When you build the chip, you reclaim control over the chain.
Global, September 2025
OpenAI is preparing a decisive leap: the launch of its own artificial intelligence chip, expected for 2026. The initiative, carried out in collaboration with semiconductor manufacturer Broadcom, marks an effort to reduce dependency on Nvidia and secure autonomy over the critical infrastructure that sustains advanced AI models.
The project has been evolving quietly since early 2025, when executives at OpenAI identified that reliance on external providers of high-performance GPUs had become a vulnerability. Growing demand, escalating costs, and limited supply left the company exposed. By designing proprietary silicon, OpenAI seeks to ensure both performance and cost-efficiency, tailoring hardware specifically to the needs of its models.
The new chip is expected to be used internally, not marketed to third parties. Broadcom’s leadership has reinforced confidence in the scale of the initiative by signaling major new orders tied to AI infrastructure. The company’s stock has risen in response, with analysts pointing out that custom AI chips are becoming the next battleground for influence and capital. OpenAI thus joins other global giants such as Google, Amazon, and Meta, all of which have invested in proprietary processors to strengthen performance and cut reliance on outside suppliers.
The partnership with Broadcom is not the only element of OpenAI’s strategy. Executives are also exploring fabrication options with Taiwan Semiconductor Manufacturing Company to bring designs into advanced production, aiming for cutting-edge three-nanometer processes. This diversification would provide redundancy against supply chain bottlenecks and strengthen energy efficiency, while offering a fallback to the cost-intensive pipeline dominated by Nvidia.
The implications extend far beyond technical specifications. In financial markets, Broadcom’s surge is compared to Nvidia’s earlier boom, as investors recognize that whoever controls custom AI silicon controls the foundations of tomorrow’s computational economy. Industry analysts note that hardware and software are no longer separate domains but increasingly intertwined elements of technological power.
For OpenAI, the pursuit of custom chips is part of a broader repositioning. The company is expanding its presence across multiple domains: testing employment-matching platforms that could compete with LinkedIn, acquiring design expertise through the integration of firms led by former Apple designers, and investing in user-facing hardware devices. These moves hint at OpenAI’s ambition to become not only a leader in software and models but also a vertically integrated technology company with end-to-end control over its ecosystem.
The shift can be read along three dimensions. The first is infrastructure sovereignty, as the company seeks to own the hardware that sustains its intelligence models, insulating itself from shortages or political disruptions. The second is supply chain diversification, spreading dependencies across partners in the United States and Asia to avoid being cornered by a single provider. The third is strategic independence, reinforcing OpenAI’s capacity to define its own trajectory instead of adapting to the priorities of external suppliers.
Observers highlight three dynamics in play. Computational self-sufficiency ensures that OpenAI can innovate at its own pace, shaping the architecture of its models without waiting for the next release from outside manufacturers. Market positioning reflects a maturing company that no longer wishes to be only a tenant in the data infrastructure space, but rather a builder of the foundation itself. Finally, geopolitical resilience arises from the ability to negotiate across multiple regions, balancing U.S. industrial policy, Asian manufacturing expertise, and global demand.
The decision to build custom chips is not without risks. Designing and producing semiconductors at scale requires enormous investment, complex coordination, and tolerance for high technical uncertainty. Yet in the logic of today’s AI race, remaining dependent on external suppliers may prove riskier still. In the eyes of its leadership, control over silicon is control over destiny.
By entering the hardware arena, OpenAI signals that the future of artificial intelligence will not be decided only by algorithms, but by the architecture that powers them. This evolution mirrors the path of other technology giants that realized long ago that to dominate a frontier, they needed to command every layer of its infrastructure.
Phoenix24: truth is structure, not noise.
Phoenix24: la verdad es estructura, no ruido.