less sophisticated models.

It’s an understatement to say these advanced models are coding wizards.

But true AGI? That would mean building complex games with GTA-6-esque graphics from scratch, solving for new physics, or even creating the next generation of AI models.

We're not there yet.

Looking at model quality data, we're witnessing a convergence in performance across different models, suggesting a commoditisation of capabilities.

This phenomenon is also evident from the top models' market share trend over the past two years.

It's becoming less about who has the most potent model and more about the quality of the UX and how these tools are implemented and integrated into customer workflows.

We'll need to wait until o3 reaches the general public to evaluate its impact truly.

But one thing's clear: while we're seeing remarkable progress in AI reasoning capabilities, we're still far from the sci-fi version of AGI that keeps appearing in headlines.

The real story isn't about a single model achieving human-like intelligence—it's about how these increasingly capable tools augment knowledge work.

This is a trend that’s more likely to continue in 2025 than Skynet coming online.

Agents

AI agents are the hottest buzzword in tech for 2025. It’s becoming an impossible-to-ignore narrative.

Every big tech company is pushing this story hard, and investors are eating it up like candy.

But before we get wrapped up in the story, let’s go back to definitions. What exactly are these agents that everyone keeps talking about?

Think of an AI agent as a digital worker with a specific job description.

It's an AI system that can perceive its environment (your computer, security cameras, or business data), make decisions based on what it sees, and take actions without human intervention.

The possibilities are fascinating.

• A developer agent like Devin can write code, debug issues, and manage entire development workflows.

• Security agents can monitor camera feeds and coordinate responses to threats.

• Customer service agents can handle support tickets 24/7, learning and adapting from each interaction.

What differentiates these agents from traditional automation is their ability to handle uncertainty and adapt to new situations.

They do this by reasoning about problems and finding novel solutions, much like humans do daily. This is a departure from traditional automation (like a Zapier workflow), which follows pre-programmed rules irrespective of changes to its environment.

The narrative around AI agents is currently more powerful than their actual capabilities. Companies are racing to position themselves as leaders in this space, even though many of these systems are still in their early stages.

Some limitations are apparent:

• Most agents still require significant human oversight.

• Their reliability varies wildly across different tasks.

• The infrastructure to support truly autonomous agents is still developing.

However, a tale as old as time is how markets don’t necessarily move because of reality. They move because of the stories we tell about the future.

This doesn't mean the progress isn't real. Companies like Anthropic, OpenAI, and Google are making genuine breakthroughs.

But in 2025, the narrative of AI agents revolutionising work will enormously impact markets.

Whether agents automate 5% or 95% of knowledge work by the end of 2025, one thing's for sure: the companies that can sell this vision of the future while delivering even incremental progress will make a killing.

Markets are moved by stories first and reality second.

Humanoid robots

Humanoid robots were the AI Agents of 2024, becoming the hot new AI narrative of the year.

Aside from all the buzz, the harsh reality for humanoid robots is that I'm still folding my own laundry, and you probably are, too.

This is the thing about narratives and the stories that we tell ourselves.

While we don’t all have a domesticated robot, the big three players in robotics have been throwing everything at the problem.

• Tesla's Optimus is slated for high-volume production "sometime in 2026" (classic Elon timeline).

• NVIDIA just dropped their Jetson Thor platform for humanoid computing.

• Figure's robots with OpenAI brains are looking impressive.

And not to mention the flood of Chinese humanoid robot releases at CES 2025.

The competition is heating up faster than expected, even if the real-world applications are still limited.

Here's what's happening on the ground:

• Tesla's Optimus can now handle basic tasks like bartending and social interactions, but we're far from the domestic helper Elon promised.

• NVIDIA is playing the long game, building the entire tech stack from training to deployment

• BMW has requisitioned Figure robots for targeted work within its factories.

The progress is tangible but slower than the short-term hype suggests. And like any transformative technology, it’ll probably be even more transformative than we imagine in the long term.

