Intel Vision In his Intel Vision Keynote on Tuesday CEO Pat Gelsinger outlined a scenario in which AI will eventually automate entire offices – or potentially even whole businesses.
He described this as the age of AI functions, in which agents – think highly tuned, application-specific models – start to interact with other agents to complete tasks. “Literally entire departments become AI automated solutions,” he envisioned. “Maybe we’ll have the age of the first one-person, billion-dollar company.” In StartupLand parlance, that’s a “Unicorn”.
Naturally, Gelsinger would prefer if the hardware used to train and run these AI agents and functions were created, or at the very least manufactured, by Intel. Throughout his roughly hundred-minute keynote address, he touched on everything from AI PCs capable of running models at home and Xeon 6 servers capable of running modestly sized chatbots like LLama2-70B, to the Habana team’s flagship Gaudi3 accelerator. The latter, Intel contends, is capable of going toe-to-toe with Nvidia’s H100.
Gelsinger is counting on widespread adoption of AI infrastructure and services to fuel Intel’s future growth. “I simply call it: every company becomes an AI company and this becomes the driving force of this expanding semiconductor TAM [total addressable market] from about $600 billion to more than $1 trillion by the end of the decade,” he predicted.
And – as Gelsinger later pointed out – to sell its hardware and claim a meaningful share of the AI infrastructure market, Intel is going to have to deliver tangible business outcomes.
According to Accenture’s chief AI officer Lan Guan, one of the technology’s biggest obstacles is “ambiguous value realization.”
This shouldn’t come as much of a surprise. The AI accelerators used to train and run large language models – like those underpinning Microsoft’s Copilot and Google’s Gemini – are incredibly expensive.
“Many clients are telling me it is really hard to realize value from their AI investments,” Guan told Gelsinger on stage.
Because of this, the idea of replacing whole swaths of knowledge workers with AI is no doubt a tantalizing prospect for enterprises – and a terrifying one for workers.
Intel is well aware of this. The x86 giant has already committed to training 30 million people to work with AI – either in their current role, or a new one – by 2030. And last week, the chip biz joined other tech titans to form a consortium to determine which jobs are most likely to be initially eliminated by AI and which workers can be retrained.
The good news, according to Gelsinger, is these changes aren’t going to happen overnight. It seems there’s still a lot of work to be done before AI functions come for our jobs.
“We see this enterprise AI transformation happening right around us today, but we see it unfolding in three different stages,” he opined.
The first he described as the “age of the AI copilot,” in which smart assistants like those from Microsoft, Google, and others help to boost worker productivity. “We expect productivity improvements of maybe 25 percent from those kinds of capabilities,” Gelsinger observed.
The next step, he predicted, will see the rise of AI agents which can handle more sophisticated tasks and automate entire workflows – like generating business reports. It will only be at that point when we can start stringing them together into AI functions, he concluded. ®