OpenAI has been making it rain infrastructure deals. The AI startup just dropped another $38 billion on Amazon Web Services, adding to agreements with Oracle, Microsoft, AMD, and a plethora of other deals that total over $1 trillion this year so far. But there’s one company conspicuously missing from the invite list to this very expensive party. Intel, the former belle of the Silicon Valley ball, is now watching everyone else dance.
The cold shoulder is particularly awkward given that the Trump administration took a chunk of Intel earlier this year. For an administration that treats dealmaking like a competitive sport and sees making advanced chips at home a matter of national security, you’d think getting OpenAI and Intel together would be irresistible — a political slam dunk wrapped in an American flag.
Intel’s absence from OpenAI’s vendor list reflects more than just technical shortcomings. It’s a Silicon Valley soap opera of missed opportunities, bruised egos, and the harsh reality that in the AI arms race, second place is no place at all.
When OpenAI was founded in 2015, Intel was the undisputed emperor of silicon. The company commanded a $150 billion market cap, controlled at least 70% of the PC processor market, and its “Intel Inside” stickers were on virtually every computer on the planet. Nvidia? They made graphics cards for gamers, worth maybe $18 billion on a good day.
So when OpenAI came knocking in 2017, hat in hand, asking Intel to invest in their fledgling AI startup, Intel said no thanks. Why would the king of computing need to bet on some nonprofit’s sci-fi dreams?
Fast forward to today. Nvidia has exploded to a $4.5 trillion valuation, making it worth 40 times more than Intel. The company that once turned its nose up at OpenAI has cycled through three CEOs and watched the AI boom happen without them. Now, Intel is the one begging for a seat at the table while OpenAI doesn’t seem to be interested.
“Why go back to a firm that has an inferior product and also snubbed you?” said Tim Derdenger, an associate professor at Carnegie Mellon’s Tepper School of Business.
Intel’s cold shoulder looks especially foolish now that OpenAI is actively hunting for anyone — anyone but Intel, apparently — to break Nvidia’s 90% stranglehold on the AI chip market. The startup’s recent deal with AMD, worth tens of billions of dollars, follows a classic playbook from the PC era.
Derdenger pointed out that OpenAI’s strategy mirrors how Microsoft once managed Intel’s dominance by cultivating AMD as a second source. The approach allows OpenAI to credibly threaten to shift orders between suppliers rather than being held hostage by a single vendor.
But leveraging Intel as that second or even third source is another story entirely. An OpenAI-Intel deal “feels possible but still unlikely near term,” according to Brad Gastwirth, global head of research and market intelligence at Circular Technology. The fundamental problem? Performance.
“Intel’s lag in training-class GPU performance makes a significant partnership less probable today,” Gastwirth said. “OpenAI already has deeper relationships and more advanced hardware programs than what Intel currently offers.”
Intel’s Gaudi AI processors, while marketed for artificial intelligence, haven’t made a dent in the market dominated by Nvidia and AMD. The company’s new GPU, called Crescent Island, won’t ship until next year, and it’s already drawing skepticism for using slower memory than the high-bandwidth memory found in competitors’ chips.
Even when Crescent Island arrives, Intel faces a software challenge. Data centers aren’t just about raw silicon. They require software to run everything. Nvidia’s CUDA platform has become the de facto standard for AI development, creating switching costs that transcend simple price comparisons.
“When you build out a data center, it’s not just the chip. You have to have the software that works and communicates with the chip,” Derdenger said. “It’s the system effect that matters.”
The Trump administration’s stake in Intel creates an unusual dynamic rarely seen in American capitalism. The government owns 10% of Intel after converting $9 billion in federal grants into equity this summer, a move Commerce Secretary Howard Lutnick defended as getting “the benefit of the bargain” for taxpayers.
Trump’s approach to reshaping the chip industry goes beyond just owning a piece of Intel. In August, he announced plans for a 100% tariff on all semiconductors made outside the U.S., though companies that commit to American production can get exemptions. Intel, with its existing U.S. fabs and the government as its largest shareholder, should be perfectly positioned to benefit from this protectionist push. Yet OpenAI and other AI companies continue to source their chips from wherever the technology is best, which isn’t Intel.
Since then, OpenAI has hinted that it might need the feds’ help. OpenAI just asked the Trump administration to expand a 35% Chips Act tax credit to cover AI data centers and server producers, according to an October 27 letter to the White House. The startup has also advocated for government grants and loan guarantees to help the AI industry counter China, support that could be crucial as OpenAI pursues its massive infrastructure buildout.
This alignment should make an OpenAI-Intel partnership politically irresistible. Trump wants to dominate AI manufacturing domestically. OpenAI needs government support for infrastructure. Intel desperately needs AI customers. The government literally owns a piece of Intel. It’s the kind of arranged marriage that could benefit everyone.
Yet even with all this political matchmaking potential, Intel can’t seem to get an invite to the AI party. The path forward for Intel is clear, but could be a stress test on new hardware that isn’t ready for prime time. A “small-scale or exploratory engagement perhaps tied to custom accelerators or inference use cases remains plausible,” according to Gastwirth, but a headline deal requires Intel to “deliver at competitive performance and scale.”
The stakes couldn’t be higher. As OpenAI commits to building 30 gigawatts of computing resources, Intel watches from the bleachers as billions flow to its competitors. In the AI gold rush, Intel isn’t even selling shovels; it’s still trying to convince miners that its tools exist.
Leave a Comment
Your email address will not be published. Required fields are marked *