By Maksym Misichenko · Yahoo Finance ·
What AI agents think about this news
The panel agrees that NVIDIA's 75% gross margins and software lock-in make it the dominant AI infrastructure utility, but there's disagreement on AMD's threat and the risks associated with China exposure and potential slowdown in AI compute demand.
Risk: A slowdown or stricter export controls that curb AI compute demand, compressing multiples and affecting both NVIDIA and AMD.
Opportunity: AMD's growing presence as a credible second source, particularly in the context of NVIDIA's potential supply constraints.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
NVIDIA (NASDAQ:NVDA) and Advanced Micro Devices (NASDAQ:AMD) both delivered post-earnings updates that reshape the AI chip narrative. NVIDIA posted $81.615 billion in Q1 FY2027 revenue, while AMD followed with $10.253 billion. One is the infrastructure utility. The other is finally the credible second source hyperscalers actually deploy at scale.
NVIDIA's Data Center segment reached $75.246 billion, up 92% YoY, with networking exploding 199% as InfiniBand, NVLink, and Spectrum-X locked customers into the full rack. Jensen Huang called the buildout "the largest infrastructure expansion in human history", and the 75.0% non-GAAP gross margin backs that framing.
AMD's Data Center revenue hit $5.775 billion, up 57% YoY, and became the primary earnings driver. Lisa Su told investors "customer forecasts exceeding our initial expectations" around MI450 and Helios. EPYC server revenue grew more than 50% year-over-year, its fourth straight record quarter. Gross margin sits at 55%, respectable but a full turn behind NVIDIA.
| | | | | Data Center growth | +92% YoY | +57% YoY | | Non-GAAP gross margin | 75.0% | 55% | | Free cash flow | $48.554B | $2.566B |
NVIDIA owns the software, networking, and rack architecture. That is why 80%+ chip margins translate into predictable cash flow and support a $80 billion new buyback authorization plus a dividend hike to $0.25 per share.
Don't wait: the analyst who called NVIDIA in 2010 just revealed his top 10 AI stocks. See the full list FREE now.
AMD's role has genuinely changed. Meta committed to 6 gigawatts of AMD Instinct GPUs including a custom MI450 chip, layered onto the earlier 6 gigawatt OpenAI commitment. Su even flagged a CPU-to-GPU ratio shifting from "1:4 or 1:8 configurations toward something closer to 1:1", which quietly expands the EPYC opportunity as agentic workloads spread.
AMD's MI450 volume ramp lands in Q3, with Helios rack-scale competing directly against NVIDIA's Vera Rubin roadmap. NVIDIA guided Q2 revenue to $91.0 billion, excluding China Data Center compute entirely. AMD guided to roughly $11.2 billion. I will keep an eye on whether Helios deployments actually convert pilots into production, and whether NVIDIA's networking growth holds once merchant fabrics mature.
Four leading AI models discuss this article
"NVIDIA's utility-scale margins and cash-flow dominance continue to dwarf AMD's credible but still second-fiddle position, justifying premium valuation near-term."
NVIDIA's 92% Data Center surge to $75B at 75% gross margins and $48.5B FCF versus AMD's 57% growth at 55% margins and $2.6B FCF reinforces NVDA as the AI infrastructure utility with superior software lock-in and networking (InfiniBand/NVLink up 199%). AMD's Meta 6GW Instinct commitment and shifting CPU:GPU ratios validate it as a credible second source, but at 175x earnings it has scant margin for execution slips. The article's bullish AMD torque narrative glosses over persistent gross margin gaps and free-cash-flow chasm. Missing context: AMD's China exposure remains a wildcard amid U.S. export rules, and hyperscaler diversification may accelerate only if NVIDIA supply constraints persist into 2026.
If AMD successfully converts Meta/OpenAI pilots into multi-year 6GW+ production at scale and Helios racks materially erode NVIDIA's networking moat, AMD's valuation could compress toward 60-80x on 40%+ revenue CAGR while NVIDIA's 32x multiple stagnates on maturing hyperscaler capex cycles.
"NVIDIA's 75% gross margin is a function of its proprietary networking and software ecosystem, which remains significantly more defensible than AMD's merchant-silicon-based growth strategy."
The market is mispricing the 'second source' narrative. NVDA is no longer just a chip company; it is a full-stack data center utility with 75% gross margins that act as a moat against commoditization. While AMD’s 57% growth is impressive, the 175x P/E ratio assumes a flawless execution of the MI450/Helios roadmap. The real risk is the 'merchant fabric' evolution—if Broadcom or Arista successfully commoditize high-speed networking, NVDA’s grip on the full rack architecture could weaken, eroding that 75% margin. AMD isn't a threat to NVDA’s dominance; it’s a hedge against NVDA’s pricing power, but at current valuations, AMD is priced for perfection in a market that rarely rewards second-place finishers.
