A Multi-Year Alliance to Build the World’s Largest AI Infrastructure, Unlocking Frontier Intelligence for Humanity
- Massive Scale Deployment: OpenAI and Broadcom are partnering to roll out 10 gigawatts of custom AI accelerators across racks equipped with advanced Ethernet networking, starting in late 2026 and wrapping up by 2029, to fuel the explosive growth in AI demand.
- Custom Hardware Innovation: By designing its own chips and systems, OpenAI embeds lessons from its frontier models like ChatGPT directly into the hardware, enabling unprecedented levels of AI capability and efficiency.
- Broader Ecosystem Impact: This collaboration not only boosts OpenAI’s mission to benefit humanity through AGI but also sets new standards for scalable, power-efficient AI datacenters, reinforcing Ethernet’s role in the global AI infrastructure race.
In a landmark move that’s set to reshape the landscape of artificial intelligence, OpenAI and semiconductor giant Broadcom have unveiled a strategic collaboration to deploy a staggering 10 gigawatts of custom AI accelerators. Announced on October 13, 2025, from the tech hubs of San Francisco and Palo Alto, this multi-year partnership goes beyond mere supply agreements—it’s a bold co-development effort aimed at creating next-generation AI clusters that can handle the surging global demand for intelligent systems. With OpenAI’s expertise in frontier AI models and Broadcom’s prowess in networking and chip fabrication, the duo is positioning itself to deliver racks of accelerators and Ethernet-based connectivity solutions, starting deployment in the second half of 2026 and completing the massive buildout by the end of 2029. This isn’t just about hardware; it’s about embedding the very essence of AI innovation into silicon to unlock new frontiers of capability.
At the heart of this alliance is OpenAI’s decision to design its own AI accelerators and integrated systems, a strategic pivot that allows the company to directly incorporate insights from developing groundbreaking products like ChatGPT. Sam Altman, OpenAI’s co-founder and CEO, emphasized the partnership’s role in building the infrastructure essential for AI’s transformative potential. “Partnering with Broadcom is a critical step in building the infrastructure needed to unlock AI’s potential and deliver real benefits for people and businesses,” Altman stated. By crafting custom chips, OpenAI can optimize hardware for the specific demands of training and running massive language models, moving away from reliance on off-the-shelf solutions from competitors like Nvidia. This approach promises to enhance performance, reduce costs, and accelerate the path toward artificial general intelligence (AGI), where AI systems could rival human-level reasoning across diverse tasks.
Broadcom, a leader in semiconductor solutions, brings its end-to-end portfolio of Ethernet, PCIe, and optical connectivity to the table, ensuring the racks are scaled entirely with standards-based networking for both scale-up (within a single system) and scale-out (across distributed clusters) operations. Hock Tan, Broadcom’s President and CEO, hailed the deal as a “pivotal moment in the pursuit of artificial general intelligence,” noting OpenAI’s pioneering role since the ChatGPT breakthrough. The collaboration builds on long-standing agreements between the two companies for co-developing and supplying AI accelerators, now formalized through a term sheet that outlines the deployment of these systems across OpenAI’s facilities and partner data centers worldwide. Charlie Kawwas, Ph.D., President of Broadcom’s Semiconductor Solutions Group, highlighted how custom accelerators pair seamlessly with Ethernet to create cost- and performance-optimized AI infrastructure, setting new industry benchmarks for open, scalable, and power-efficient datacenters.
From a broader perspective, this partnership underscores the intensifying race to secure the computational backbone of AI. OpenAI, which has exploded to over 800 million weekly active users and seen robust adoption by global enterprises, small businesses, and developers, is under immense pressure to scale its operations amid skyrocketing energy and hardware needs. The 10-gigawatt deployment—equivalent to the power output of several large nuclear plants—addresses this by prioritizing efficiency and Ethernet’s flexibility over proprietary alternatives, potentially lowering barriers for widespread AI adoption. Greg Brockman, OpenAI’s co-founder and President, captured the excitement: “Our collaboration with Broadcom will power breakthroughs in AI and bring the technology’s full potential closer to reality. By building our own chip, we can embed what we’ve learned from creating frontier models and products directly into the hardware, unlocking new levels of capability and intelligence.”
This initiative reinforces OpenAI’s core mission to ensure AGI benefits all of humanity, rather than concentrating power in the hands of a few. For Broadcom, it solidifies its leadership in AI infrastructure, validating Ethernet as the go-to technology for datacenters handling exabytes of data and trillions of parameters in AI training. As AI permeates every sector—from healthcare diagnostics to autonomous vehicles—this collaboration could democratize access to advanced intelligence, fostering innovations that drive economic growth and solve global challenges. Yet, it also raises questions about energy consumption and sustainability in an era where AI’s thirst for power is one of its biggest hurdles. Ultimately, OpenAI and Broadcom’s alliance isn’t just a tech deal; it’s a blueprint for the intelligent future, where custom hardware meets visionary software to propel humanity forward.