TickerTain
TickerTain
NewsroomShortsPortfolioConvergence
NewsroomShortsPortfolioConvergence
$ANTHROPIC·$MA····$INTC····$BLUE-ORIGIN·$SPCX····$CRWV····$CRM····$MSFT····$NVDA····$ORCL····$CURSOR·$AAPL····$OPENAI·$AMZN····$UBER····$GOOGL····$META····$TSLA····$DATABRICKS·$PERPLEXITY·$LYFT····$NBIS····$TSM····$LITE····$ANDURIL·
$ANTHROPIC·$MA····$INTC····$BLUE-ORIGIN·$SPCX····$CRWV····$CRM····$MSFT····$NVDA····$ORCL····$CURSOR·$AAPL····$OPENAI·$AMZN····$UBER····$GOOGL····$META····$TSLA····$DATABRICKS·$PERPLEXITY·$LYFT····$NBIS····$TSM····$LITE····$ANDURIL·
←

kav ul shah

T3 · host / generalist

Kav Ul Shah leads Pebble, a company optimizing data center energy efficiency for AI workloads through GPU power-performance curve optimization and grid-responsive computing.

1 call·1 name·100% bull·last heard 3 months ago·SemiAnalysis
track record

no scored calls yet — needs a stated position or a categorical verdict, with a matured window vs SPY

top calls

highest conviction · one per company
1sthigh conviction
$PEBBLEPebbleposition

Pebble optimizes GPU power-performance curves to boost tokens per watt and eyes 100 GW flexible power market

Pebble exploits the non-linear, saturating relationship between GPU power and performance — especially for memory-bound inference workloads — by dynamically capping power and clock frequencies per GPU, and is extending this to make AI clusters grid-responsive to unlock 100 GW of flexible power capacity in the US.

SemiAnalysis2026-06episode →

most discussed · click a bar to filter

  • $PEBBLE

recurring themes

  • AI Infrastructure1
  • Grid & Power Infrastructure1
1 total
$PEBBLE
Pebble
HIGHkav ul shah·SemiAnalysis·3 months ago·The GPU Power-Performance Curve Most Clusters Ignore | Researcher Conversations at GTC· position
Pebble optimizes GPU power-performance curves to boost tokens per watt and eyes 100 GW flexible power market
Pebble exploits the non-linear, saturating relationship between GPU power and performance — especially for memory-bound inference workloads — by dynamically capping power and clock frequencies per GPU, and is extending this to make AI clusters grid-responsive to unlock 100 GW of flexible power capacity in the US.
"most of these workloads running on like GPU clusters, they the relationship between power and performance is it's not linear... there is a point on on sort of like the power perfo…"
0:32
7
AI Infrastructuretailwind
GPU power-performance curve saturates, creating optimization opportunity for memory-bound inference
The power-performance relationship in GPU clusters is non-linear and saturates because inference workloads are memory-bound — decode stages consume power reading weights into HBM while SMs idle — enabling dynamic per-GPU power/frequency capping to improve tokens per watt without hurting latency SLOs.
7
Grid & Power Infrastructuretailwind
Data centers can unlock 100 GW flexible power by becoming grid-responsive
Today's data centers are static power consumers; by making AI workloads curtailable during peak grid periods without violating SLOs, operators can access 100 GW of flexible power capacity in the US, turning energy flexibility into a new asset class.