Following the AI stack from models to inference. Thoughts on Kimi, DeepSeek, agents, benchmarks, and what's actually useful.
Santa Calra, California
- Tweets 946
- Following 2,262
- Followers 1,828
- Likes 1,389
Ethan retweeted
Your compute shouldn't be stuck on "loading..."
Modular data centers. Up and running in just 3 months.🟢
#AIInfrastructure #ModularDataCenter #DataCenters
2
5
11
439
Ethan retweeted
What is an NVIDIA Reference Architecture?
A validated design for AI infrastructure—bringing compute, networking, storage, and software together into one scalable system.
1/5🧵
1
3
8
246
Ethan retweeted
AI compute starts on the factory floor.
From structural assembly to equipment installation, Canopy Wave’s modular AI data center is taking shape—building the foundation for GPU workloads.
#AIInfrastructure #ModularDataCenters
3
8
298
From architecture to production-ready delivery
Built on NVIDIA Reference Architecture. Delivered with Speed.
Canopy Wave designs and delivers complete GPU Clusters, with designs, ensuring full alignment with NVIDIA Cloud AI Infrastructure standards from day one.
Our expert team integrates compute, networking, and storage, speeding up deployment through preconfiguration and automation.
Design → Deploy → Validate → Deliver
One team, from architecture to production-ready delivery.
Talk to our engineers👇
17
Ethan retweeted
Take your AI performance further with NVIDIA HGX B300 on Canopy Wave.
Powered by @nvidia's Blackwell Ultra architecture, B300 delivers approximately 2.3x the peak FP8 compute performance of H200.
Built for production AI with:
✅99.99% uptime
✅SOC 2 Type II
✅24/7 Support
Scale with confidence. Reserve your B300 capacity today👇
canopywave.com/hgx-b300?utm_…
#NVIDIA #B300 #CanopyWave
1/5
2
5
12
383
starts here
From Power to Token starts here.
Inside the factory, Canopy Wave’s modular AI data center is taking shape.
We’re building the physical foundation that connects power, data center capacity, GPU compute, and AI inference.
Built module by module. With scale in mind.
#AIInfrastructure #ModularDataCenters #AICompute #FromPowerToToken #CanopyWave
24
From Power to Token
MiMo V2.6 Flash on Canopy Wave delivered nearly 3x higher total throughput than the other provider under the same test conditions.
From power to token, speed is built into every layer.
Here's what makes that possible.🧵
#AIInference #AIInfrastructure #MiMo #CanopyWave #FromPowertoToken
29
Ethan retweeted
Day 0: MiMo V2.6 Flash is live on Canopy Wave
Stronger agents. Bigger performance gains.
- Up to 3.6x higher performance than MiMo V2.5 across agent tasks.
- Competitive with GPT-5.6 Sol performance at roughly 1/30 to 1/70 the cost.
Start building now👇
4
6
24
1,952
Ethan retweeted
Build with DeepSeek-V4.1-Flash through Canopy Wave’s OpenAI-compatible API.
5
10
560
Ethan retweeted
Built for the latest-generation @nvidia GPUs. Designed to Tier III standards.
Canopy Wave’s modular AI data center is engineered for high-density GPU environments:
• 24/7 power delivery
• 99.8% measured availability
• N+1 redundant configuration
Explore more: canopywave.com/modular-ai-da…
6
9
1,660
Ethan retweeted
90% factory-based prefabrication. Targeting core deployment in 3 months.
Canopy Wave’s modular AI data center moves integration and testing into the factory—before arrival on site for assembly.
Compared with traditional construction that relies heavily on on-site installation and integration, our approach moves more of that work upfront.
More preparation in the factory. A faster path to deployment.
Explore:canopywave.com/modular-ai-da…
5
12
448
Jensen was live on stage. Then Trump called 📞
MUST SEE: Amazing moment as President Trump calls Nvidia CEO Jensen Huang while he's on stage at the All-In Summit.
@POTUS on AI Doomerism: “I'm telling you, it's all a hoax… and we're not going to let that happen.”
1
31
Ethan retweeted
Introducing Canopy Wave’s modular AI data center.
- Targeting 3-6-month core deployment cycles
- Designed to Tier 3 standards
- Built for the latest-generation NVIDIA GPUs
Built for the AI factories of today and tomorrow.
🧵1/6
3
6
12
6,488
Ethan retweeted
⚡Day 0 on Canopy Wave: DeepSeek-V4.1-Flash is now live.
Built on DeepSeek’s new architecture, V4.1 Flash brings native visual understanding alongside stronger coding and agent capabilities.
🖼️Native image + text understanding
🤖Built for agentic workflows
📚Up to 1M context
⚡Faster, more efficient inference
From understanding visual inputs to tackling long-context coding and multi-step agent tasks, V4.1 Flash expands what the Flash series can handle.
Try DeepSeek-V4.1-Flash on Canopy Wave 👇
3
7
22
1,138
Choosing the GPU is only half the infrastructure decision.
Self-built. Public cloud. Private GPU cloud.
There’s no single "best" way to run production AI workloads — only the model that fits your infrastructure needs.
We compare the trade-offs across deployment speed, control, security, and operations.
Full breakdown 👇
1
30
Ethan retweeted
Reservations for NVIDIA HGX B300 are now open on Canopy Wave.
Powered by @nvidia’s latest-generation Blackwell Ultra GPUs, B300 delivers 288GB of GPU memory for demanding AI workloads.
Plan ahead. Scale from development to production on secure B300.
Reserve now👇
1
6
12
1,172
GPT-6 just turned an iPad mini into a gaming PC.
I used GPT-6 Astra to get AAA PC games running on my iPad mini!
Skyrim, Batman: Arkham City, Hades, and Age of Empires II.
The actual Windows versions, running locally with touch controls. No streaming.
Demo below and I'll share a repo and setup guides soon!
29
Finding the right GPU cloud just got a little easier.
We're now listed on GetDeploying.⚡️
If you're comparing GPU cloud options for your next AI workload, you can now find Canopy Wave there too: getdeploying.com/canopywave
32