At a critical juncture for the global semiconductor industry, Lip-Bu Tan, Intel board member and a legendary figure in chip design, provided a revealing interview to Bloomberg Tech. His central message was clear: the era of traditional Generative AI is giving way to Agentic AI, and this transition demands a radical rethink of how tech giants collaborate.
Tan, whose experience spans from leading Cadence Design Systems to providing strategic guidance for Intel’s turnaround, argued that Agentic AI is not merely an evolution of Large Language Models (LLMs). It represents a fundamental paradigm shift where AI ceases to be a passive conversationalist and becomes an active "agent" capable of planning, making decisions, and executing complex tasks across multiple systems without constant human intervention.
The Architecture of Autonomy: Why Hardware Isn't Enough
During the discussion, Tan emphasized that Intel can no longer rely solely on its manufacturing prowess. "Agentic AI requires a holistic approach that combines hardware, software, and data networks," he noted. The complexity of these new systems means that no single organization, no matter how large, possesses all the answers. Intel, under Pat Gelsinger’s leadership and with Tan’s input, is investing heavily in what it calls "partner networks."
These networks are not just commercial agreements but deep technical integrations. The need for low latency and high energy efficiency in Agentic AI means that the silicon must be designed in direct coordination with the software it will run. Tan pointed out that partnerships with companies like Microsoft, Google, and smaller specialized startups are the only way to ensure that Intel’s x86 architecture—and its new forays into GPUs and AI accelerators—remain relevant in an autonomous world.
The Role of Open Ecosystems
One of the most compelling points of the interview was the emphasis on open standards. In contrast to Nvidia’s closed CUDA ecosystem, Intel is betting on OneAPI and initiatives like UALink (Ultra Accelerator Link). Tan argued that Agentic AI will only thrive if there is interoperability. "Our customers don't want to be locked into a single vendor," he stated. "They want the freedom to move their AI agents from the cloud to the edge and back without having to rewrite their code from scratch."
This approach is part of Intel’s broader IDM 2.0 strategy, which includes opening its foundries to third parties. Tan sees Agentic AI as the primary driver of demand for Intel’s next-generation lithography (such as Intel 18A). As AI agents become more autonomous, the need for custom silicon will skyrocket, and Intel aims to be the platform upon which this future is built.
Challenges and Geopolitical Context
Despite the optimism, Lip-Bu Tan did not shy away from tough questions regarding competition and geopolitics. Nvidia’s market dominance in AI chips remains the most significant hurdle for Intel. Furthermore, export restrictions to China and turbulence in the global supply chain complicate the creation of the international partner networks Tan envisions.
However, he believes that the focus on Agentic AI offers a "second chance" for Intel. While the first phase of AI was about training models (where Nvidia dominates), the next phase is about execution and autonomy in real-time. Here, Intel’s experience in data processing and its presence in millions of edge devices—from PCs to industrial robots—gives it a strategic advantage, provided it can effectively activate its partner network.
Conclusion: Collective Intelligence as a Business Model
Lip-Bu Tan’s intervention underscores a fundamental truth for 2026: artificial intelligence is no longer a solo sport. Agentic AI requires a "collective intelligence" at the industry level. For Intel, success will not be judged solely by the number of transistors it can fit on a chip, but by the bonds it can forge with the broader ecosystem. As Tan concluded, "The future belongs to those who can collaborate faster than they can compete."
