NVIDIA Corporation
Technology company specializing in semiconductors and AI chips, IPO in 1999, Taiwan semiconductor dependency
Every fact below is drawn from Via News's knowledge graph and linked to the source document it came from. Nothing here is unsourced. How we source →
Key metrics · each point sourced
Stated objectives
Regain billions of dollars in lost business through China market
sourceShare 15% or more of revenue from China sales with US government
sourceCommission over 1 million square feet for manufacturing Blackwell chips in Arizona and AI supercomputers in Texas
sourceRamp up mass production of AI supercomputers in Texas by September 2026
sourceQ4 FY2026 revenue target
sourceQ4 FY2026 GAAP gross margin target
sourceQ4 FY2026 non-GAAP gross margin target
sourceStart shipping H200 chips to China
sourceDeliver up to 10x reduction in inference token cost compared to Blackwell platform
sourceReduce number of GPUs needed to train MoE models by 4x compared to Blackwell
sourceRelationship graph · 5341 connections
Where sources disagree
We surface conflicts between sources rather than hiding them.
Fact A claims 1.8x faster task completion vs x86 CPUs (May 31), while Fact B claims 50% faster (1.5x) vs traditional rack-scale CPUs (March 16). The speed improvement claims conflict: 1.8x vs 1.5x represents a 20% discrepancy. While they compare against different baselines (x86 specifically vs general rack-scale), the claimed performance improvement should be relatively consistent across comparable processor classes. The later claim (Fact A) is more optimistic, which could reflect updated benchmarks, different workloads, or marketing adjustments rather than genuine product differences.