AMD and NVIDIA both posted explosive AI quarters, but they are betting on opposite visions of how the AI economy will actually be won. One of those visions fits the next phase of the buildout far better than the market\u2026<\/p>\n
AMD (NASDAQ: AMD<\/a> | AMD Price Prediction<\/a>) and NVIDIA (NASDAQ: NVDA<\/a>) both posted blockbuster AI quarters this summer. AMD\u2019s Data Center business more than doubled. NVIDIA\u2019s did too, but on a base almost ten times larger. Both are chasing the same explosion in AI compute demand<\/a> from opposite ends of the market, and their earnings reveal very different playbooks for how to win it.<\/p>\n Data Center Doubles Lift Both, But Only One Story Is Getting Told<\/p>\n AMD\u2019s Q2 FY2026 revenue hit $11.54B, up 50.1% year over year, with Data Center at $6.72B and up 107%. EPYC gained x86 server share<\/a>, and Lisa Su said \u201cData Center revenue more than doubled year-over-year.\u201d The Helios rack system is shipping, with Anthropic committing to up to 2 GW of MI450 Series GPUs and Microsoft deploying Helios on Azure.<\/p>\n NVIDIA is operating on a different plane. Revenue landed at $96.22B, up 105.8% year over year, and Data Center reached $89.02B. Vera Rubin is in full production, with Jensen Huang calling it \u201cthe fastest product ramp in NVIDIA\u2019s history.\u201d<\/p>\n Metric Revenue Data Center YoY Non-GAAP Gross Margin Next Quarter Guide Platform Empire Versus Open Coalition Challenger<\/p>\n NVIDIA is selling a closed AI factory. Huang framed it plainly: \u201cNVIDIA is a platform, an entire AI factory platform that spans the entire AI lifecycle that you can use in any cloud.\u201d The company is even layering revenue-sharing structures onto neocloud infrastructure to capture rental economics on top of hardware sales.<\/p>\n AMD is playing coalition builder. Helios pairs EPYC Venice, MI450 GPUs, and Pensando networking with open ROCm software, and management claims it delivers \u201cUp to 30% more tokens per dollar than the competition.\u201d That total cost of ownership pitch, combined with 50% more HBM per Helios rack, is exactly the argument the inference-heavy phase of the AI economy rewards.<\/p>\n Lens Core Bet Key Risk Inference Economics Will Decide Who Gains Share<\/p>\n I will be watching Helios shipments through Q4 and Anthropic\u2019s first gigawatt beginning in the first half of 2027. For NVIDIA, the tell is whether Vera Rubin actually reaches about 20% of data center revenue in Q3 and whether that $40 billion per gigawatt economic footprint holds as competitors close in.<\/p>\n Why AMD Looks Like the Sneakier AI Trade Right Now<\/p>\n You can see the market is starting to notice. AMD is up 123% year to date, while NVIDIA has added 23.67%. If you want the safest AI franchise on the planet, NVIDIA still owns that title, and its 75% gross margin is hard to argue with. But I lean toward AMD for the next leg. The customer roster now includes Anthropic, Meta, OpenAI, and Microsoft, and the open-standard, memory-rich pitch fits where inference is heading (the same data-center buildout keeps pulling in the power, cooling, and networking names we rounded up in a free AI infrastructure report<\/a>). I would change my mind quickly if Helios yields slip or HBM allocation tightens.<\/p>\n
\nAMD Q2 FY26
\nNVIDIA Q2 FY27<\/p>\n
\n$11.54B
\n$96.22B<\/p>\n
\n+107%
\n+117%<\/p>\n
\n56%
\n75.0%<\/p>\n
\n~$13B
\n$108B<\/p>\n
\nAMD
\nNVIDIA<\/p>\n
\nOpen, memory-rich TCO
\nFull-stack lock-in<\/p>\n
\nChina export controls<\/a>, Gaming down 31%
\nSupply-constrained through FY28<\/p>\n