The Hyperscale Lottery: How State-Space Models Have Sacrificed Edge Efficiency
Robin Geens ∗ † {}^{*\text{\;}\dagger} , Jonas De Schouwer † , Marian Verhelst ∗ , Thierry Tambe † · 2026
Read The Hyperscale Lottery: How State-Space Models Have Sacrificed Edge Efficiency by Robin Geens ∗ † {}^{*\text{\;}\dagger} , Jonas De Schouwer † , Marian Verhelst ∗ , Thierry Tambe † (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-07935.txt.
Description
Geens and three colleagues say state-space models traded edge efficiency for cloud throughput, a pattern they call the Hyperscale Lottery. The older Hardware Lottery says research follows available silicon; here, architectures chase hyperscale GPU saturation. Mamba models were praised for linear complexity suited to edge use, yet Mamba-3's changes impose a steep edge penalty. Latency rises 28 percent at 880 million parameters and worsens to 48 percent for 15 million parameter models. The authors urge separating cloud saturation tricks from core design to keep single-user real-time edge use viable.
Catalog record
| CREATOR | Robin Geens ∗ † {}^{*\text{\;}\dagger} , Jonas De Schouwer † , Marian Verhelst ∗ , Thierry Tambe † |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 11881 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.07935