The Biden administration is weighing new controls on the export of graphics cards and related AI hardware to select destinations, aiming to curb the flow of advanced computing technology to China and Russia. A recent Bloomberg report describes a staged framework that could be rolled out in three levels, each calibrated to different export conditions. The plan is framed as a national security measure intended to limit access to high end GPUs and AI accelerators used in data centers, research labs, and advanced manufacturing, while preserving supply for the United States and its closest allies. The regime would begin with a baseline that imposes no export restrictions, followed by a second level that restricts computing power in exportable devices, and a third level that would ban the export of AI chips altogether. The timing under discussion points to a gradual rollout, with new safeguards introduced as the security landscape evolves. The approach would apply to sophisticated graphics cards and AI accelerators while keeping Canada, Australia, Japan, and Western European partners outside the tightest controls. Bloomberg reports.
Under the first tier, shipments would continue without new licensing hurdles for covered countries. The second tier would tie export licensing to the device computing power, effectively restricting the sale of devices that exceed specified performance thresholds. The target discussed in policy circles envisions limiting overall export volumes to roughly 50,000 GPUs through 2027. Enforcement would likely consider factors such as processor speed, memory capacity, tensor cores, and AI acceleration features. The plan would avoid disrupting the United States market and its closest allies, ensuring continued access for government offices, universities, and industry customers in Canada, Australia, Japan, and Western Europe through existing channels.
The top tier would prohibit the export of AI chips entirely. That ban would cover chips designed for AI workloads as well as high end GPUs that play a central role in AI model training and inference. In Russia, observers warn that such a policy could push prices higher as unofficial imports and gray market routes adapt to new restrictions. Nvidia has signaled a price point of about 1,999 dollars for its flagship GeForce RTX 5090 when it reaches shelves, which translates to roughly 203,700 rubles at current exchange rates. The timing of sanctions details will influence how the market responds to the RTX 5090 launch, and the ultimate effect on Russian pricing remains uncertain as channels adjust.
Nvidia has previously argued that restrictions on AI chip exports could hinder the United States by slowing innovation and limiting access to essential technologies for researchers and enterprises. As the policy takes shape, industry players seek clarity on licensing criteria, enforcement rules, and the scope of products covered. The broader implication is that geopolitics will continue to steer technology trade, prompting buyers, suppliers, and researchers to monitor regulatory signals as the global GPU ecosystem recalibrates to a tighter export regime. Bloomberg reports.