Talk Your Book: Public & Private Opportunities in AI

AI Is Big Enough to Matter and Messy Enough to Misprice

The strongest case for AI is no longer just technical brilliance. It is commercial adoption. Businesses are using AI across more functions, budgets are moving beyond pure experimentation, and executives are getting less interested in flashy demos and more interested in one stubborn question: does this thing save money, make money, or both?

That shift matters because markets eventually stop paying for possibility and start paying for evidence. Early in a technology cycle, investors often reward anything with a credible story. Later, they get picky. They want margins, retention, usage data, customer concentration, and proof that the product is sticky once the novelty wears off. AI is entering that phase now.

That does not make the opportunity smaller. It makes it broader. The investable AI universe is no longer limited to the obvious names that manufacture chips or rent out cloud capacity. It now stretches across data-center landlords, power and cooling suppliers, industrial automation, cybersecurity, data management, software companies that embed AI effectively, and private startups building everything from evaluation tools to vertical assistants for law firms, hospitals, manufacturers, and insurers.

In plain English, AI is growing up. And grown-up technologies usually create more than one class of winner.

Why Public Markets Still Own the First Draft of the AI Story

Semiconductors, Memory, and Networking Still Matter

The public side of AI begins with the companies supplying the physical guts of the boom. Compute is not a metaphor. Someone has to design the chips, package them, cool them, connect them, and keep them running without melting the electric bill into modern art. That is why the first leg of the AI trade has been so visible in public equities. These businesses are easier to analyze than an early-stage startup because investors can see revenue growth, backlog trends, capital spending, and margin progression in real time.

There is also a brutal logic here: before AI can automate a call center, draft legal summaries, or write better code, somebody has to build the machines that make those tasks possible. The picks-and-shovels trade may not always be the most glamorous part of the story, but it often gets paid first.

Cloud Platforms and Data-Center Exposure Are Not Side Characters

The next public opportunity sits in cloud infrastructure and digital real estate. AI models require enormous computing capacity, and enterprise customers usually do not want to build all of that from scratch. They rent, lease, and outsource. That creates a chain of public-market beneficiaries: hyperscale platforms, colocation providers, fiber and networking players, and even the property owners tied to data-center buildouts.

This is one reason the AI conversation has expanded beyond software and into infrastructure. AI is increasingly an industrial buildout story. It needs servers, land, electricity, cooling, and grid upgrades. Investors who only stare at chatbot headlines can miss the duller but highly monetizable reality: when a technology wave starts pulling on physical infrastructure, the opportunity set gets much wider.

Power, Cooling, and Electrical Equipment Are the Sneaky AI Trade

One of the more interesting turns in the AI boom is how quickly energy became part of the thesis. Data centers do not run on optimism. They run on power. That means utilities, power equipment makers, cooling specialists, and firms exposed to electrical upgrades have moved from background extras to important supporting actors.

There is a weirdly beautiful finance lesson here. Investors wanted an AI revolution. What they also got was a reminder that the digital world still sits on top of transformers, substations, pipes, backup systems, and giant bills for electricity. Sometimes the future looks less like a robot and more like a very overworked utility engineer.

Enterprise Software Adopters Could Become the Second-Wave Winners

Public AI opportunity is not just about builders. It is also about adopters. The next class of winners may be publicly traded businesses that use AI to expand margins, improve productivity, reduce service costs, and speed up product development. This is where the story becomes more selective and, frankly, more fun.

Not every software company with an AI button deserves applause. Some are dressing up ordinary automation in a futuristic jacket. But companies that genuinely use AI to improve sales productivity, customer support, coding velocity, fraud detection, underwriting, or workflow automation can create real economic value. That can show up in better margins, faster growth, or both.

In other words, the public-market AI story is moving from pure infrastructure to actual operating leverage. That is usually when stock picking gets harder and more interesting.

Where the Private-Market Upside Lives

Foundation Models and Model Infrastructure

The private side of AI still attracts huge attention because some of the most explosive upside remains off the public exchanges. Foundation model companies, inference optimization firms, developer tools, evaluation layers, and model-governance platforms all sit in territory where market leaders can still be defined. That is thrilling for venture investors and mildly terrifying for everyone else.

