Lisa Su has been playing chess while the rest of the chip industry was playing checkers. That much became clear at AMD's first-ever Financial Analyst Day, where the company laid out a strategy that goes way beyond just trying to nip at Nvidia's heels in the AI chip race.
The headliner? A staggering $45 billion in custom chip design revenue projected to start rolling in by 2026. Not million. Billion. With a B.
I've been covering semiconductor battles for years, and this kind of declaration isn't just ambitious—it's a gauntlet thrown directly at Jensen Huang's feet. Though Su would never be so crass as to say it outright.
Look, the semiconductor industry has always been cyclical. Companies rise and fall with architectural advantages that inevitably become obsolete. Remember AMD's brief moment in the sun during the mid-2000s with their Athlon 64 chips? I do. I was building PCs back then, swapping out Intel processors for AMD's superior offerings. Then Intel countered, AMD stumbled, and they nearly went bankrupt before the Zen architecture brought them roaring back.
This time feels different, though.
What struck me most wasn't actually the headline numbers (though they're eye-popping enough). It was AMD's commitment to annual product launches instead of their traditional biennial cycle. In chip design, cadence is everything. Absolutely everything. You can have the best architecture on paper, but if you're six months late to market? Might as well be six years.
The company is targeting some seriously ambitious market share numbers—double-digit AI data center share in 3-5 years, 50% server share, 40% PC share, and 70% adaptive computing share. For context, Nvidia currently controls something like 80-90% of the AI chip market.
(Sidenote: I spoke with several industry analysts after the presentation who confirmed these targets are... aggressive, to put it mildly.)
What's particularly fascinating is AMD's diversification strategy. While everyone else is fixated solely on the data center AI gold rush—which, let's be honest, is indeed growing at a mind-bending 80% compound annual growth rate—Su is positioning AMD across multiple sectors: aerospace, defense, automotive, communications.
This reminds me of AMD's approach during their first comeback. Start with price-sensitive customers, build credibility, improve your products, then expand upmarket.
The OpenAI mention wasn't accidental either. In today's landscape, getting designed into foundation models like GPT is the semiconductor equivalent of landing your chips in the iPhone. It's the ultimate validation. If AMD is pursuing more deals of this caliber, they clearly believe their MI300 and future chips can compete on performance, not just price.
But here's where we need a reality check.
The semiconductor graveyard is filled with ambitious targets that never materialized. Intel's foundry services anyone? IBM's semiconductor dreams? Execution matters more than announcements, and Nvidia's advantage isn't just hardware—it's their CUDA software ecosystem that's proven remarkably difficult to displace.
AMD's ROCm platform needs to achieve true compatibility and performance parity for these market share goals to be anything more than PowerPoint promises. The software stack in AI isn't a nice-to-have; it's everything.
Yet... if there's one executive in semiconductors who's earned the benefit of the doubt, it's probably Lisa Su. She took AMD from the brink of irrelevance to genuine contender. When she says they can deliver on a roadmap, history suggests you should probably believe her.
In semiconductors, standing still means death. AMD's accelerated product cadence suggests they understand this viscerally. Whether they can execute against Nvidia—a company currently firing on all cylinders under Huang's leather-jacketed leadership—remains the trillion-dollar question.
But I wouldn't count them out. Not by a long shot.
After all, in this industry, today's underdog is sometimes tomorrow's top dog. AMD has made that journey once before. Who's to say they can't do it again?
