How Chip Giant Intel Rejected OpenAI and fell behind

For U.S. chip giant Intel, the darling of the computing age, before the AI era hit rock bottom, things could have been much different.

About seven years ago, the company had an opportunity to buy a stake in OpenAI, a recently founded nonprofit research organization working in a little-known field called generative artificial intelligence, four people with direct knowledge of the discussions told Reuters.

In the months of 2017 and 2018, executives from both companies discussed options, including Intel buying a 15 percent stake for $1 billion in cash, three of the people said. They also discussed Intel taking an additional 15 percent stake in OpenAI if it produced hardware for the startup at cost, two of the people said.

Intel ultimately decided against the deal, in part because then-CEO Bob Swan didn’t think its generative AI models would be commercially viable in the near future and thus recoup the chipmaker’s investment. OpenAI is interested in investing in Intel because it would reduce its reliance on Nvidia chips and allow the startup to build its own infrastructure, two of the sources said. The deal also fell through because Intel’s data center unit didn’t want to produce products at cost, the sources added.

An Intel spokesperson did not respond to a request for comment, and OpenAI declined to comment.

Intel’s decision not to invest in OpenAI, which launched its groundbreaking ChatGPT in 2022 and is now reportedly valued at around $80 billion, has not been publicly disclosed.

It’s a series of strategic misfortunes that have hampered a company that led the way in computer chips in the 1990s and 2000s in the AI era, according to Reuters interviews with nine people familiar with the matter, including former Intel executives and industry experts.

Last week, Intel’s second-quarter profits sent its stock price down more than a quarter of its value, its worst trading day since 1974.

For the first time in 30 years, the tech company is valued below $100 billion. The former market king, whose marketing slogan “Intel Inside” has long represented the gold standard for quality, is still struggling to bring the most powerful AI chips to market.

Intel has now been overtaken by $2.6 trillion rival Nvidia, which has shifted from graphics for video games to AI chips needed to build, train and run large-scale AI systems, such as OpenAI’s GPT4 and Meta Platforms’ Llama models. Intel also trails AMD’s $218 billion.

Asked about AI advancements, an Intel spokesperson referred to recent comments by CEO Pat Gelsinger, who said the company’s third-generation Gaudi AI chip, targeted for launch in the third quarter of this year, would outperform competitors.

Gelsinger said the company has “more than 20 customers” for its second- and third-generation Gaudi AI chips, and that the next-generation Falcon Shores AI chip will launch in late 2025.

The spokesperson told Reuters: “We are nearing the end of a historic milestone in our technology, design and process innovation, and we are encouraged by the product portfolio we are building to capture a greater share of the AI market in the future.”

Game chips dominate AI market

In the OpenAI space, Microsoft invested in it in 2019, pushing itself to become a leader in the AI era, fueled by the 2022 launch of ChatGPT and a flurry of activity among the world’s largest companies in AI adoption.

In retrospect, though, the deal is a missed opportunity for Intel, as the company has been losing the battle for AI supremacy for more than a decade. As former executives and industry experts have said,

"Intel failed in AI because it didn't deliver a coherent product strategy to customers," said Dylan Patel, founder of semiconductor research group SemiAnalysis.

For more than two decades, Intel believed that CPUs, or central processing units, like the units that power desktop and laptop computers, could more efficiently handle the computational tasks needed to build and run AI models, according to four former Intel executives with direct knowledge of the company's plans.

Intel engineers initially viewed the graphics processing unit (GPU) architectures for video games used by competitors Nvidia and Advanced Micro Devices as "not pretty," one of the people said.

However, in the mid-2000s, researchers discovered that gaming chips were much more efficient than CPUs at handling the massive amounts of data needed to build and train large AI models. Because GPUs were designed for gaming graphics, they could do a lot of calculations at once.

Nvidia engineers have spent years since then refining GPU architectures to optimize them for AI applications and building the necessary software to exploit those capabilities.

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

NERVANA AND HABANA

Since 2010, Intel has made at least four attempts to produce efficient AI chips, including two startup acquisitions and at least two large home-grown efforts. None of them have managed to hurt Nvidia or AMD in this rapidly expanding and lucrative market, according to three people with direct knowledge of the company’s internal activities.

Intel’s entire data center business is expected to generate $13.89 billion in sales this year. That includes the company’s AI chips, but also many other designs, while analysts expect Nvidia to generate $105.9 billion in data center revenue.

In 2016, Intel CEO Brian Krzanich tried to buy his way into the AI business, buying Nervana Systems for $408 million. Intel executives were intrigued by Nervana’s technology, which was similar to the tensor processing unit (TPU) chips made by Google, according to two former executives.

The TPU, which is designed specifically for building or training large-scale generative AI models, strips away features of traditional GPUs useful for video games and focuses solely on optimizing AI computations.

Nervana has had some success with customers, including Meta Platforms for its processors, though not enough to prevent Intel from switching manufacturers and abandoning the project.

In 2019, Intel bought a second chip startup, Habana Labs, for $2 billion before shutting down its Nervana efforts in 2020.

https://github.com/rhysd/dot-github/issues/7

https://github.com/reprograma/github/issues/4

https://github.com/Semantic-Org/example-github/issues/7

https://github.com/jpardogo/GoogleProgressBar/issues/30

https://github.com/LeadCoding/3-weeks-Google-Prep/issues/5

https://github.com/darcyliu/google-styleguide/issues/15

https://pastelink.net/z4k0l6ay

https://linksome.me/awdrsfcsce

https://open.firstory.me/user/clzkmz4en02yl01tk1hjgd4gw

https://wokwi.com/projects/405614618890522625

https://herbalmeds-forum.biolife.com.my/d/114077-how-chip-giant-intel-rejected-openai-and-fell-behind