The attraction is obvious. If a private company becomes the default layer for building, routing, evaluating, securing, or customizing AI, the upside can be enormous. The downside is just as obvious: intense competition, falling prices, huge infrastructure costs, and the very real possibility that what looked like a moat turns out to be a very expensive puddle.

Vertical AI Applications May Produce the Quiet Giants

Some of the best private opportunities may not be the biggest model labs. They may be the companies applying AI to highly specific workflows in industries that hate generic software. Think healthcare documentation, legal review, insurance claims, industrial maintenance, procurement, logistics, accounting, drug discovery, and customer service in regulated sectors.

Why does this matter? Because horizontal AI is crowded. Vertical AI, by contrast, can pair models with proprietary workflows, domain-specific data, compliance expertise, and painful real-world problems customers will actually pay to solve. That combination can create better retention, deeper integration, and pricing power that does not disappear the second a cheaper model shows up.

Agentic Software Is Exciting, but Execution Still Wins

Agentic AI has become the phrase everyone wants on stage and in pitch decks. The idea is appealing: software that does not merely assist but coordinates tasks, makes decisions within rules, and completes multi-step workflows. In theory, that opens the door to much larger productivity gains than simple copilots.

In practice, agents still need guardrails, high-quality data, clean system integrations, and plenty of human supervision. That is why private investors need to separate theater from traction. The interesting companies are not the ones showing the coolest demo. They are the ones proving that agents can operate safely inside messy enterprise systems, save real labor hours, and avoid turning the compliance team into a support group.

Embodied AI and Robotics Are the Longer-Duration Bet

There is also a more speculative but increasingly serious corner of private AI: embodied systems, robotics, and automation that connect digital intelligence to physical action. This is a longer-duration opportunity, but it matters because the AI story may eventually spill from screens into warehouses, factories, hospitals, and homes.

This category comes with higher technical risk, heavier capital needs, and slower commercialization. It also carries the kind of upside that can make investors act as if sleep is optional. The winners here will likely combine software, hardware, and data loops in ways that are hard to replicate. The losers will produce excellent demo videos and very little else.

Public vs. Private AI: The Real Trade-Offs

Public AI investing is about visibility. You get liquidity, price discovery, audited financials, and a faster feedback loop. The market can still overhype public companies, but at least investors have quarterly data, management commentary, and a real-time sense of sentiment. Public markets are also where the largest and most immediate AI monetization has shown up so far, especially in infrastructure and at-scale software.

Private AI investing is about optionality. You are paying for the possibility that a company becomes essential before everyone else notices. The catch is that everyone else may have noticed already, and the valuation may assume six miracles before lunch. Private investors also deal with illiquidity, slower information flow, governance questions, and exit timing that often depends on broader market conditions rather than pure business quality.

That does not make one better than the other. It makes them different tools for different jobs. Public exposure can give investors access to the current monetization wave. Private exposure can offer entry into the next one. Smart portfolios usually respect both truths.

What Investors Keep Getting Wrong About AI

They Treat All AI Revenue as Equal

Revenue tied to one-time experimentation is not the same as recurring revenue embedded in mission-critical workflows. A company selling GPU demand today may look different from a company capturing durable workflow spend five years from now. The market often blurs those categories until reality sorts them out.

They Underestimate the Importance of Proprietary Data

General-purpose models are powerful, but proprietary data is where many businesses create defensibility. If every competitor can access roughly similar models, then the edge shifts to distribution, workflows, brand, regulatory readiness, and unique data. This is especially true in enterprise and vertical AI.

They Ignore the Adoption Gap

Buying AI infrastructure is easier than reorganizing a company around AI. Plenty of enterprises have budget, interest, and pilot programs. Far fewer have clean data, governance, change management, and the courage to redesign workflows. That gap between technical possibility and organizational readiness is where both disappointment and opportunity live.

They Forget That Falling Costs Can Help as Much as They Hurt

Yes, lower model prices can compress margins for some builders. But falling costs can also expand adoption, widen the customer base, and create room for new applications that were previously uneconomic. AI price pressure is not automatically bad news. It often shifts value from one layer of the stack to another.