Think about smartphones—it took years after the iPhone's launch for the app economy to really transform our lives.

Humanoid robots will likely follow a similar trajectory. The foundations are being laid now, but the real impact might not hit until 2030.

Brain chips

Read this slowly so it sinks in: we've successfully put a computer chip in someone's brain, and it worked.

Not a lab rat or a monkey—a human being.

In January 2024, Noland Arbaugh became the first human to receive Neuralink's brain-computer interface implant.

A 30-year-old quadriplegic who couldn't move below his shoulders, Noland can now control a computer cursor with his thoughts.

He literally plays Chess using his mind.

The sheer complexity of what was achieved here is staggering.

Think about what Neuralink has accomplished:

• They figured out how to interface with the most complex organ we have.

• They designed a chip small enough to sit in said organ without disrupting it.

• They developed surgery techniques that were precise enough to insert it.

• They created software that can translate neural signals into digital commands.

Despite early hiccups, the fact that this technology works at all feels like something straight out of a William Gibson novel.

Yet, this moment hasn't captured the public imagination like ChatGPT did. Maybe it feels distant—most of us aren't quadriplegics, so we can’t fathom Neuralink’s immediate impact on our lives compared to Noland’s.

But that's missing the bigger picture.

Similar to how the first iPhone was technically just a phone that could browse the web, it could’ve easily been dismissed early on.

These early brain chips might seem limited in scope but the potential applications are mind-bending. Think about the possibilities of:

• Direct brain-to-computer communication with other humans. Or Aliens.

• Enhanced memory and cognitive capabilities.

• Seamless control of smart devices.

Like humanoid robots, progress will probably be slower than we’d like.

We're years away from having brain chips as standard as smartphones. The regulatory hurdles facing biotech innovation like this alone are massive.

While robots might change how we interact with our physical world, brain chips could fundamentally alter what it means to be human.

Are we still human if we can directly interface our thoughts with computers?

This exponential augmentation might lead to full-blown evolution.

But, as I said, barring something crazy—bookmark that until 2030.

VR disillusionment

The AR/VR space in 2024 felt like watching a Michael Bay Transformers movie—otherworldly special effects but a weak plot.

We had all the incredible fight scenes—Apple's Vision Pro launch, Meta's Orion reveal, whispers about Google's Project Astra—but without any substance.

Let's talk about the Apple Vision Pro.

The classic Apple launch playbook swept us away:

• An iconic ad

• Premium pricing at $3,499

• Breakthrough features like EyeSight

• Even renaming the category “spatial computing.”

But here's the thing—they've already had to scale back production due to lower-than-expected demand.

It's not exactly the “iPhone moment” they were hoping for.

The real story here isn't about technical capabilities but user experience.

The Vision Pro can help surgeons perform operations and create immersive workspaces, which is incredible.

But for the average person, we're still strapping miniature TVs to our faces.

AR/VR will remain a niche industry until it solves the “smartphone problem.” Our phones are just too damn good. They're always with us, they're socially acceptable to use anywhere, and they just work.

The killer AR device needs to be:

• As lightweight as regular glasses

• Socially acceptable to wear, i.e. look cool

• Capable of perfect real-world passthrough

Until then, specific use cases like surgery or industrial design will drive limited adoption.

Just as the smartphone didn't take off until it solved fundamental user experience issues, AR/VR needs its own breakthrough in form factor and interface design.

When strapping on an Apple Vision Pro is as cool as whipping out your iPhone, we’ll see real mass market success for AR/VR devices.

Post-quantum world

Google's Willow quantum chip announcement in late 2024 caused quite a stir. And rightly so—a 105-qubit chip that can perform calculations faster than the fastest supercomputer on Earth is frightening.

Particularly frightening is the possibility that a computer powered by Willow could break all modern encryption standards—basically, it would be a full-blown digital crisis.

