Have We Run Out of Data for AI?

Back in December,?Ilya Sutskever, AI-guru and ex OpenAI co-founder, said that we have reached?"peak data."?Elon Musk echoed the sentiment in January, saying that AI has “gobbled up” all human-produced data to train itself.

We only have one Internet they said, implying that we ran out of human data.

These make for punchy headlines for all of us who follow the AI boom. But for enterprises, it could lead to a false complacency, or worse, to cause them to ignore what they are sitting on: the goldmine of their own proprietary enterprise data.

The Three Layers

Modern AI, particularly generative AI, is built on the interplay of three critical layers: Compute Infrastructure, Foundational Models, and Data.

1. Compute Infrastructure: This is the backbone of AI, comprising high-end GPUs, hyperscale data centers, and deep neural networks. The energy demands of these systems are so immense that they’ve sparked renewed interest in nuclear energy to power them sustainably.
2. Foundational Models: These are the cutting-edge AI models developed by top-tier talent, often based on groundbreaking research like the seminal paper?"Attention is All You Need,"?which introduced the transformer architecture and revolutionized generative AI. These models are the engines driving today’s AI advancements.
3. Data: The often-overlooked layer, yet arguably the most critical. Massive amounts of data are required to train these models effectively. Without data, even the most sophisticated models are powerless

The Power of Proprietary Data in AI

If you are a regular user of LLMs, you will know that their true magic happens when they’re fine-tuned or augmented with?your specific context. (let's put aside the scary part about privacy for now, which deserves many other posts on their own). The full power of LLMs is on display when you ask it questions, and it already has all the finer points of your situation in memory and can then couple that with its foundational knowledge to generate the specific right answer for you.

The same principle applies to enterprises:

AI delivers its full potential when it’s trained and continuously aligned with your organization's specifics, quirks, edge cases, and industry nuances.

Several emerging AI strategies make this possible:

• RAG- or Retrieval-Augmented Generation
• Fine-tuning and Domain Adaptation
• Structured Grounding

These approaches ensure that your Enterprise AI doesn’t just rely on broad, generic knowledge—it integrates with your organization’s unique insights, producing tailored, high-value outputs.

The Goldmine of Enterprise Data

Your organization has proprietary data that no one else does.

Here are just a few categories:

• Customer Data – Insights into behavior, preferences, and feedback.
• Product & Services Data – Features, performance, and usage patterns.
• Market & Competitive Data – Trends, competitor insights, industry dynamics.
• Internal Communications – Emails, meeting notes, collaboration tools.
• Process & Workflow Data – How your business operates, including inefficiencies.
• Resource Data – Physical and digital assets, employee skills, and performance.
• Financial Data – Revenue, costs, profitability, and key financial metrics.
• Performance Data (KPIs) – Success metrics that track organizational goals.

This proprietary data is what makes you different and can help you win or lose in your market and industry.

When you hire top-tier consultancies like McKinsey, Bain or BCG - they come to you with their deep knowledge of industry and strategy. But they still need to immerse themselves in the specifics of your organization. That is why you have a discovery phase where they interview your key players for hours and turn every stone they can find.

Modern AI works the same way, it will need first to absorb and learn from all your proprietary data in order to deliver maximum value.

The Real Question: Is Your Enterprise AI- Data Ready?

So no, -we haven't not run out of data to train AI, we never will!

The more pressing question for enterprise leaders is:?What is your AI data readiness?

• Do you have a clear understanding of your data?
• Where is it stored, and how is it collected?
• Is it clean, secure, and usable?
• Is it properly labeled, indexed, and structured?
• What steps are you taking to leverage it for AI?

In all this AI frenzy and exuberance, data is more important than ever—especially proprietary data.

The organizations that succeed will be those that recognize the power of their data and invest in making it AI-ready. This means:

• Implementing robust data governance
• Ensuring data quality
• Building the right infrastructure to integrate AI into operations

In the AI era,?every company is a data company.

So the real question is: What are you doing with your data?

Eric's Note:

I enjoy writing these posts and I want to warmly thank you for being one of my faithful readers. Want to read all my best posts, on less restrained subjects, without LinkedIn between us? Subscribe (absolutely free!) to my other Newsletter on Substack. Click here and See you there!