How to Think About the AI Opportunity Set Without Becoming a Hype Intern

A useful framework is to divide AI opportunities into four buckets. First, builders: the companies creating models, chips, compute, and infrastructure. Second, enablers: the firms providing data tools, cybersecurity, orchestration, governance, networking, cooling, and power. Third, adopters: the businesses using AI to improve economics inside existing operations. Fourth, private optionality: the earlier-stage companies trying to become the next platform, vertical champion, or automation layer.

This approach helps avoid a classic mistake: assuming AI is one trade. It is not. It is a stack, and each layer behaves differently. Builders may win first. Enablers may compound quietly. Adopters may create the most underestimated margin upside. Private companies may deliver spectacular returns or spectacular excuses. The point is not to predict a single winner. It is to understand where the value is being created at each stage of the cycle.

And yes, valuation still matters. A wonderful business can be a terrible investment if bought at a ridiculous price. AI has not repealed arithmetic. It has merely given it a nicer user interface.

The Experience of AI Investing on the Ground

Talk to founders, CIOs, product leaders, and portfolio managers, and the lived experience of AI in 2025 and 2026 sounds less like a single narrative and more like a noisy orchestra tuning up before a very expensive concert. Everyone agrees the music is coming. Nobody agrees on which section will carry the melody.

On the public-market side, the experience has often been surprisingly straightforward at first and much more nuanced later. Early on, investors flocked to obvious beneficiaries: the companies making the chips, renting the compute, and selling the cloud capacity. That phase had a clear logic. Demand was visible, spending was real, and earnings revisions had teeth. It felt like standing near a construction site and correctly guessing that the businesses selling cement, steel, and power tools were going to have a pretty good quarter.

Then the experience changed. Investors started asking whether second-order beneficiaries could win too. Could utilities benefit from data-center demand? Could enterprise software companies use AI to drive better margins? Could industrial firms that support cooling, electrical systems, and automation become part of the AI trade without ever building a model? Suddenly, the conversation became less about one giant theme and more about diffusion. The winners were no longer just the most obvious names. They were the companies quietly plugged into the broader buildout.

Private-market experience has felt different. It is more crowded, more emotional, and frankly more exhausting. Founders report that customers are interested but cautious. Enterprises want AI, but they also want security, compliance, reliability, integrations, and proof that the product will not hallucinate its way into a legal incident. Venture investors see giant upside, but they also see crowded cap tables, high burn, and intense competition from both incumbents and open models. It is a thrilling place to hunt for outliers, but nobody should confuse thrilling with easy.

CIOs often describe the middle ground best. They say the hardest part is not getting access to AI tools. The hardest part is cleaning data, prioritizing use cases, training teams, and deciding whether to build internally, buy from a vendor, or do an awkward combination of both. That experience matters for investors because it explains why adoption can feel both fast and slow at the same time. Fast in interest. Slow in implementation. Fast in pilots. Slow in enterprise-wide transformation.

There is also a psychological experience attached to AI investing that deserves mention. Every cycle produces fear of missing out, but AI adds a layer of existential panic. If you are an executive, you worry your company is behind. If you are a founder, you worry your feature will be commoditized. If you are an investor, you worry the most expensive stock will keep rising after you sell it and the private deal you passed on will become legendary at someone else's annual meeting. It is a very modern form of stress.

Yet beneath the noise, one pattern keeps showing up: the durable opportunities usually sit where technical capability meets workflow reality. That is where AI stops being a party trick and starts becoming a business. Public and private investors who understand that tend to sound calmer, ask better questions, and lose less sleep. Or at least they pretend more convincingly.

Conclusion

AI is creating opportunities across both public and private markets, but not in the same way and not on the same timeline. Public markets offer access to the current buildout: chips, cloud, data centers, energy, networking, and software companies translating AI into revenue and margin improvement. Private markets offer access to the next wave: vertical applications, workflow automation, model infrastructure, agents, and embodied systems that may define future categories.

The smartest way to think about AI is not as a single bet on a single company. It is as an ecosystem of builders, enablers, adopters, and emerging challengers. Some of the biggest winners will be obvious. Others will be hiding behind boring labels like electrical equipment, workflow software, or compliance tooling. That is usually how real technological revolutions work. The headlines celebrate the magic. The money often follows the plumbing.

So yes, talk your book. Just make sure your book includes more than a few glamorous names and a heroic amount of wishful thinking.