The US decision to rescind the previous Joe Biden administration's export control rule on AI chips and GPUs, which was slated to take effect on May 15, has come as a relief for Indian cloud companies and data centre providers who were on edge.
While revoking the Artificial Intelligence (AI) Diffusion Rule, the US commerce department's Bureau of Industry and Security (BIS) announced some measures to fortify export controls on AI chips.
It warned that use of Chinese firm Huawei’s Ascend AI chips anywhere globally would be seen as a violation of US export controls. It also warned against utilising US AI chips for the training and inference of Chinese AI models.
The BIS also published guidance to US companies on strategies to protect their supply chains against diversion tactics.
While the restriction on Huawei Ascend chips will not impact Indian companies much as they are not used here, the US move to track movement of its chips has raised some concerns.
It means that trackers will have to be put in every shipment, no matter which country it goes to, industry executives said.
“Disabling the chip may be a drastic step, but establishing a foolproof tracking system at the physical level will require time to implement,” Deepika Giri, assistant vice president, big data analytics (BDA) and AI research, at IDC Asia Pacific, said, acknowledged the complexity of implementing the new controls.
“An alternative could be to use keys, similar to those in software licences, which are sent to the purchaser and prevent unauthorised usage by non-licensed users,” she said.
The AI Diffusion Framework—which proposed an annual cap of 50,000 GPUs for India—would have severely impacted India’s ability to import high-end chips like Nvidia’s B200, with estimates suggesting a limit of just 20,000 units under the proposed total processing performance (TPP) limit, industry insiders had said.
The Donald Trump administration said it intends to pursue a more “bold, inclusive strategy” for the global dissemination of American AI technology among trusted allies while safeguarding it from adversaries.
Cloud company ESDS Software Solution utilises Nvidia chips, including the A100 and H100, and anticipates the B200 GPU rollout in India within the next three to four months.
Piyush Somani, founding chief executive of ESDS, expressed concern about the potential delay of B300 GPUs, which might take over a year to reach India without policy intervention.
Somani, who is also the president of the Cloud Computing Innovation Council of India, highlighted the exponential surge in India's AI compute requirements, having increased over fivefold in the last 18 months, driven by demand from sectors like banking, agritech, and government agencies.
He emphasised the critical importance of data sovereignty and self-reliance in AI infrastructure for India. "If India is to truly lead in AI, we can’t afford to depend on foreign cloud platforms where data exits our country, sometimes even unknowingly,” Somani said.
Prateek Jhawar, managing director and head of infrastructure and real assets investment banking at Avendus Capital, said Huawei Ascend chips are only being used in China and they are unlikely to come to India anytime soon.
“But fundamentally, they can provide a low-cost alternative to Nvidia chips, and that is why the US is trying to create blockage around them from being used by other countries,” he said.
Jhawar also raised concerns about the practicality of enforcing the new regulations. “If they intend to track, they will have to include trackers in every shipment, no matter which country it goes to.”
He, however, added, “We're yet to see the formal guidance from the US on the consequences of not following this rule.”
While revoking the Artificial Intelligence (AI) Diffusion Rule, the US commerce department's Bureau of Industry and Security (BIS) announced some measures to fortify export controls on AI chips.
It warned that use of Chinese firm Huawei’s Ascend AI chips anywhere globally would be seen as a violation of US export controls. It also warned against utilising US AI chips for the training and inference of Chinese AI models.
The BIS also published guidance to US companies on strategies to protect their supply chains against diversion tactics.
While the restriction on Huawei Ascend chips will not impact Indian companies much as they are not used here, the US move to track movement of its chips has raised some concerns.
It means that trackers will have to be put in every shipment, no matter which country it goes to, industry executives said.
“Disabling the chip may be a drastic step, but establishing a foolproof tracking system at the physical level will require time to implement,” Deepika Giri, assistant vice president, big data analytics (BDA) and AI research, at IDC Asia Pacific, said, acknowledged the complexity of implementing the new controls.
“An alternative could be to use keys, similar to those in software licences, which are sent to the purchaser and prevent unauthorised usage by non-licensed users,” she said.
The AI Diffusion Framework—which proposed an annual cap of 50,000 GPUs for India—would have severely impacted India’s ability to import high-end chips like Nvidia’s B200, with estimates suggesting a limit of just 20,000 units under the proposed total processing performance (TPP) limit, industry insiders had said.
The Donald Trump administration said it intends to pursue a more “bold, inclusive strategy” for the global dissemination of American AI technology among trusted allies while safeguarding it from adversaries.
Cloud company ESDS Software Solution utilises Nvidia chips, including the A100 and H100, and anticipates the B200 GPU rollout in India within the next three to four months.
Piyush Somani, founding chief executive of ESDS, expressed concern about the potential delay of B300 GPUs, which might take over a year to reach India without policy intervention.
Somani, who is also the president of the Cloud Computing Innovation Council of India, highlighted the exponential surge in India's AI compute requirements, having increased over fivefold in the last 18 months, driven by demand from sectors like banking, agritech, and government agencies.
He emphasised the critical importance of data sovereignty and self-reliance in AI infrastructure for India. "If India is to truly lead in AI, we can’t afford to depend on foreign cloud platforms where data exits our country, sometimes even unknowingly,” Somani said.
Prateek Jhawar, managing director and head of infrastructure and real assets investment banking at Avendus Capital, said Huawei Ascend chips are only being used in China and they are unlikely to come to India anytime soon.
“But fundamentally, they can provide a low-cost alternative to Nvidia chips, and that is why the US is trying to create blockage around them from being used by other countries,” he said.
Jhawar also raised concerns about the practicality of enforcing the new regulations. “If they intend to track, they will have to include trackers in every shipment, no matter which country it goes to.”
He, however, added, “We're yet to see the formal guidance from the US on the consequences of not following this rule.”
You may also like
Coco Gauff serves to empty court at Italian Open with opponent absent
Mike Tindall responds to Zara's 'brutal admission' about the first time they met
India's apparel exports clock strong growth in April driven by US market
Shark Tank's Anupam Mittal is taking his family for Kashmir trip. Shares a message in viral post
Delhi CM supports '100 per cent boycott' of Turkish apples, institutions
