Tenstorrent Galaxy Blackhole superclusters are deployed at scale with customers @aiand_ , @Cirrascale
, @VirtuFinancial
, and Turiyam across a broad range of use cases from neoclouds to financial to sovereign AI.
COMPUTING
-
Tenstorrent Galaxy Blackhole Superclusters Deployed at Scale with Customers
By
–
-
Tenstorrent Powers Equinix Distributed AI Hub for Enterprises
By
–
Tenstorrent is the compute layer for Equinix's Distributed AI Hub enabling a full end-to-end agentic AI stack for enterprises. @Equinix @orionvm
-
Tenstorrent Outperforms GPUs at 350+ TSU With Flat Costs
By
–
In theory, GPUs serve at 300 tsu. In practice, no one is serving at those speeds because of poor economics. Tenstorrent serves at 350+ tsu while keeping costs flat. @JasminaVas
-
Tenstorrent Blackhole Chip Built for Scalable General Purpose AI Compute
By
–
AI is constantly changing. Tenstorrent is built for the future from first principles focused on scale, general purpose, and lower cost compute. Big or small, it runs on Blackhole. @jimkxa
-
AI Agents Drive New Infrastructure Demands for Modern Inference
By
–
AI isn’t just generating answers anymore. It’s running workflows.
— SambaNova (@SambaNovaAI) 1 mai 2026
That shift changes everything about infrastructure. CPUs, GPUs, and RDUs each play a role, working together to keep agents moving.
This is what modern inference looks like 🦾 pic.twitter.com/GkSjTvOFi3AI isn’t just generating answers anymore. It’s running workflows. That shift changes everything about infrastructure. CPUs, GPUs, and RDUs each play a role, working together to keep agents moving. This is what modern inference looks like
-
RTX PRO 4000 vs RTX 3090 VRAM Bandwidth and NVLink Compared
By
–
Just some numbers so you don’t get misled RTX 3090 (7 years old)
> 24GB VRAM
> Bandwidth: 936.2 GB/s
> Bi-directional NVLink 112GB/s RTX PRO 4000
> 24GB VRAM
> Bandwidth: 672 GB/s > No Bi-directional NVLink,
> need 32 Gen. 5 PCIe Lanes to pool 2 at 64GB/s x.com/LLMJunky/statu… -
MakeMyTrip Reduces Personalization Latency with Apache Spark Structured Streaming
By
–
.
@MakeMyTrip reduced personalization latency from about 1.23 seconds to 44 ms P50 by adopting Real-Time Mode in Apache Spark Structured Streaming, achieving millisecond performance without adding a second engine. Processing traveler searches as continuous streams delivered:
– -
Webinar: Build Production-Ready Apps Faster with Modern Frameworks and Serverless Postgres
By
–
You want to build, not spend hours setting up infrastructure. In this webinar, discover how to go from an idea to a production-ready app faster using modern frameworks, real-time iteration, and serverless Postgres. Join experts from Databricks and @cursor_ai on May 5 (AMER),
-
Run AI Agent On-Device in Chrome Using Gemma 4 2B Model
By
–
You can now run an AI agent on-device, right inside Chrome.
— AlphaSignal AI (@AlphaSignalAI) 1 mai 2026
No servers. No data leaving your device.
The browser extension uses Gemma 4 E2B, a 2B parameter model with native tool calling baked into its chat template.
The model decides when to call tools and which ones, all… pic.twitter.com/obpX6r1prRYou can now run an AI agent on-device, right inside Chrome. No servers. No data leaving your device. The browser extension uses Gemma 4 E2B, a 2B parameter model with native tool calling baked into its chat template. The model decides when to call tools and which ones, all
-
Remote Infrastructure Management via SSH WireGuard DNS and GPU Job Dispatch
By
–
~2,000 miles from home – SSH into the mothership
– WireGuard tunnel like I’m on LAN
– Resolve everything over private DNS
– Routes via reverse proxies – Dispatch jobs to GPUs
– Sync state across agents
– Search conversations, traces, artifacts, logs Read how below