But let's put this in perspective: Willow uses 105 physical qubits to create and stabilise a single logical qubit. Meanwhile, we need about 4,000 stable logical qubits to break RSA-2048 encryption and 8,000 to crack SHA-256 (which secures Bitcoin).

The math isn't hard here:

• We're at one stable logical qubit

• We need thousands for practical cryptographic attacks

• Each logical qubit requires over 100 physical qubits

This means that we’re about a decade away from this threat becoming real. The gap between current capabilities and what's needed to break modern cryptography is still massive.

Don’t only take my word for it. Just look at Google’s roadmap:

And while everyone's freaking out about quantum computers breaking Bitcoin, they're missing the bigger picture. Our encryption standards won’t stand still.

The world is already developing post-quantum cryptographic solutions that should resist quantum attacks long before they become a real threat.

This reminds me of Y2K—a legitimate technical challenge that drove meaningful upgrades to our systems but not the catastrophe many predicted. Unless something truly extraordinary that no one expects happens in 2025, quantum supremacy remains a future concern, not a present crisis.

Aside from the potential threats to current systems, we can look forward to a future in which quantum computing unlocks limitless possibilities for drug discovery, materials science, and optimisation problems we can't even approach with classical computers.

AI, Meet Developer

AI is reshaping the entire software development landscape.

We're seeing two parallel revolutions:

• For developers: AI acts as a force multiplier, making them 10x more productive and creative.

• For non-technical folks: AI bridges the coding gap, turning “I can't code” into “I can actually build that.”

Tools like v0 and Bolt fundamentally change what it means to build software.

You can now go from idea to working prototype in minutes, not months. This has staggering implications for the future of software.

On a shorter-term horizon, here's what I believe will define 2025:

Developers will no longer choose frameworks based on technical merit. They'll flock to ecosystems with the best AI tooling.

While this will be the trend to follow in this space, the real prize for companies is not about winning over developers. The company that figures out how to make its AI tools accessible to non-technical users will win.

Maybe this is Replit. Or perhaps an unknown upstart appears and takes it all.

It’s open season for AI-driven development in 2025.

Crypto glitches

The crypto world in 2024 was like watching a masterclass in market psychology.

Bitcoin hit the ephemeral $100,000 milestone that seemed outlandish a year ago, and MicroStrategy turned borrowing into an art form.

People’s expectations for crypto in 2025 are enormous. You can feel it in the air.

Outsized expectations usually result in predictions like Bitcoin to $250k, ETH to $10k, and SOL to $1,000.

But these forecasts miss a fundamental truth about markets. The most significant moves often come from events no one sees coming.

Think about it:

• Who foresaw Sam Bankman-Fried being outed as a fraud?

• Who predicted UST's collapse in 2022 would trigger a crypto winter?

• Who saw Bitcoin ETF approval fundamentally changing market dynamics?

For 2025, instead of making bold price predictions, I'm more interested in watching how the market adapts to unprecedented scenarios.

The real opportunities will emerge from being positioned to capture the upside of unexpected events while staying resilient to the downside.

Remember: the most transformative market events are rarely the ones we can forecast.

Trumponomics

In a world filled with uncertainty, Trump’s return to the Oval Office is arguably the biggest wild card.

He’s already stacked his administration with high-profile figures—Elon Musk being the most notable—and has the tech community buzzing with anticipation.

But here’s the paradox of expectations: the higher they are, the harder they can fall.

Compare this to President Truman, who took office in 1945, to a collective shrug. He delivered surprisingly strong leadership precisely because no one saw it coming.

Conversely, Trump enters with front-page fanfare, so any perceived misstep could trigger a wave of disappointment.

Conversely, an unexpected policy breakthrough could spark an outsized market rally—just the sort of tail event that can reshape entire sectors overnight.

For tech in 2025, the question isn’t whether Trump will move markets but how.

• Will he kick off an era of deregulation, propelling AI and crypto to the stratosphere?

