Table Of Contents
Larry Ellison, cofounder and Chief Technology Officer of Oracle, has delivered a sharp critique of the current artificial intelligence boom, arguing that the Biggest Problem With AI Models today is not performance, scale, or speed, but sameness.
Speaking during Oracle’s fiscal Q2 earnings call, Ellison said that leading AI systems such as ChatGPT, Gemini, Grok, and Llama share a fundamental limitation. According to him, they all learn from largely the same publicly available data, which makes them increasingly difficult to differentiate.
His comments arrive at a moment when governments, enterprises, and investors are pouring billions into AI infrastructure, expecting long-term competitive advantages. However, Ellison believes many of those expectations may be misplaced unless the industry changes direction.
Why Ellison Says AI Models Are Becoming Commodities
Ellison explained that most large language models rely heavily on open internet data, including public websites, forums, books, and online documents. As a result, these models absorb similar knowledge, learn similar patterns, and often generate similar outputs.
Because of this shared foundation, AI systems now compete more on branding and pricing than on meaningful intelligence gaps. In Ellison’s view, this trend pushes AI toward commoditisation, where no single model offers a truly unique advantage.
He warned that this dynamic represents the Biggest Problem With AI Models, especially for enterprises hoping to gain strategic value from artificial intelligence rather than generic automation.
Moreover, Ellison stressed that simply building larger models or adding more computing power will not solve the issue. Instead, he argued that scale alone only accelerates the same limitations.
The Shift Toward Private and Proprietary Data
To address this challenge, Ellison believes the next phase of AI must focus on private, enterprise-specific data rather than public internet sources. He described this transition as the industry’s “second wave” of artificial intelligence.
Unlike public data, private datasets contain detailed operational records, financial information, customer behavior, medical histories, and supply-chain intelligence. When AI systems can securely analyze this information, they can deliver insights that generic models cannot match.
Importantly, Ellison emphasized that privacy, security, and regulatory compliance must remain central to this shift. Enterprises cannot afford to expose sensitive data while experimenting with AI tools. Therefore, he sees secure data access as a prerequisite for meaningful AI progress.
This transition, he argued, directly addresses the Biggest Problem With AI Models by restoring differentiation and business value.
Oracle’s Strategy and the Industry Debate
Oracle’s current AI roadmap aligns closely with Ellison’s assessment. The company is investing heavily in cloud infrastructure and AI platforms designed specifically for enterprise workloads.
A major focus lies in enabling AI systems to interact with private databases through techniques such as retrieval-augmented generation. This approach allows models to reference live internal data without permanently absorbing it into their training sets.
Ellison also pointed to Oracle’s massive capital expenditure plans, which include tens of billions of dollars in cloud and AI infrastructure investments. These investments aim to support large-scale AI workloads while maintaining strict data controls.
Because Oracle already manages vast amounts of enterprise data globally, Ellison believes the company holds a natural advantage as organizations move toward data-centric AI deployment.
However, Ellison’s view has not gone unchallenged. Some AI researchers argue that synthetic data, federated learning, and improved model architectures could still deliver differentiation without relying exclusively on private datasets.
At the same time, competitors such as Microsoft, Google, and Amazon are aggressively expanding their own enterprise AI offerings. Each company is racing to position itself as the default AI platform for businesses.
Still, Ellison remains firm. He insists that without access to unique, high-quality data, AI systems will continue to converge in capability rather than diverge.
A Defining Moment for AI’s Future
As artificial intelligence moves from experimentation to widespread deployment, Ellison’s warning highlights a critical inflection point. The industry must decide whether AI becomes a utility tool with limited differentiation or a deeply integrated intelligence layer tailored to each organization.
Ultimately, his message is clear: solving the Biggest Problem With AI Models requires shifting focus away from public data scale and toward secure, proprietary intelligence. How quickly the industry adapts may determine which companies lead the next decade of AI innovation.
Also Read: Anthropic CEO Warns The World About Unseen AI Dangers Ahead
