️ ELON MUSK: "99% of compute long-term will be for inference."

What happened

Elon Musk said that 99% of computing power in the long term will be used for inference, which means running AI models to produce results rather than training them.

Confirmed

Global impact / market context

When most compute is needed for inference, demand will move toward hardware and services that run AI models efficiently, benefiting chip makers and cloud providers that focus on inference‑optimized solutions.

Analyst inference

The AI surge has driven huge spending on training hardware, but a shift to inference could redirect capital toward data‑center capacity and specialized chips, influencing valuations of firms in those segments.

Analyst inference

What to watch

1. Chip makers' product roadmaps – watch for announcements of inference‑optimized processors, signaling expectations of higher inference demand and potential shifts in R&D spending. Analyst inference
2. Cloud providers' capacity plans – monitor new data‑center builds or upgrades aimed at inference workloads, indicating how they allocate capital to support AI services. Analyst inference
3. AI startups' business models – track shifts toward inference‑as‑a‑service offerings, reflecting broader market movement from training‑focused to inference‑focused revenue streams. Analyst inference

Evidence

• x.com