Perplexity AI chooses Nvidia infrastructure exclusively over AMD for inference, cementing Nvidia's ecosystem lock-in with AI application layer.
NVIDIA (NASDAQ: NVDA) and AMD (NASDAQ: AMD) both closed strong quarters, but the ground shifted this week when Perplexity selected NVIDIA's new Vera CPUs over x86 server chips for its multi-agent AI coding stack, running 1.5 times faster than standard server processors. That decision reframes the earnings comparison. One company now sells GPUs, CPUs, networking, and software as one bundle. The other is still assembling its answer.
NVIDIA's Q1 FY27 showed revenue of $81.61 billion, up 85.2% year-over-year, with Data Center at $75.25 billion (+92%) and Networking at $14.8 billion (+199%). Jensen Huang framed it plainly: "The buildout of AI factories, the largest infrastructure expansion in human history, is accelerating at extraordinary speed." Blackwell, NVLink Fusion, and Spectrum-X are selling as one unified stack.
AMD's Q1 FY26 was solid but smaller. Revenue hit $10.25 billion (+37.9% year-over-year). Data Center reached $5.78 billion (+57%), and non-GAAP EPS came in at $1.37. Lisa Su leaned on the pipeline: "Customer engagement around MI450 Series and Helios is strengthening, with leading customer forecasts exceeding our initial expectations." The MI450 volume ramp lands in the second half of 2026.
The Perplexity win validates NVIDIA's push into a new $20 billion CPU vertical that once belonged to x86 vendors. NVIDIA is converting GPU market share into host-CPU sockets, exactly the margin territory AMD's EPYC franchise defends. AMD fights a two-front war: chase NVIDIA in accelerators with MI450 and Helios while protecting server CPU turf from Vera.
Customer deals look close on paper. OpenAI signed 10+ gigawatts with NVIDIA and 6 GW with AMD. But NVIDIA's CUDA-X and Dynamo software layer keeps inference workloads sticky. ROCm 7 is improving but still short of moat status.
Two questions will prove decisive. First, whether MI450 shipments in the second half of 2026 meet the "exceeding expectations" language Su used, given AMD's 184x trailing P/E leaves no room for a slip. Second, whether Vera CPU adoption spreads beyond Perplexity into hyperscalers. NVIDIA guided Q2 revenue to $91.0 billion, and analysts carry a target of $301.62.
I lean NVIDIA here. The Perplexity decision signals that inference buyers default to Blackwell plus NVLink plus CUDA when latency matters. That is a moat. AMD remains healthy. Data Center up 57% and free cash flow of $2.57 billion (+253%) show the business is compounding. But shares already ran 140.99% year-to-date, and last week gave back 11.15%. For the platform winner, NVIDIA is cleaner. For the higher-variance catch-up trade, AMD works, provided Helios ships on time.