Olivier Blanchard πŸ’¬ (@oablanchard) 's Twitter Profile
Olivier Blanchard πŸ’¬

@oablanchard

Research Director and Practice Lead at The Futurum Group: AI Devices + Semis, Automotive, Policy.
Co-author of Human/Machine + Futureproof πŸ‡«πŸ‡·|πŸ‡ΊπŸ‡²

ID: 14217862

linkhttp://www.futurumgroup.com calendar_today25-03-2008 18:49:51

127,127K Tweet

41,41K Followers

22,22K Following

Shay Boloor (@stocksavvyshay) 's Twitter Profile Photo

$NVDA JUST SAID AI AGENTS USE 10,000X MORE TOKENS THAN ONE-SHOT LLMS Compute demand for inference is about to explode -- inference is the tax on intelligence. This is why $AMD made the Futurum AI Fifteen: scaling the chips driving the next wave of agentic computing πŸ‘€

$NVDA JUST SAID AI AGENTS USE 10,000X MORE TOKENS THAN ONE-SHOT LLMS

Compute demand for inference is about to explode -- inference is the tax on intelligence. 

This is why $AMD made the Futurum AI Fifteen: scaling the chips driving the next wave of agentic computing  πŸ‘€
Daniel Newman (@danielnewmanuv) 's Twitter Profile Photo

$AMD - Shots fired. 40% more tokens per dollar versus Blackwell. Like I saidβ€”AMD sees inference as an inflection for growth. $NVDA

$AMD - Shots fired. 

40% more tokens per dollar versus Blackwell.

Like I saidβ€”AMD sees inference as an inflection for growth. 

$NVDA
Patrick Moorhead (@patrickmoorhead) 's Twitter Profile Photo

.Signal65 didn’t publish these numbers, but it is good to see AMD showing FP4 inference performance on MI355X. Something I’ve been harping on. The 355 is a much improved product. AI at Meta seems pretty excited about it for recommendation workloads. β€œBang for the buck” and β€œTCO”

.<a href="/Signal_65/">Signal65</a> didn’t publish these numbers, but it is good to see <a href="/AMD/">AMD</a> showing FP4 inference performance on MI355X. Something I’ve been harping on. The 355 is a much improved product.

<a href="/metaai/">AI at Meta</a> seems pretty excited about it for recommendation workloads. β€œBang for the buck” and β€œTCO”
Patrick Moorhead (@patrickmoorhead) 's Twitter Profile Photo

Lisa Su: β€œI'm happy to announce that MI355 production shipments actually started earlier this month, and we have the initial wave of partners on track to launch platforms and public cloud instances here in the third quarter.” There we go…. $AMD

Lisa Su: β€œI'm happy to announce that MI355 production shipments actually started earlier this month, and we have the initial wave of partners on track to launch platforms and public cloud instances here in the third quarter.” There we go….
$AMD
Olivier Blanchard πŸ’¬ (@oablanchard) 's Twitter Profile Photo

Great to see AMD finally launch a developer cloud for ROCm. (Much clapping and elation in the room.) The demo makes it look very accessible. #AdvancingAI

Great to see <a href="/AMD/">AMD</a> finally launch a developer cloud for ROCm. (Much clapping and elation in the room.)
The demo makes it look very accessible.
#AdvancingAI
Ryan Shrout (@ryanshrout) 's Twitter Profile Photo

The results we are seeing with the AMD MI355X are pretty impressive so far. Yes, these results are looking at some specific use cases, but these are not corner cases, these are real-world AI deployment scenarios. Check out the work from Signal65!

Olivier Blanchard πŸ’¬ (@oablanchard) 's Twitter Profile Photo

Something else to keep in mind when considering AMD's roadmap: Look for Sovereign Cloud wins like this. What was the line at #AdvancingAI today about the Kingdom having a lot of land and energy, and a very young population? Don't sleep on Sovereign Cloud opportunities.