Perplexity's Chip Choice: NVIDIA Vs AMD

Perplexity's decision to run its multi-agent AI coding stack on NVIDIA's Vera CPUs rather than standard x86 server processors is a small procurement choice with a larger signal attached. Vera ran 1.5 times faster than standard server chips in Perplexity's deployment, according to reporting on the decision. That performance gap matters less on its own than what it represents: NVIDIA is no longer competing only on GPUs. It's now selling GPUs, CPUs, networking hardware, and software as a single connected stack — Blackwell, NVLink Fusion, and Spectrum-X, sold together rather than assembled piece by piece.

AMD, by comparison, is still building out the equivalent answer. CEO Lisa Su pointed to strengthening customer engagement around the MI450 series and its Helios platform, saying customer forecasts are exceeding the company's initial expectations. The real test of that statement lands in the second half of 2026, when MI450 volume shipments are expected to begin.

Key Points

• Perplexity chose NVIDIA's new Vera CPUs over standard x86 server chips for its multi-agent AI coding stack, citing 1.5 times faster performance
• NVIDIA's Q1 FY27 revenue hit $81.61 billion, up 85.2% year over year, with Data Center revenue at $75.25 billion
• AMD's Q1 FY26 revenue reached $10.25 billion, up 37.9% year over year, with Data Center revenue of $5.78 billion
• The Perplexity decision signals NVIDIA is expanding into server CPU territory that has traditionally belonged to AMD's EPYC line
• AMD's MI450 chip volume ramp is expected in the second half of 2026, a milestone the market is watching closely given the stock's valuation

The Numbers Behind the Narrative

NVIDIA's Q1 FY27 revenue reached $81.61 billion, up 85.2% year over year. Data Center revenue alone was $75.25 billion, up 92%, and Networking revenue grew 199% to $14.8 billion. CEO Jensen Huang described the current moment as the largest infrastructure expansion in history, driven by AI factory buildouts.

AMD's Q1 FY26 revenue was $10.25 billion, up 37.9% year over year, with Data Center revenue at $5.78 billion, up 57%, and non-GAAP earnings per share of $1.37. Free cash flow rose 253% to $2.57 billion. Both companies are growing. The gap is in scale: NVIDIA's non-GAAP gross margin sits at 75%, compared with AMD's 55%, and NVIDIA's networking business alone now generates more revenue than AMD's entire company.

Why the CPU Move Matters More Than It Looks

Vera's entry into server CPUs pushes NVIDIA into a roughly $20 billion market that has historically belonged to x86 vendors, with AMD's EPYC line as one of the strongest incumbents in that space. If NVIDIA can convert GPU relationships into CPU sockets the way the Perplexity deal suggests, AMD ends up fighting on two fronts simultaneously: chasing NVIDIA in AI accelerators with MI450 and Helios, while defending server CPU share it has held for years.

Software may end up mattering as much as hardware here. NVIDIA's CUDA-X and Dynamo software layer is widely credited with keeping inference workloads locked into its ecosystem, since switching platforms means rebuilding on a different software stack. AMD's ROCm 7 has narrowed that gap but has not yet reached the same level of default adoption. Customer commitments on paper look closer than the earnings numbers suggest — OpenAI has signed for more than 10 gigawatts of NVIDIA capacity and 6 gigawatts of AMD capacity — but sticky software tends to compound advantages over time in ways a single hardware deal doesn't fully capture.

What This Means Beyond the Chip Market

For teams tracking AI infrastructure as part of a broader AI adoption strategy, this dynamic is worth watching independent of any stock position: the companies building AI-dependent products are increasingly locked into whichever hardware and software stack offers the least friction, and that lock-in shapes cost, speed, and reliability further down the chain. The Perplexity decision is one data point, not a verdict, and AMD's Data Center growth and free cash flow gains show a business that is still expanding in absolute terms even as NVIDIA pulls further ahead on scale.

This article covers publicly reported earnings data and a reported vendor decision. It is not financial advice, and any investment decision should involve independent research or a licensed financial advisor.