Public anger: users switching from NVIDIA
Social posts show at least one user celebrating a switch from NVIDIA to AMD and another musing about moving to macOS for personal reasons — a small but noisy sentiment backlash that could echo into procurement conversations if it scales. These posts are raw sentiment, not technical procurement signals, but they’re visible in dev channels. (x.com (x.com)
Legendary programmer John Carmack posted on X that his DGX Spark delivered roughly half the quoted compute and was drawing about 100W, a post that sparked wide technical pushback. (me.pcmag.com) NVIDIA announced DGX Spark shipments on Oct. 13, 2025 as a compact developer AI system with up to 128GB unified memory and petaflop-class sparse throughput. (investor.nvidia.com) Independent testing and developer forums soon reported thermal throttling and lower-than-advertised sustained performance from units in the field. (tomshardware.com) Tech outlets and forum threads documenting those Spark failures also recorded public posts from users saying they were moving to AMD hardware or alternative platforms after their Spark experiences. (tomshardware.com) At the same time, Oracle announced a plan to deploy 50,000 AMD Instinct MI450 accelerators, signaling a concrete procurement pathway away from NVIDIA for cloud and enterprise customers. (cnbc.com) AMD has been beefing its software and hardware stack—acquiring compiler/optimization startup Brium on June 4, 2025 and pushing MI300X-based full-rack AI systems to compete in data-center deployments. (techcrunch.com) Internal NVIDIA emails show CEO Jensen Huang personally directed teams to engage critics and fix issues as Spark criticism mounted, and the company promoted firmware and SDK updates at CES 2026 to address developer complaints. (businessinsider.de) NVIDIA still offers startup-focused procurement incentives—its Inception program and NVIDIA AI Enterprise pricing let qualifying startups buy reduced-price one‑year subscriptions (up to 64 per member) to the company’s software stack. (docs.nvidia.com)