• Or slam the brakes on Big Tech with aggressive antitrust moves?

Either way, it’s going to be a roller coaster.

Monopoly losers

Big Tech is facing its biggest regulatory reckoning yet. In 2025, antitrust enforcers will directly target the hyperscalers, and the fallout could reshape the industry.

Here are some of the legal battles we get to look forward to:

• The U.S. Department of Justice has declared Google a monopoly in search, and regulators are considering forcing it to divest Chrome. This would disrupt its search-advertising dominance and potentially weaken its AI ambitions—similar to how Microsoft’s antitrust battle in the 90s slowed its innovation and let rivals gain ground.

• The FTC is aggressively targeting Meta’s hold on Instagram and WhatsApp, with increasing pressure to break them apart. If regulators succeed, social media’s power dynamics could shift overnight, allowing new players to challenge.

• Amazon is under fire for alleged anticompetitive marketplace tactics, including squeezing sellers with high fees and price controls. It could upend e-commerce and redefine online retail competition if forced to restructure.

Antitrust battles are slow-moving, but the pressure is mounting.

If any of these lawsuits were successful at a grand scale, forcing a company into irrelevance, it would erase a trillion dollars in market cap from the S&P.

On the other hand, if regulators fail, Big Tech will tighten its grip, cementing its dominance for years to come.

Either way, the stakes couldn’t be higher—the outcome will shape the future of tech and the global economy.

Anti-Cloud movement

For years, the dominance of cloud giants like AWS, GCP, and Azure seemed unshakable. But in 2025, a growing anti-cloud movement is challenging that assumption.

Companies are re-evaluating whether relying on centralised cloud services is the best long-term strategy instead of returning to on-premise infrastructure.

The irony is brutal: the very companies that propelled Big Tech to trillion-dollar valuations could be the ones to undermine it by moving away from their services.

If a mass migration back to on-prem were to gain traction, it would deal a staggering blow to the Magnificent 7’s stock prices, potentially rivalling the effects of the antitrust lawsuits.

After all, the only way to kill a monopoly is to stop using its products.

But there may be a catch—can startups realistically afford to abandon the cloud?

AI upscaling and building AI-native applications require enormous computing power, something that on-prem solutions struggle to match without astronomical costs.

Conversely, cloud providers offer unparalleled scalability, something a scrappy startup can’t replicate with its own hardware.

So, while the anti-cloud movement raises valid concerns about cost control and vendor lock-in, it’s unclear whether it will ever reach a critical mass.

Cloud computing remains the backbone of AI innovation, and unless infrastructure costs fundamentally shift, Big Tech’s stranglehold on computing power will not disappear anytime soon.

Whether businesses can afford to move away from cloud dominance will determine whether this movement is a real existential threat or just another passing tech rebellion.

With all that said, fate loves irony.

What a time to be alive

2025 is set up to be a wild year. Product drops from CES already hint at what this year has in store for tech.

A lot of the above predictions will probably be entirely wrong.

Some of the predictions may only be partially right.

Very few, if any, will come true.

This is the nature of our highly unpredictable, ever-changing world.

But like I said, many of these tech trends are actually decade-long projects that will keep grinding forward.

So, you might ask, why make predictions at all?

Let me tell you why this process is so essential by quoting one of my favourite writers, Morgan Housel, in his bestselling book The Psychology of Money:

If predicting what stock markets will do in 2025 is hard, then predicting how tech will advance is almost impossible.

Progress may be stagnant, or we could discover ASI. We just don’t know.

But it’s reasonable for us to try to make sense of the world no matter how wrong we might be.

That’s what I’m trying to convey above.

I hope this can serve as a robust lens to view what this year may or may not hold.

Now that predicting is over, it's time to get to work.

Was this newsletter forwarded to you? If you liked it, sign up here.

And, if you loved it, forward it to your friends and family.

Thanks for reading,

— Luca