Just when investors were thinking they understood the AI Boom, the rules shifted again…

Forget everything you thought you knew about the “Magnificent 7.” We’ve moved beyond the individual gladiators and entered the era of Consolidated AI Supergroups.

Until recently, the tech narrative has centered around Nvidia‘s (NVDA) meteoric rise, Microsoft‘s (MSFT) AI integrations, Google‘s Gemini ambitions, and Meta‘s (META) open-source gambit. Adorable, really – like watching kids play with LEGOs before someone introduces them to a master builder with an industrial crane.

But now we are no longer in a playground skirmish. It has become a full-blown, no-holds-barred, multi-hundred-billion-dollar arms race. It’s being fought by three distinct, increasingly formalized, and ridiculously well-funded super-teams. And each believes that they alone possess the key to Artificial General Intelligence (AGI).

Welcome to the Great AI War of 2026.

Why the AI Industry Is Entering a New Arms Race

At the moment, it doesn’t matter who’s winning – not yet anyway. 

We’re still early in this marathon. Predicting a winner now is like declaring victory for the guy who sprinted out of the starting gate; an impressive initial burst, perhaps, but historically, a terrible predictor of who crosses the finish line first.

What is already unequivocally clear, though, is that these titans are going to spend money like it’s going out of style. And the vast majority of that capital will cascade directly into the AI supply chain.

Now, let’s meet our competitors, shall we?

“Elon Co.”: Vertical Integration at the Edge of the AI Industry

By merging SpaceX, xAI, and Tesla (TSLA), Musk isn’t just building AI; he’s building a vertically integrated, energy-independent, space-faring AI entity. The core idea is “Orbital Compute” – taking the entire data center off-world, leveraging the vacuum of space for cooling and perpetual solar for power, all connected by Starlink. And let’s not forget Optimus, the physical manifestation of his AI ambitions, designed to perform Earthly tasks while their brains are in orbit.

To accomplish this, he’s betting on radical innovation with a complete disregard for terrestrial limitations.

• Why They’re Spending: To build literally everything from scratch – new rockets for bigger satellites, new radiation-hardened chips, new AI models, new robots.

The Googlopoly: Custom Silicon and the Fight for AI Efficiency

Google, Meta, Anthropic, and Broadcom (AVGO) are coalescing around a strategy of sheer, unadulterated efficiency through custom silicon. Gemini (Google’s offering) aims to own the consumer AI space – embedded everywhere, ambient, invisible, and indispensable. Claude (Anthropic’s creation) is gunning for enterprise AI, becoming the go-to “reasoning engine” for every boardroom and data analytics department. The critical piece? Google’s Tensor Processing Units (TPUs), now being custom-designed with Broadcom and sold directly to partners like Anthropic. Public reports suggest Broadcom/Google AI silicon alliances could drive high-single-digit to low-double-digit billions in commitments over the next several years.

Google is rooting itself far and wide to ensure it is necessary.

• Why They’re Spending: To perfect the custom ASIC, to drive the cost-per-token down to levels no one else can match, and to entrench its models so deeply into enterprise and consumer ecosystems that extracting them would be like trying to remove a deeply embedded splinter from the global economy.

The OpenAI Empire: Brute-Force Scale in the AI Industry

Nvidia, Amazon (AMZN), Microsoft, and SoftBank (SFTBY) have effectively formed a financial and infrastructural behemoth around OpenAI to create a “Sovereign AI” cloud – a global network of “AI Factories” so vast and powerful they make existing data centers look like glorified server closets. They’re betting that raw computational power, fed into increasingly sophisticated large language models, will be the ultimate differentiator.

In other words, OpenAI is becoming an undisputed heavyweight champion with access to unprecedented pools of capital and compute.

• Why They’re Spending: Because they can. They are aggressively pre-purchasing H200s, future Rubin GPUs, electricity, and land to house their monstrous AI factories. Their thesis is simple: the bigger the model, the better the AI. And bigger models require bigger everything.

The Investment Case: How AI Spending Really Translates Into Profits

So, you see the problem, right? 

Each of these super-teams is fundamentally altering the definition of what it means to “build AI.” They’re not just iterating on software; they’re building entirely new industries, reinventing energy grids, and pushing the boundaries of physical infrastructure.

And this is what makes a shrewd investor smile – because while everyone else is debating whether GPT-5 is smarter than Claude 4 or if Optimus will be externally available by Christmas, you should be buying the picks and shovels.