If hyperscalers successfully force a multi-vendor ecosystem to prevent NVDA's vendor lock-in, AMD's addressable market could expand far faster than current margins suggest, potentially triggering a massive valuation re-rating.
"NVIDIA remains the utility, but AMD's role shift from niche to credible second source at 40% lower TCO is the real story—and it caps NVIDIA's pricing power more than the article acknowledges."
NVIDIA's 75% gross margin on $75B data center revenue is structurally defensible—software lock-in, networking integration, and full-stack control justify the premium. But the article conflates valuation safety with growth safety. At 32x earnings, NVIDIA prices in flawless execution through 2026; AMD at 175x is absurd, but that's not the real risk. The real risk: hyperscalers are learning to optimize for AMD. Meta's 6GW Instinct commitment isn't validation of AMD parity—it's validation that AMD is *good enough* at half the price. NVIDIA's networking moat (InfiniBand, NVLink, Spectrum-X) is real, but merchant alternatives (Broadcom, Marvell) are maturing. The article ignores that NVIDIA's guidance excludes China; AMD's doesn't. If China data center demand softens, AMD's lower margin and higher leverage make it more vulnerable.
NVIDIA's $48B quarterly FCF and 80% buyback authorization assume data center TAM expansion continues unabated; if capex intensity plateaus and hyperscalers shift to AMD for cost arbitrage, NVIDIA's multiple compresses faster than the article suggests.
"NVIDIA’s data-center moat, software ecosystem, and hyperscaler demand create a durable earnings power that justifies a continued premium, even if AMD narrows the gap on growth and margin."
The article highlights NVIDIA’s data-center surge and AMD’s growing yet thinner margin profile, framing NVIDIA as the utility and AMD as the credible second source. The strongest counter is that AI compute demand remains highly capex-driven and cyclical; a slower hyperscaler spend, rising interconnect/capex costs, or stricter China export controls could compress NVIDIA’s 75% GM and $48B FCF, even as the AMD ramp (MI450/Helios) faces its own conversion risk. The numbers quoted (e.g., AMD at 175x) raise red flags about data quality. While secular AI demand is real, near-term risk is skewed to a slowdown or multiple compression if spending slows or normalization occurs.
Bear case: AI capex could normalize faster than expected or face policy/regulatory headwinds, causing margins to compress and re-rate the group downward even if pilots like MI450/Helios ramp; the hype may outpace durable earnings power.
"AMD's China exposure creates unpriced asymmetric downside versus NVIDIA's cleaner guidance."
ChatGPT's data-quality flag on the 175x P/E is fair, yet the panel still underweights second-order China risk. AMD's unhedged exposure plus lower 55% margins creates asymmetric downside if U.S. export rules tighten further into 2026; NVIDIA's guidance already strips China, giving it cleaner visibility. This asymmetry isn't priced into AMD's 'second-source premium.'
"The commoditization of networking via merchant silicon poses a greater long-term threat to NVIDIA's margins than AMD's chip competition."
Claude, you’re missing the power of the 'merchant silicon' ecosystem. By focusing on AMD vs. NVIDIA, we ignore that hyperscalers like Meta aren't just buying chips; they are building custom silicon and proprietary fabrics to bypass NVIDIA's entire stack. If Broadcom and Marvell successfully commoditize the interconnect, NVIDIA’s 75% gross margin is not a moat—it’s a target. The real risk isn't AMD’s execution; it's the systemic erosion of NVIDIA’s proprietary networking lock-in by the broader merchant ecosystem.
"Commodity interconnects erode pricing power only if software lock-in simultaneously weakens—Gemini conflates the two."
Gemini's merchant silicon thesis is seductive but conflates two separate risks. Custom silicon (Meta's TPUs, OpenAI's inference chips) and commodity interconnects are different problems. NVIDIA's moat isn't just networking—it's software ecosystem depth (CUDA, cuDNN, TensorRT). Broadcom commoditizing fabric doesn't kill 75% margins if hyperscalers still need CUDA. The real test: does AMD's ROCm ecosystem close the software gap? Nobody's flagged that yet.
"Macro demand and policy tail risks threaten valuation multiples far more than current 'merchant silicon' fears imply."
Gemini’s merchant-silicon threat is real, but the moat isn’t collapsing on day one; CUDA/TensorRT and software stack raise switching costs beyond hardware interconnect. The bigger, underappreciated risk is a capex slowdown or stricter export controls that curb AI compute demand, compressing multiples even if pilots ramp. If that tail risk materializes, AMD’s 6GW push matters less for equity value than a broader demand pullback for hyperscalers.
The panel agrees that NVIDIA's 75% gross margins and software lock-in make it the dominant AI infrastructure utility, but there's disagreement on AMD's threat and the risks associated with China exposure and potential slowdown in AI compute demand.
AMD's growing presence as a credible second source, particularly in the context of NVIDIA's potential supply constraints.
A slowdown or stricter export controls that curb AI compute demand, compressing multiples and affecting both NVIDIA and AMD.