Will There Be One AI to Rule Them All?

For the past few years, the AI story has been dominated by the race to build colossal, all-purpose models. Digital equivalents of Swiss Army knives that can handle everything from writing poetry to coding software.

That’s because the bigger the model, the better its answers.

Financial news outlet Bloomberg created “BloombergGPT” to gain an edge in analysis and forecasting. It fed the model every scrap of data Bloomberg had gathered over 40+ years.

GPT-4, which wasn’t specifically trained for finance, smoked it in tasks like figuring out the sentiment of earnings calls. It didn’t matter how much high-quality, “proprietary” data Bloomberg had because GPT-4 was 100X bigger.

It was the same story when Google’s general-purpose AI model went head to head with a model specifically trained for healthcare. Google’s general model won handily.

Lesson: If you want the best AI, just make it massive

But there’s a major shift underway, one that savvy investors must understand. The future isn’t one AI to rule them all. It’s a flourishing ecosystem of specialized, on-premises models optimized for specific jobs.

Three powerful forces are driving this shift:

1. Keep your crown jewels safe. If you’re running a bank, hospital, or defense contractor, you simply cannot send your most sensitive data to GPT-4 or Claude. Full stop. Even for companies without regulatory constraints, the risks of exposing proprietary data are too great. Would Coca-Cola Co. (KO) send its secret formula to an external AI? Not a chance. These companies will bring the AI in-house where their data never leaves the security perimeter.
2. Call an expert. China’s AI model DeepSeek handed us a gift when it pioneered the “Mixture of Experts” (MoE) approach. Instead of one massive model where every parameter works on every problem, MoE divides the model into specialized “experts.”
   When you ask a question, the system routes it only to the relevant experts. For example, the math expert for calculations or the reasoning expert for logic problems. Since only a fraction of the model activates for any given task, you get faster responses, lower costs, and better performance. Every AI company is now adopting MoE.
3. A unique edge. The brutal truth about AI models is that you can train them on as many GPUs as you want, but without unique data, they’re commodities racing to the bottom. Open-source models are catching up to proprietary ones at breakneck speed.
   The only sustainable advantage is unique data no one else can access. This is why I’m bullish on:
   • xAI’s Grok, fueled by X’s social network;
   • Meta’s models, trained on Facebook, Instagram, and WhatsApp;
   • Google’s Suite, which can leverage the vast treasure troves of YouTube and Gmail.

This is also a huge advantage for specialized, on-premises models built on proprietary company data.

The next wave isn’t building one AI to rule them all. It’s building thousands of specialized AIs, each ruling its own domain.

Companies building specific AIs on their unique data assets will create billions of dollars in value. These models are being crafted right now, largely out of the public eye.

Feeding the AIs

Feeding data into AI models isn’t a simple copy-paste job. According to S&P Global, a staggering 60% of an AI developer's time isn’t spent building cool chatbots or features. It’s spent on the grueling work of gathering, preparing, and cleaning data.

I’m sure you’ve heard the analogy “data is the new oil.” It’s imperfect, as data can be reused infinitely and grows rather than depletes. But in one crucial respect, the comparison is spot-on: Both must be refined before they’re valuable.

When crude oil comes out of the ground, it’s a toxic, unusable sludge. It must go through complex refinement processes before becoming the gas that powers your car.

Data is the same. The raw information companies collect is messy, inconsistent, and often incomprehensible to AI systems.

A global retailer like Target (TGT) has reams of customer data: purchase histories, browsing patterns, demographic information, in-store traffic… all scattered across hundreds of systems in dozens of formats.

Refining it into machine-readable data that can be fed into AIs is the real moneymaker.

Companies that can clean and structure this data won’t just participate in the AI revolution, they’ll lead it. And this trend of refining data becomes even more important as we move into the world of unstructured data.

AI is moving fast, and the biggest winners won’t always be the most obvious plays. If you want to stay ahead of the crowd and learn more about the next wave of AI opportunities before they hit the mainstream, sign up for The Jolt—my free investing letter.

______________

Stephen McBride is Chief Analyst, RiskHedge. To get more ideas like this sent straight to your inbox every Monday and Friday, make sure to sign up for The Jolt, a free investment letter focused on profiting from disruption.