Every dollar of the hundreds of billions these three entities are pouring into this race will flow through a very specific set of choke points in the AI supply chain. 

The AI Supply Chain Bottlenecks Investors Shouldn’t Ignore

The Cooling Conglomerates: Nvidia’s Blackwell and the upcoming Rubin GPUs run hot – too hot for air cooling to be effective. Liquid cooling is rapidly becoming mandatory for next-generation, high-density AI racks. Companies like Vertiv (VRT), which reported multi-billion-dollar backlogs for cooling and critical infrastructure solutions, are positioned to benefit as liquid cooling becomes standard.

The High-Speed Network Weavers: These AI models aren’t just running on one server; they’re running on tens of thousands, connected by a seamless “fabric.” The data has to move at exceptional speeds between GPUs. Arista Networks (ANET) leads in 800G and next-gen Ethernet, technologies essential for high-throughput AI clusters.

The Power Grid Guardians: Data centers are energy devourers. We’re talking about gigawatts of power; and one gigawatt is enough to power roughly 750,000 to 1 million homes. Eaton (ETN) and similar electrical infrastructure providers supply critical power distribution and grid management gear for high-density AI facilities. Every new data center, expansion, and added megawatt is a direct revenue stream.

The Custom Silicon Tailors: While Nvidia dominates, the “Googlopoly” (and even Elon Co., to some extent) is pushing hard into custom Application-Specific Integrated Circuits (ASICs) to break vendor lock-in and optimize for specific workloads. Broadcom (AVGO) – designing chips for Google – and Marvell Technology (MRVL) – providing specialized interconnects and DSPs for custom AI silicon – are perfectly positioned. Even Arm Holdings (ARM) benefits from broad adoption of Arm-based CPUs and SoCs in power-efficient AI workloads.

The Energy Giants: Traditional utilities are now AI plays. Companies like Vistra (VST) and Constellation (CEG) are signing multi-year power purchase agreements with hyperscalers for gigawatt-scale capacity. For context, Constellation reported over 2.6 GW of power purchase agreements with major cloud and AI customers as of late 2025 – an amount roughly equivalent to the capacity of a large nuclear plant.

The AI Host Specialists: Forget the old-school real estate investment trusts (REITs). Today’s AI factories need specialized data center environments with extreme power density and advanced cooling. Applied Digital (APLD), with its focus on high-density AI hosting, is a direct beneficiary as hyperscalers outsource high-density AI workloads.

The Circular Spending Debate – and Why AI Infrastructure Still Wins

Now, a word of caution – some market watchers highlight what they call “internal spending loops” – where capital circulates among partners. It goes something like this: Nvidia invests $30 billion into OpenAI, which then spends that money on Microsoft Azure, which Microsoft then uses to buy more chips from Nvidia. It’s a recursive business model.

But here’s why – for now – that risk is irrelevant to the supply chain thesis.

This still involves real purchases of chips, infrastructure, and services, supporting supplier revenue regardless of balance-sheet mechanics.

Regardless of whose money it originally was or how many times it circles between these super-teams, it still represents a tangible purchase of physical – and profitable – goods and services. 

The race is heating up. The spending is accelerating. And the physical infrastructure needed to win (or even just compete) in this Great AI War is immense. Don’t try to pick the ultimate victor in mile three of a marathon. Instead, invest in the companies that are selling the sneakers, the water bottles, and the medical supplies to all the runners… Because one thing is clear: they’re all going to be running – and spending – until the very last mile.

And that finish line is a long way off.

If there’s one lesson from every major technology arms race, it’s this: the biggest fortunes aren’t always made inside the super-teams themselves.

They’re made by spotting the unexpected companies riding the same exponential curve – often in places most investors never think to look.

That’s why I recently flew to Silicon Valley to pressure-test what this Great AI War really means on the ground – not just for Nvidia or Big Tech, but for the next wave of companies quietly using AI to scale faster than ever before.

I believe one small, under-the-radar stock could emerge as the next “Amazon moment,” in a completely unexpected sector that AI is now reinventing from the inside out.

In my latest presentation, I break down my “Hyperscale” investing approach, explain why AI is creating the fastest wealth transfer in modern history, and reveal how everyday investors can position themselves before exponential progress does the heavy lifting.

If you want to see where I believe the next breakout could come from – and how I’ve already identified winners early in past tech cycles – you can watch that new briefing here.

The post The AI Industry Has Split Into Supergroups. Investors Need to Adapt. appeared first on InvestorPlace.

Source link