[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]  FundamentalBottom [@FundaBottom](/creator/twitter/FundaBottom) on x 2499 followers Created: 2025-07-17 13:18:05 UTC After Mark Zuckerberg announced the 5GW data center construction plan, we quickly interviewed a data center expert to understand the latest developments. $Meta Prometheus is expected to bring 1GW online in 2H26, equipped with GB300 GPUs for training scenarios. Hyperion is expected to bring 1.4GW online in 2027, equipped with Vera Rubin GPUs, expanding to 5GW by 2030. This project has a training:inference ratio of 35%:65%, with experts projecting AMD could secure a XX% share of the inference portion. Meta Prometheus in Ohio: Total 1GW capacity expected to be fully online by 2H26. This cluster is for training, primarily equipped with NVIDIA GB300 (460,000 GB300 cards). Experts expect this cluster won't procure AMD cards. The project's total CAPEX is estimated at $25-26 billion. Meta Hyperion in Louisiana: 1.4GW will come online in 2027, primarily equipped with Vera Rubin GPUs, prioritized for training. Experts expect this project to expand to 5GW by 2030, with a training:inference ratio of 35%:65%. For inference, experts project AMD could secure a XX% share. Meta Titan project is still in early stages, and experts don't believe Meta has such substantial additional demand beyond the total 6GW from Prometheus + Hyperion. Experts believe Meta's 18-month lease for 100K B-series cards from Oracle will likely not be renewed after expiration as Prometheus comes online. XXXXX engagements  **Related Topics** [tiktok](/topic/tiktok) [meta](/topic/meta) [vera](/topic/vera) [$meta](/topic/$meta) [mark zuckerberg](/topic/mark-zuckerberg) [meta platforms](/topic/meta-platforms) [stocks communication services](/topic/stocks-communication-services) [Post Link](https://x.com/FundaBottom/status/1945835399131406499)
[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
FundamentalBottom @FundaBottom on x 2499 followers
Created: 2025-07-17 13:18:05 UTC
After Mark Zuckerberg announced the 5GW data center construction plan, we quickly interviewed a data center expert to understand the latest developments.
$Meta Prometheus is expected to bring 1GW online in 2H26, equipped with GB300 GPUs for training scenarios. Hyperion is expected to bring 1.4GW online in 2027, equipped with Vera Rubin GPUs, expanding to 5GW by 2030. This project has a training:inference ratio of 35%:65%, with experts projecting AMD could secure a XX% share of the inference portion.
Meta Prometheus in Ohio: Total 1GW capacity expected to be fully online by 2H26. This cluster is for training, primarily equipped with NVIDIA GB300 (460,000 GB300 cards). Experts expect this cluster won't procure AMD cards. The project's total CAPEX is estimated at $25-26 billion.
Meta Hyperion in Louisiana: 1.4GW will come online in 2027, primarily equipped with Vera Rubin GPUs, prioritized for training. Experts expect this project to expand to 5GW by 2030, with a training:inference ratio of 35%:65%. For inference, experts project AMD could secure a XX% share.
Meta Titan project is still in early stages, and experts don't believe Meta has such substantial additional demand beyond the total 6GW from Prometheus + Hyperion.
Experts believe Meta's 18-month lease for 100K B-series cards from Oracle will likely not be renewed after expiration as Prometheus comes online.
XXXXX engagements
Related Topics tiktok meta vera $meta mark zuckerberg meta platforms stocks communication services