Intel is redefining its AI strategy

Over the past year, under new CEO Pat Gelsinger, Intel has continued to expand its chip lineup and increase its workforce, rolling out new artificial intelligence software to improve AI-powered smart chatbots, Applications such as facial recognition and movie recommendations.

Intel is known for its CPU, x86 architecture and its dominance of the market for servers that run corporate networks and the Internet. But over the past decade, Intel has lost some of its appeal to investors as Nvidia has gradually taken over the market for chips designed for artificial intelligence, especially those for training AI models.

Nvidia now accounts for about 80 percent of AI-specific computing revenue in big data centers, according to the Omdia division of U.K. research and consulting firm Informa PLC, though that doesn't include AI computing on Intel's general-purpose CPUs. Two years ago, Nvidia surpassed Intel in AI chip dominance as the most valuable U.S. chip company.

AI chips are a relatively small but fast-growing segment of the overall chip market. The growing demand for faster and more efficient AI computing has spawned dozens of chip startups, while leading chipmakers are investing heavily. The artificial intelligence chip market was worth about $8 billion in 2020, but is expected to grow to nearly $200 billion by 2030, according to a report from Portland, Oregon-based Allied Market Research.

Intel's strategy is to build stable chips and open-source software to meet the broad computing needs of artificial intelligence becoming more commonplace. For example, it could sell customers a software package that allows them to offload some tasks to specialized chips that excel at specialized chips like image recognition, while handling other work on general-purpose chips.

Intel hopes the efficiency of this division of labor will help the company optimize the performance of its specific AI tasks and save money by reducing power consumption. This may make sense for clients (large corporations and well-funded startups) with lots of data and doing a lot of AI processing.

An important change Intel has made to implement this strategy is the addition of graphics processing units to its product line. The chips, introduced more than two years ago, could help it better against Nvidia, which focuses on GPUs originally developed for computer games but suitable for machine learning tasks. Intel acquired Israeli startup Habana in 2019, which makes chips designed for training AI models and generating output from those models.

Another change is the way Intel integrates its AI products for customers.

“It’s not even a question of whether we need more investment — we’re investing quite a bit in AI,” said Sandra Rivera, a longtime Intel executive who Gelsinger hired last summer. Leading data center business and AI strategy. "But we haven't gotten the clout of those investments yet when we have different strategies and different execution priorities for various products".

Since taking on her new role, Sandra Rivera has brought in several new executives, including Kavitha Prasad, from a machine learning startup she co-founded after her earlier stint at Intel. Ms Prasad, who directly manages the AI ​​strategy, said the focus is now on using AI to achieve clients' business goals, rather than providing a menu of chips and letting clients figure out the rest.

"Intel has all of these technologies, but from a customer perspective, what brings them together so that it's cohesive so that customers can deploy it faster so they can achieve business outcomes faster?" she Say. "It's not about finding a solution, it's about bringing them together meaningfully to make it happen."

Bringing it all together is largely the work of Intel's software architects, led by CTO Greg Lavender, whom Gelsinger brought in from VMware, where Gelsinger was CEO.

Of course, success is not a sure thing. Nvidia, already well ahead of Intel and other rivals, is rapidly rolling out its own chips, announcing a new generation of superfast processors in March. While analysts say Intel's strategy could help make it a stronger competitor to Nvidia, its ability to tap the AI ​​market depends on delivering AI-targeted chips and related software on time. But that could be a challenge, as Intel has faltered in chip-making technology in recent years, trailing South Korea's Samsung Electronics Co and Taiwan Semiconductor Manufacturing Co. In the high-stakes race to make chips with the smallest transistors and the best performance. Some of its latest CPU chips for servers have been delayed.

Gelsinger, whose goal is to reverse that trajectory and put the company back into manufacturing, has announced that he will spend tens of billions of dollars over the next few years on new chip factories and build a business that makes chips based on other people's designs. It's an open question whether the company can execute on Gelsinger's plan to take back the technology leadership from its Asian rivals in the coming years.

"There's a lot of things they haven't executed over the last five or six years," said Matt Bryson, an analyst at Wedbush Securities. "Obviously Intel is investing more in product development under Pat Gelsinger and if you invest more in development you should have better execution, but after you show the product and start to see traction Before, it came back to the question of how do you know?"

Sandra Rivera said Intel is ready to make that leap: "We have the customer relationships, we have the market position, we have the unique differentiation -- we just have to execute our strategy."