Want to Know if AMD is Gaining on Nvidia? Look Here.
AMD just posted its best data center growth in years while Nvidia guided margins lower, and one number buried in both earnings calls reveals whether the AI chip race is actually tightening or if the gap is as wide as…
This post may contain links from our sponsors and affiliates, and Flywheel Publishing may receive compensation for actions taken through them.
Advanced Micro Devices (NASDAQ:AMD | AMD Price Prediction) and NVIDIA (NASDAQ:NVDA) both reported blockbuster quarters this August, and each result speaks directly to the AI accelerator race. Lisa Su’s team posted $11.54 billion in revenue with Data Center up 107%. Jensen Huang countered with $96.22 billion and a $108.0 billion forward guide. The gap is still enormous, but the trajectory is what matters.
Gross Margins Tell the Real Story on Pricing Power
Gross margin is where a chipmaker’s pricing power lives. If AMD were truly eroding NVIDIA’s grip on AI accelerators, you would expect NVIDIA’s margin to buckle while AMD’s climbs. Right now, both are moving, but not in the way bulls on either side might frame it.
NVIDIA posted a non-GAAP gross margin of 75.0%, then guided Q3 to 74%, plus or minus 50 basis points, with a Q4 trough in the 71% to 72% range before settling at 72% to 73% for fiscal 2028. CFO commentary blamed “extreme pricing conditions in memory“. That distinction matters.
AMD, meanwhile, expanded gross margin to 56%, up over 200 basis points year-over-year, and guided the same 56% into Q3. Jean Hu also flagged that Instinct margins run “slightly below corporate average”. Translation: as MI350 and Helios scale, AMD’s mix gets richer in revenue but not necessarily richer in margin.
| Metric | AMD Q2 | NVIDIA Q2 |
| Non-GAAP gross margin | 56% | 75.0% |
| Data Center revenue | $6.72B | $89.02B |
| Next-quarter guide | ~$13B | $108.0B |
Helios and Anthropic vs. Vera Rubin and CUDA
AMD’s competitive proof points are real. Helios claims “up to 15% more throughput at the same rack power, and up to 30% more tokens per dollar than the competition”, and Anthropic committed to up to two gigawatts of MI450 series GPUs. Microsoft is deploying Helios on Azure for frontier inferencing. That is genuine share of wallet.
Huang’s answer is scale and stickiness. Vera Rubin is now in full production, revenue per gigawatt has climbed from roughly $18 billion on Hopper to about $40 billion on Vera Rubin, and NVIDIA is fulfilling only about 70% of demand. When supply is the bottleneck, you keep pricing.
What I Am Watching Into 2027
The signal I care about is whether NVIDIA’s actual fiscal 2028 margin lands closer to 73% or slips toward 70%. If it settles in the guided 72% to 73% band after price increases take effect, memory is the villain and AMD is not really eroding NVIDIA’s pricing power. If it undershoots, that is the first hard evidence AMD’s tokens per dollar pitch is forcing concessions.
Why I Still Lean NVIDIA on This Quarter
Personally, I lean NVIDIA here, and it is largely a valuation and durability call. AMD shares are up 113.42% year to date against NVIDIA’s 20.47%. A lot of AMD’s future is already priced in. If you are a turnaround-and-momentum investor, AMD’s Helios ramp is the more exciting story. If you want the compounder with the widest moat and the cleanest margin structure, NVIDIA still screens as the more durable margin story until the gross margin gap actually narrows (and if you would rather sidestep the chipmaker debate entirely, the power, cooling, and networking suppliers feeding this buildout are their own trade, which we covered in a free report on seven AI infrastructure names that aren’t chipmakers).
Contact [email protected] for any questions or corrections.







