US firms adopt DeepSeek despite scrutiny

Despite mounting regulatory scrutiny, Chinese company DeepSeek's artificial intelligence breakthrough is seeing widespread adoption among US technology firms, with industry experts predicting the efficiency gains will accelerate AI implementation across various sectors.

Microsoft is among the first to embrace the innovation, integrating DeepSeek's reasoning model R1 into its developer platforms Azure AI Foundry and GitHub. Amazon has followed suit, enabling developers to leverage what it calls the "powerful, cost-efficient" R1 model through Amazon Web Services.

Nvidia, a leading chip designer, has also adopted R1, allowing developers to deploy and use AI programs on their personal systems. The company praised the model's "state-of-the-art reasoning capabilities", "high inference efficiency" and "leading accuracy" for tasks requiring logical inference, reasoning, mathematics, coding, and language understanding.

Tech executives have responded positively, with many addressing DeepSeek's effect during recent earnings calls.

Microsoft CEO Satya Nadella highlighted the technology during a Jan 29 earnings call, Yahoo Finance reported.

"I think DeepSeek has had some real innovations," Nadella said. "For a hyperscaler like us, a PC platform provider like us, this is all good news as far as I'm concerned."

Qualcomm CEO Cristiano Amon expressed particular enthusiasm about DeepSeek's innovations, suggesting they could accelerate AI-driven demand for smartphones and personal computers.

"DeepSeek-R1 and other similar models recently demonstrated the AI models are developing faster, becoming smaller, more capable and efficient, and now able to run directly on device," Amon said during an earnings call earlier this month.

Cutting AI expenses

DeepSeek's breakthrough is particularly appealing to smaller enterprises looking to cut AI expenses or reduce reliance on OpenAI. High-priced, dominant models remain out of reach for many, while DeepSeek offers significantly lower pricing for application programing interface access than OpenAI or Anthropic.

AI search startup Perplexity quickly integrated DeepSeek's R1 model across its platforms for its cost-effectiveness. In an email to users, CEO Aravind Srinivas endorsed the technology: "You can experience the latest breakthrough in AI by turning on Pro Search with R1 ... I highly recommend you try it out today — the experience is truly remarkable."

DeepSeek's breakthrough represents a significant opportunity for corporate adoption by substantially reducing the cost of developing AI-powered products and services, Goldman Sachs executives said in a recent episode of the investment bank's podcast AI Exchanges.

DeepSeek's R1 model, released on Jan 20, has demonstrated capabilities comparable to OpenAI's closed-source GPT models in certain areas while achieving significantly lower training costs.

However, its rapid rise has sparked so-called security concerns among US officials and lawmakers.

Virginia has joined Texas and New York in banning DeepSeek from government devices over "security and safety" concerns. At the federal level, Congress has introduced legislation to ban DeepSeek on federal government devices, citing national security grounds.

The regulatory response has drawn comparisons to previous restrictions on TikTok and sparked pushback from some experts.

Eva Galperin, director of cybersecurity at the Electronic Frontier Foundation in California, questioned the approach in a recent article by StateScoop, suggesting a double standard in security considerations.

She pointed to the privacy concerns stemming from the US Department of Government Efficiency's unrestrained access to systems containing personal and financial information of millions of US citizens.

"I think that the trend of singling out Chinese products is particularly pernicious — especially while allowing practices that actively endanger US interests and the privacy and security of our very sensitive systems and data," Galperin said.

liazhu@chinadailyusa.com