SANTA CLARA — Advanced Micro Devices closed Friday’s session at $476.67, down 0.89%, a minor dip that tells a complicated story about where AMD sits in the AI chip race. The drop happened on the same day NVIDIA reported a blockbuster earnings quarter that sent its own shares surging, and the contrast captures something real: AMD is building genuinely competitive AI hardware, but the benchmark it’s being held to keeps moving.
The MI300X accelerator has become AMD’s strongest commercial argument in years. Data center revenue grew 107% year-over-year in the second quarter, reaching figures that would have seemed implausible eighteen months ago. Second-quarter revenue hit $11.5 billion, roughly double the year-earlier number, and AMD shares are up approximately 120% year-to-date, a run that puts it among the stronger performers in the semiconductor sector for 2026.
Those numbers are genuinely impressive. They are also the wrong numbers to measure AMD against. NVIDIA’s data center division brings in multiples of AMD’s GPU revenue every quarter, and the gap between H100/H200/B200 deployment volumes and MI300X installations remains significant across hyperscaler and enterprise workloads. AMD holds approximately 5-7% of the AI accelerator market by revenue, according to several analyst estimates, against NVIDIA’s approximately 75-81%. That’s a gap that 107% data center revenue growth won’t close in the next fiscal year.
Chief Executive Lisa Su has been consistent about this framing. AMD isn’t trying to displace NVIDIA across every AI workload. It’s building the case that a credible alternative is necessary for customers who don’t want to source all their AI compute from a single vendor. That pitch has found real buyers: Microsoft, Meta, and several large cloud providers have disclosed MI300X deployments this year. The question is whether those deployments scale from pilot to production at a pace that justifies AMD’s current valuation.
Friday’s trading played out against the backdrop of the Jackson Hole symposium, where Federal Reserve Chair Kevin Warsh delivered his first major address at the event since taking over from Jerome Powell in February. Warsh signaled no near-term rate cuts and reaffirmed the Fed’s commitment to its inflation mandate. The 10-year Treasury yield held around 4.679%. For AMD, which trades at a significant premium to traditional semiconductor multiples on AI-growth expectations, elevated rates compress the present value of future earnings. The stock’s resilience — a 0.89% decline on a day when rate signals were hawkish and NVIDIA drew all the sector attention — suggests the market has largely priced the macro headwind in.
Anyone tracking AMD without watching NVIDIA’s session today is reading half the story. The two companies are increasingly benchmarked against each other by the same enterprise procurement teams, and NVIDIA’s strong quarter resets the performance expectations AMD needs to meet.
The 52-week range runs from $149.22 to $584.73. At $476.67, AMD trades in the upper middle of that band, nowhere near the year’s low but comfortably below its peak. The stock’s trajectory in the second half of 2026 depends primarily on how quickly the MI300X pipeline converts to revenue at hyperscale, and whether AMD’s software stack — ROCm, the company’s answer to NVIDIA’s CUDA — has matured enough to reduce friction for developers. CUDA’s ecosystem advantage remains AMD’s hardest obstacle, and it’s not a hardware problem.
What AMD has demonstrated is that the duopoly framing — NVIDIA and everyone else — is becoming less accurate than it was in 2024. The MI300X has shown up in real production environments. The data center revenue growth rate is a reported result, not a projection. The argument that AMD is a viable second source in AI compute has moved from credible pitch to demonstrated fact over the past twelve months.
What hasn’t been demonstrated yet is the conversion rate from initial MI300X deployments to the kind of sticky, expanding commitments that drive multi-year revenue compounding. NVIDIA’s remaining performance obligations give investors forward visibility that AMD’s equivalent disclosures have not matched, which is why analyst price targets cluster more tightly around current levels. Morgan Stanley’s semiconductor team held a $520 target heading into Friday; Barclays sat closer to $500. Neither implies dramatic upside, and neither implies the AI thesis has collapsed.
On the broader Nasdaq today, the index fell 0.5% as Warsh’s remarks reduced rate-cut expectations across the board. AMD’s mild underperformance relative to that baseline reflects how closely it correlates with sector risk appetite on days when no company-specific catalyst is present.
The genuinely open question heading into AMD’s next earnings report is what happens to MI300X market share over the next eighteen months. If ROCm compatibility improves and hyperscaler deployments expand, the 5-7% market share estimate is a floor. If NVIDIA’s next-generation architecture maintains its performance advantage and the software gap persists, AMD’s data center growth decelerates from its current rate. Friday’s session didn’t resolve that question. The next earnings report, covering MI300X shipment volumes and updated data center guidance, will tell investors more than any single trading session can.
Related Posts
