Up 3x, Has AMD Cracked The Nvidia Code

Up 3x, Has AMD Cracked The Nvidia Code?

For most of the AI boom, Nvidia (NVDA) has been the only name that mattered in data center hardware. That dynamic is shifting. AMD (AMD) stock has risen 55% year-to-date and remains up 3.5x over the last 12 months, powered by record data center revenues and an increasingly competitive GPU lineup.

With hyperscalers actively seeking alternatives to single-vendor dependence, the structural case for AMD is stronger than the tenfold gap in market capitalization might suggest. Investors now see a credible second player in the AI gold rush - and the data backs it up. Data center revenue hit a record $5.4 billion over the last quarter, up 39% year-over-year, and strong Q1 guidance of $9.8 billion confirmed AMD as one of the key beneficiaries of the estimated $160 billion plus in hyperscaler capex for the quarter.

In fact, the company now expects data center revenue to improve by 60% year-over-year over a multi-year period.

Is This Structural Or Cyclical?

This appears like a structural gain.

AMD is successfully executing a two-front war: taking server CPU share from Intel while ramping its Instinct GPUs as a viable alternative to Nvidia. The GPU growth is particularly interesting

The hardware case for AMD is strong. Its Instinct MI325X offers up to 432GB of HBM4 memory - nearly double Nvidia's latest Rubin architecture. For inference workloads running massive models like Llama-4, that means fitting the entire model on a single chip rather than splitting it across several, cutting complexity and latency meaningfully. The economics reinforce the hardware story. With AMD GPUs, the cost per token for inference is significantly lower - which is why the likes of Meta and Microsoft are buying them in bulk. Cheaper inference directly improves the profitability of AI services at scale.

Finally, no hyperscaler wants a single-vendor dependency. Oracle and Meta integrate AMD partly as strategic leverage against Nvidia in price negotiations - making AMD a procurement necessity. Adding fuel to that structural case, Nvidia's once-impenetrable CUDA software moat is showing some cracks, as the industry is embracing open-source frameworks such as PyTorch and JAX.

So Is The Stock Priced For Perfection?

It appears so. The stock trades at around 130x trailing earnings; the multiple shrinks to 67x based on FY '26 consensus earnings. (See AMD valuation multiples) These are premium metrics, but not irrational given the secular AI tailwind. That said, several risks deserve close attention. Growth is projected at 35% for this year and almost 45% next year, marking an increase from last year. Operating margins also expanded to 17.1% last quarter.

Still, this valuation leaves little room for error, and there are some notable risks.

First, Nvidia remains formidable. Its CUDA ecosystem still dominates, despite the rise of open source. As the 'Rubin' architecture delivers a massive performance leap, it could slow AMD's momentum.

Second, hyperscalers are quietly moving loads away from GPUs. Google, Amazon, and Meta are shifting inference workloads - the high-volume, everyday AI "thinking" - onto custom chips like Trainium 3 and Maia 200, which can run up to 60% cheaper than third-party GPUs. AMD must prove its Instinct GPUs are more than a gap-filler for when Nvidia is out of stock, or it risks losing the fastest-growing part of the market to internal silicon. Marvell (MRVL) - a maker of interconnects and ASIC chips - stands to win big from this transition. Marvell Stock: The Good News Keeps Coming

Third, supply may be the quiet constraint. AMD relies on TSMC for leading-edge nodes, competing directly with Nvidia and Apple (AAPL) for scarce 3nm capacity. Any allocation imbalance or delay can disproportionately limit AMD's ability to convert demand into actual revenue.