Accelerator forecasts bundle memory; sell-side still sees over $1T market by 2030
BenBajarin · x · 2026-11-17
Analyst Ben Bajarin notes that all the big X-hundred-billion-dollar AI accelerator forecasts include memory in their totals. His shop keeps two models—one for GPU/XPU package value alone, one including memory—but both require separate assumptions for each component to reach the dollar figure.
He adds that sell-side estimates for 2030 remain north of $1 trillion, reflecting continued optimism about long-term AI compute demand.
More from Infra
- Philox RNG-to-tensor assignment is extremely sensitive to hardware details — ezyang · 2026-11-17
- Why tensor parallelism breaks RNG reproducibility in model init — ezyang · 2026-11-17
- Bittensor's Engy tops Venice's AI inference volume despite 28x smaller valuation — markjeffrey · 2026-11-17
- Fractile raises $720M for AI inference chips, betting inference cost breaks AI economics first — YvesMulkers · 2026-11-17
- DeepSeek V4.1 Flash on Vera Rubin NVL72: 63-93x per-chip throughput vs GB300 in vLLM benchmarks — vllm_project · 2026-11-17
- First DeepSeek V4.1 Flash results on Vera Rubin NVL72: up to 93.3x per-chip throughput vs GB300 — vllm_project · 2026-11-17