Nvidia CEO asserts unmatched AI performance per dollar

11 articles · Updated · TradingView · Apr 29

Jensen Huang emphasized Nvidia's AI leadership, citing MLPerf benchmarks showing the GB200 NVL72 platform achieving up to 30 times higher throughput than H200 in certain tests.
Nvidia highlighted total cost of ownership, including hardware, power, cooling, and software, as key to its value proposition, while recent software improvements further boosted Blackwell performance.
Despite Nvidia's claims, competition intensifies as Broadcom reports 106% AI chip revenue growth and Alphabet projects $175–$185 billion in 2026 AI infrastructure spending, reflecting evolving industry strategies.

As rivals build custom silicon, can anyone truly break Nvidia's grip on the AI market?

Nvidia claims a 35x lower cost-per-token. What does this mean for AI's real-world affordability?

Google claims its custom TPU has 4.7x better price-performance. Is this the chink in Nvidia's armor?

Inference is 80% of AI cost. Are specialized chips the smarter long-term investment over GPUs?

Nvidia has reserved most of TSMC’s capacity. How can competitors possibly catch up?

With AI racks demanding 120kW, is our power grid the real bottleneck for AI's future?