Intel’s decision to bypass collaboration with OpenAI has led to its current lag in technological advancements and industry competitiveness
Things could have been significantly different for Intel, the U.S. chip behemoth that was the darling of the computer age before it encountered difficulties in the AI era.
Four individuals with direct knowledge of the discussions told Reuters the company could acquire a stake in OpenAI. This fledgling non-profit research organization worked in a little-known field called generative artificial intelligence approximately seven years ago.
According to three individuals, executives at the two companies deliberated on various alternatives for several months in 2017 and 2018, including Intel’s acquisition of a 15% stake for $1 billion in cash.
Additionally, two sources deliberated on the possibility of Intel acquiring an additional 15% stake in OpenAI if it produced hardware for the startup at cost.
Ultimately, Intel (INTC.O) decided against a deal, in part because then-CEO Bob Swan did not believe that generative AI models would make it to the market shortly, thereby repaying the chipmaker’s investment, according to three of the sources, who all requested anonymity to discuss confidential matters. Opens a new tab.
Two individuals stated that OpenAI was interested in investment from Intel because it would reduce their dependence on Nvidia’s processors and enable the startup to establish its infrastructure.
The individuals added that the vidualsvisualsthat the transaction was also unsuccessful due to Intel’s data center unit’s refusal to manufacture products at cost.
An Intel spokesperson did not address questions regarding the prospective transaction. Swan did not respond to a request for comment, and OpenAI declined to answer.
Intel has not previously disclosed its decision to refrain from investing in OpenAI, which subsequently introduced the innovative ChatGPT in 2022 and is currently estimated to be worth approximately $80 billion.
According to Reuters interviews with nine individuals 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 earnings resulted in its worst trading day since 1974, during which the stock price declined by over 25%.
The technology company’s valuation has fallen below $100 billion for the first time in three decades.
The former market leader, whose marketing slogan “Intel Inside” long symbolized the highest quality, still struggles to introduce a groundbreaking AI processor product.
Intel is currently dwarfed by its $2.6 trillion competitor, Nvidia (NVDA.O), which has transitioned from video game graphics to AI processors required for the construction, training, and operation of large generative AI systems such as OpenAI’s GPT4 and Meta Platforms’ (META.O). Intel has also lagged behind AMD (AMD.O), valued at $218 billion.
When asked about Intel’s AI advancements, the spokesperson cited recent statements by CEO Pat Gelsinger, who stated that the company’s third-generation Gaudi AI processor, which will be released in the third quarter of this year, would surpass its competitors.
According to Gelsinger, the company had “20-plus” consumers for Gaudi’s second and third-generation generations and anticipated the release of its next-generation Falcon Shores AI chip 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,” the spokesperson stated to Reuters.
AI SWEEPS GAMING CHIPS
In 2019, Microsoft (MSFT.O) invested in OpenAI, propelling itself to the vanguard of the AI era. This move was prompted by the 2022 release of ChatGPT and a flurry of activity among the world’s largest companies to deploy AI.
Intel has been gradually losing the battle for AI supremacy for over a decade, according to the former executives and industry experts interviewed, even though the prospective transaction was a missed opportunity in hindsight.
Dylan Patel, the founder of SemiAnalysis, a semiconductor research group, stated that Intel’s failure in AI was due to its inability to present a cohesive product strategy to its customers.
Four former Intel executives with direct knowledge of the company’s plans have stated that Intel believed the CPU, or central processing unit, which powers desktop and laptop computers, could more effectively handle the processing tasks necessary to build and run AI models for over two decades.
One of the individuals stated that Intel engineers perceived the graphics processing unit (GPU) video gaming processor architecture, which is employed by rivals Nvidia and Advanced Micro Devices, as relatively “ugly.”
However, by the mid-2000s, researchers had discovered that gaming processors were significantly more efficient than CPUs in managing the extensive data processing required to construct and train large AI models. GPUs can perform an immense number of calculations in parallel due to their design for game graphics.
Since then, Nvidia’s engineers have dedicated years to modifying the GPU architecture to optimize it for AI applications and developing the software required to leverage its capabilities.
According to Lou Miscioscia, an analyst at Daiwa, Intel could not provide the appropriate processor when AI was introduced.
HABANA AND NERVANA
Intel has made at least four attempts to develop a viable AI chip since 2010, including acquiring two startups and developing at least two significant indigenous initiatives.
According to three individuals with firsthand knowledge of the company’s internal operations, none have significantly impacted Nvidia or AMD in the rapidly expanding and lucrative market.
This year, Intel’s total data center business is anticipated to generate sales of $13.89 billion, which includes the company’s AI chips and numerous other designs. Meanwhile, analysts expect that Nvidia will create a data center revenue of $105.9 billion.
In 2016, Intel CEO Brian Krzanich attempted to acquire Nervana Systems for $408 million to enter the AI industry through acquisition. According to two former executives, Intel executives were drawn to Nervana’s technology, which is comparable to a tensor processing unit (TPU) processor manufactured by Google.
The TPU, specifically engineered to construct or train large generative AI models, eliminated the features of a traditional GPU that were beneficial for video games and instead concentrated on optimizing AI calculations.
Nervana achieved some success with clients, such as Meta Platforms, for its processor; however, this was insufficient to prevent Intel from abandoning the project and switching horses.
2019, Intel acquired Habana Labs, a second semiconductor startup, for $2 billion. Nervana’s operations were terminated in 2020.
Krzanich did not respond to a request for comment regarding this article.