Last month, Wall Street was stunned by the emergence of a Chinese tech startup called DeepSeek, which released an artificial intelligence chatbot. It became the most downloaded free app on Apple’s iPhone store.
According to reporting by The Associated Press, industry analysts said the app vaulted above American companies’ generative AI — machine learning that creates such content as music, images and text — at a much lower cost, especially compared with U.S. leader OpenAI, headquartered in San Francisco.
The breakthrough is highly consequential for Seattle-area technology firms, which employ tens of thousands and are worldwide talent magnets.
Neither Microsoft nor Amazon panicked over DeepSeek.
Microsoft and OpenAI, the Redmond company’s partner and beneficiary as largest investor, are exploring whether data from OpenAI’s technology was improperly obtained by a group connected to DeepSeek, according to Bloomberg. (Elon Musk is attempting a hostile takeover of OpenAI, although it’s unlikely to succeed.)
Typically, software developers can reimburse OpenAI for a license to use its proprietary artificial intelligence.
Microsoft declined to comment, although the company said DeepSeek’s open-source R1, which rattled American investors and tech experts, is available on its Azure A1 Foundry — a product for AI apps.
Microsoft said, “We are committed to enabling customers to build production-ready AI applications quickly while maintaining the highest levels of safety and security. DeepSeek R1 has undergone rigorous safety evaluations.”
Microsoft CEO Satya Nadella told Wall Street analysts last month that DeepSeek “has some real innovations.”
He added, “I talked about how, in some sense, what’s happening with AI is no different than what was happening with the regular compute cycle.” In other words, the processing time for artificial intelligence.
Nadella said: “It’s always about bending the curve and then putting more points up the curve. There’s Moore’s Law that’s working and hyperdrive.”
Moore’s Law, named after one of Intel’s founders Gordon Moore, posits that the number of transistors on an integrated circuit will double every two years with minimal increase in cost, something applied more broadly in technology.
Both Microsoft and Amazon are investing heavily in AI.
Amazon Web Services says its customers can use DeepSeek R1, “a powerful and cost-effective AI model that excels at complex reasoning tasks,” with some of its products.
R1 models can be used on Amazon’s Bedrock AI, one of the company’s products to build AI applications, and SageMaker AI, as well as AWS Trainium, Inferentia chips and Amazon Elastic Compute Cloud.
Amazon emphasizes that using models through AWS offers security.
In plain English, that means Amazon Bedrock is a fully managed service for building and scaling generative artificial intelligence applications with high-performing models.
Meanwhile, Amazon SageMaker AI is an end-to-end service — managing a system or process from start to finish — used by hundreds of thousands of customers to help build, train and deploy AI models for any use case with fully managed infrastructure, tools and workflows.
On a recent earnings call with Wall Street analysts, Amazon CEO Andy Jassy said he was impressed by DeepSeek’s breakthrough.
“I think one of the interesting things over the last couple of weeks is, sometimes people make assumptions that if you’re able to decrease the cost of any type of technology component, in this case, we’re really talking about inference, that somehow it’s going to lead to less total spend in technology,” according a transcript of the call. “And we have never seen that to be the case.”
DeepSeek’s R1 raises the potential of an AI system that can equal American companies’ at much lower cost.
A 22-page paper released by the Chinese company says “DeepSeek-R1 achieves performance comparable to OpenAI … on a range of tasks.”
DeepSeek’s R1 is open source, meaning its code is freely available, compared with most artificial intelligence competitors.
New York Times tech reporter Kevin Roose wrote that he was doubtful that DeepSeek was a plot by Beijing to wreck the U.S. technology sector.
“But I do think that DeepSeek’s R1 breakthrough was real. Based on conversations I’ve had with industry insiders, and a week’s worth of experts poking around and testing the paper’s findings for themselves, it appears to be throwing into question several major assumptions the American tech industry has been making.”
Among those assumptions is spending large sums on data centers and powerful semiconductors when an upstart Chinese firm used a fraction of the cost to equal Western firms or leap ahead.
The company was founded 18 months ago by hedge fund billionaire Liang Wenfeng in Hangzhou, a city of nearly 12 million located near the East China Sea.
I’ve written before about how AI might replace the work of lawyers, doctors, accountants and journalists, among others.
A World Economic Forum survey found 41% of employers surveyed intended to reduce their workforces because of AI on certain jobs by 2030.
Yet historically, the acceleration and growth of technology and automation have led to more jobs across several industries. Buggy whip makers were thrown out of work by the rise of the automobile, but Detroit created far more jobs.
When DeepSeek made the news, my first thought was: This is how World War III might begin.
For all of Seattle’s peaceful pretensions, the region is surrounded by military installations, including the ballistic submarine base at Bangor, as well as Boeing. It makes us a first-strike target if a confrontation with China went nuclear.
According to the Carnegie Endowment for International Peace, many nations are racing to use AI for military uses. China is working to use the technology to leapfrog American advantages on a potential battlefield. The United States is employing AI to enhance its defense capabilities.
With AI, many human pilots might become obsolete. Warships will require smaller crews. And cyberwarfare will reach every tech user. I wrote an unfavorable column about China a year ago, and my MacBook Pro has been buggy ever since. Coincidence?
The worst might never happen, but if a conflict erupts, an artificial intelligence Pearl Harbor would be enabled by the technology embodied by DeepSeek and its competitors.
