Will xAI reach 50 million H100-equivalents in 5 years?
10
Ṁ1311
2030
38%
chance

Will resolve based on solid, uncontroversial estimates or official claims.
Can resolve early. End date is July 22nd, 2030, but this market can wait for a good estimate to come out after this point - if it seems like this estimate implies the goal was met on time, it can still resolve YES.

Get
Ṁ1,000
and
S3.00
Sort by:

This probably means <=5M Nvidia Feynman GPUs: Blackwells are 2.5x as powerful as H100s, and >=2x jumps between generations is typical (Feynman after Rubin after Blackwell).

Nvidia sells a bit over 5M GPUs/year now and that's still growing. Given further growth and chip accumulation over a few years, 50M H100-equivs in 2030 would not be infeasible for a hyperscaler. Mainly a question of whether xAI is a hyperscaler by the standards of 2030.

bought Ṁ50 NO

As of mid-2025, xAI operates the world's largest AI cluster, "Colossus," with about 200,000 H100-equivalents.
Global AI compute expected to grow 10x by Dec 2027 to ~100M H100-eq, with xAI's usage share rising from 2% to 12%. Even with very aggressive scaling, 50M H100-eq (50% of total global compute) is implausible given diversified competitors (Google, Meta, China) also expanding heavily.
Since 2016, Musk promised Tesla vehicles would achieve fully autonomous driving within about a year, enabling coast-to-coast driving without human intervention. Nearly a decade later, Tesla’s FSD still requires human oversight, has missed countless deadlines, and faces ongoing safety and legal challenges.

© Manifold Markets, Inc.Terms + Mana-only TermsPrivacyRules