OpenAI Dissatisfied With Nvidia’s AI Chips, Explores Alternative Suppliers

NVDANVDA

OpenAI is unsatisfied with some of Nvidia’s latest AI chips and has been seeking alternatives since last year, according to eight sources. This development could complicate Nvidia’s partnership with its largest AI customer and pressure future GPU sales in the face of rising competition.

1. NVDA Set to Exceed Fiscal 2027 Revenue Projections

Nvidia’s management has guided analysts toward fiscal 2027 revenue in excess of $323.3 billion, and current demand signals indicate the company will surpass that target by a wide margin. Backed by strong order backlogs for its next-generation Blackwell GPU systems and expanding rack-scale solutions, channel checks suggest quarterly sales growth of over 40% year-over-year through the first half of fiscal 2027. Institutional investors have revised their forecasts upwards in 12 of the last 14 analyst reports, citing record-high data center bookings and deferred revenue balances that now exceed $15 billion.

2. Gross Margins to Remain Near 75% Despite Cost Pressures

Even as Nvidia ramps up production volumes to meet surging AI infrastructure demand, the company expects to hold its gross margin around 75%, just below the historic high of 76.5% achieved last quarter. Management attributes this resilience to economies of scale in its TSMC 5-nanometer manufacturing process and ongoing price discipline on premium SKUs such as the H200 and blackwell-based A800 accelerators. Operating leverage from expanding software-as-a-service offerings is also expected to contribute incremental margin expansion in the second half of fiscal 2027.

3. Dominant GPU Market Share Underpins Competitive Moat

Nvidia currently commands approximately 92% of the overall GPU market, and the company is on track to defend that position against competitors including AMD and Qualcomm. Key to this dominance is the breadth of Nvidia’s CUDA software ecosystem—now supported by over 1.5 million developers worldwide—and its suite of performance-tuning tools that shorten time-to-market for generative AI applications. Industry surveys show that 87% of hyperscale cloud providers plan to standardize on Nvidia GPUs for both training and inference workloads, reinforcing the firm’s position at the center of the AI computing stack.

Sources

FFFZY
+15 more