OpenAI has confirmed an alliance with Broadcom to design and manufacture its own artificial intelligence chips , a move with which it seeks to secure computational power and adjust costs amidst the expansion of its generative models.
As both companies have explained, OpenAI will take charge of the design of the accelerators and Broadcom will be responsible for the development, manufacturing and deployment, with a schedule that points to the second half of 2026 and an infrastructure sized for around 10 gigawatts of electricity.
What the agreement includes and how the functions are distributed
The plan includes systems based on Broadcom's Ethernet stack and network technology , integrated both in OpenAI's own data centers and in external partner infrastructures, with network, memory and computing integration tailored to their workloads.
The company argues that with hardware specifically designed for its models, it will be able to gain efficiency, accelerate training and inference, and ultimately offer faster and more affordable models for businesses and users.
Sam Altman described the move as key to building the necessary infrastructure to bring the benefits of AI to scale, while Hock Tan, CEO of Broadcom, noted that OpenAI is driving some of the most advanced frontier models.
Timing, magnitude, and consumption: the 10 GW bar
The rollout would begin in late 2026 with an aggregate capacity equivalent to 10 GW , a figure that, for comparative purposes, far exceeds the electricity generated by the Hoover Dam and could supply more than eight million homes in the United States.
This leap in scale puts electrical suppliers and the data center supply chain to the test, as they will have to meet demands for energy, cooling, and components at a pace that is unusual even for the technology industry.
Industry sources estimate that a 1 GW data center can require around $50.000 billion in investment, of which about $35.000 billion is typically allocated to chips and accelerators; extrapolated, the financial effort for 10 GW would be enormous.
Less dependence on Nvidia and side deals
The alliance with Broadcom fits into the strategy of diversifying suppliers and reducing dependence on Nvidia , whose offerings and prices have strained the market. In parallel, OpenAI maintains a 6 GW contract with AMD and collaborates with Nvidia on another 10 GW of infrastructure, bringing the total commitment to around 26 GW of capacity.
In addition to these partners, the company has advanced agreements with Oracle for cloud infrastructure and discussions with Samsung and SK Hynix related to the supply of memory and critical components for its future systems.
The financial terms of the agreement with Broadcom have not been made public , nor has the financing arrangement been detailed. According to reports in the financial press, AMD granted OpenAI a warrant for up to 160 million shares , subject to capacity deployment conditions, in addition to the supply of its upcoming MI450 series.
Broadcom's market and position in the ecosystem
Following the announcement, Broadcom's stock price rebounded by nearly 10% , mirroring the movements seen in AMD and Oracle after their respective deals with OpenAI. For Broadcom, this is one of its most significant contracts and reinforces its shift from networks and telecommunications to large-scale AI infrastructure.
The company already supplies XPU accelerators and solutions to giants like Google, Meta, and ByteDance, and has seen its revenue and market capitalization grow strongly since 2022 thanks to generative AI.
Operational challenges: energy, costs and efficiency
The AI ​​business is still far from being fully profitable for many players, and building data centers of this scale involves complex permits, massive investment, and energy consumption that can strain local and regional networks.
OpenAI believes that custom chips will help contain computing costs and improve energy efficiency per task, reducing bottlenecks compared to buying commercial processors in an increasingly competitive market.
The joint commitment of OpenAI and Broadcom marks a turning point in the race for AI infrastructure : a proprietary accelerator program with a 2026 horizon, a 10 GW energy target, and parallel alliances with AMD, Nvidia, and other partners that aim to secure supply, lower costs, and scale capabilities without depending on a single provider.
