How Intel Missed The AI Wave By Turning Down OpenAI

Things could have been very different for US semiconductor giant Intel, which was the darling of the computer age before struggling in the AI era.


Four people with direct knowledge of the conversations told Reuters that about seven years ago, the corporation had the opportunity to purchase a stake in OpenAI, a budding non-profit research group working in a little-known field called generative artificial intelligence.

Over several months in 2017 and 2018, executives at the two businesses explored a variety of options, including Intel purchasing a 15% interest for $1 billion in cash, according to three sources. They also discussed Intel purchasing an extra 15% share in OpenAI if it produced hardware for the firm at cost, according to two sources.

According to three sources who asked anonymity to discuss secret matters, Intel finally opted against a merger, in part because then-CEO Bob Swan did not believe generative AI models would make it to market in the near future, repaying the chipmaker’s investment.

According to two sources, OpenAI was interested in an investment from Intel since it would have decreased the startup’s dependency on Nvidia chips and allowed it to construct its own infrastructure. The agreement also fell through because Intel’s data center unit refused to manufacture items at cost, according to the sources.

An Intel spokeswoman declined to comment on the proposed deal. Swan did not return a request for comment, and OpenAI declined to comment.

Intel’s decision not to invest in OpenAI, which went on to debut the revolutionary ChatGPT in 2022 and is now reputedly valued at around $80 billion, was not previously disclosed.

According to Reuters interviews with nine people familiar with the matter, including former Intel executives and industry experts, it is one of a series of strategic misfortunes that have caused the company, which was at the forefront of computer chips in the 1990s and 2000s, to stumble in the era of AI.

Last week, Intel’s second-quarter reports caused a stock price drop of more than a quarter of its worth on its worst trading day since 1974.

For the first time in 30 years, the tech company is valued below $100 billion. The former industry leader, whose marketing slogan “Intel Inside” once signified the gold standard of quality, is still striving to bring a blockbuster AI chip product to market.

Intel is now eclipsed by $2.6 trillion rival Nvidia, which has shifted its focus from video game graphics to AI chips required to construct, train, and run big generative AI systems such as OpenAI’s GPT4 and Meta Platforms’ Llama models. Intel has also dropped behind AMD, which is valued at $218 billion.

When asked about its AI efforts, the Intel spokesman cited recent comments by CEO Pat Gelsinger, who stated that the company’s third-generation Gaudi AI chip, which is set to launch in the third quarter of this year, will surpass competitors.

Gelsinger stated that the business has “20-plus” customers for the second and third generations of Gaudi, and that its next-generation Falcon Shores AI processor will be released in late 2025.

“We are nearing the completion of a historic pace of design and process technology innovation, and we are encouraged by the product pipeline we’re building to capture a greater share of the AI market going forward,” a representative shared with Reuters.

Gaming chips outperform AI.

On the OpenAI front, Microsoft made an investment in 2019, propelling itself to the forefront of the AI era, which will be sparked by the 2022 release of ChatGPT and a frenzy of activity among the world’s major enterprises to deploy AI.

Although the proposed transaction was a missed opportunity for Intel, the corporation has been progressively losing the war for AI supremacy for more than a decade, according to former executives and industry experts questioned.

“Intel failed in AI because they didn’t present a cohesive product strategy to their customers,” said Dylan Patel, founder of semiconductor research firm SemiAnalysis.

For more than two decades, Intel believed that the CPU, or central processing unit, used to power desktop and laptop computers, could better handle the processing tasks required to build and run AI models, according to four former Intel executives with direct knowledge of the company’s plans.

Intel developers thought the graphics processing unit (GPU) video gaming chip design employed by rivals Nvidia and Advanced Micro Devices was “ugly,” according to one of the sources.

By the mid-2000s, researchers had established that gaming chips were significantly more efficient than CPUs in handling the heavy data processing required to develop and train massive AI models. Because GPUs are built for game visuals, they can execute massive amounts of computation in parallel.

Since then, Nvidia developers have spent years refining the GPU architecture to optimize it for AI applications, as well as developing the software required to take advantage of the capabilities.

“When AI hit… Intel just didn’t have the right processor at the right time,” said Lou Miscioscia, an analyst at Japanese investment bank Daiwa.

Nervana y Habana

Since 2010, Intel has attempted at least four attempts to develop a viable AI processor, including the acquisition of two firms and at least two major in-house endeavors. None have made a dent in Nvidia or AMD’s quickly expanding and lucrative business, according to three people with firsthand knowledge of the company’s internal operations.

Intel’s whole data center business is predicted to generate sales of $13.89 billion this year, including the company’s AI chips as well as many other designs, while analysts expect Nvidia to earn data center revenue of $105.9 billion.

In 2016, Intel CEO Brian Krzanich attempted to enter the AI market by purchasing Nervana Systems for $408 million. According to two former Intel executives, Nervana’s technology was similar to Google’s tensor processing unit (TPU) hardware.

The TPU, which was specifically built for generating or training large generative AI models, removed the elements that make a traditional GPU effective for video games and concentrated solely on optimizing AI computations.

Nervana had some success with customers, like Meta Platforms, with its processor, but not enough to prevent Intel from abandoning the project.

Intel spent $2 billion in 2019 to acquire Habana Labs, a second chip firm, before discontinuing Nervana’s operations in 2020.