Founding Engineer @ LiquidEnergy (YC F26)💡 💡
Joined March 2018
- Tweets55
- Following127
- Followers32
- Likes1.4K
Christmas came early 👀
We built a new type of datacenter for AI inference. We call it NanoDC.
NanoDC is a solar-powered autonomous datacenter that can be deployed anywhere on the planet in 4 weeks.
11 Trillion will be invested in AI compute over the next 4 years. Most of this will be for inference and most of it will be severely delayed: stuck in grid interconnect queues, site permits, or supply chain issues. NanoDC sidesteps these issues completely.
NanoDC is built rapidly at a factory and shipped to the site ready to be turned on in 30 minutes. It generates its own power with solar and batteries. You open the door and the solar panels fold-out like an accordion. The compute is cooled with patent-pending high-density air cooling. It connects to the internet via Starlink.
No construction. No grid interconnect. No Water supply.
It never waits for transformers and turbine generators either.
Getting more compute is as easy as dropping more units and turning them on.
The first units will be live this month. Thousands will be shipping starting 2027.
Sunlight in. Tokens out.
az retweeted
So so proud of you, @sidravi_ 🔥
YEARS of single-pointedness, sacrifice, knowledge assimilation, questioning the limits of what’s possible, white papers on mind-bending theories the world is yet to see and sheer grit 🫡
May periva always be with you every step of the way🙏
az retweeted
So excited to see @_LiquidEnergy take off! Incredible innovation and hard work, @sidravi_ @SajKhosa @ArshSek @romistaro. So proud of you all!
Solar in, tokens out!! Let’s goooo! 🔥
We built a new type of datacenter for AI inference. We call it NanoDC.
NanoDC is a solar-powered autonomous datacenter that can be deployed anywhere on the planet in 4 weeks.
11 Trillion will be invested in AI compute over the next 4 years. Most of this will be for inference and most of it will be severely delayed: stuck in grid interconnect queues, site permits, or supply chain issues. NanoDC sidesteps these issues completely.
NanoDC is built rapidly at a factory and shipped to the site ready to be turned on in 30 minutes. It generates its own power with solar and batteries. You open the door and the solar panels fold-out like an accordion. The compute is cooled with patent-pending high-density air cooling. It connects to the internet via Starlink.
No construction. No grid interconnect. No Water supply.
It never waits for transformers and turbine generators either.
Getting more compute is as easy as dropping more units and turning them on.
The first units will be live this month. Thousands will be shipping starting 2027.
Sunlight in. Tokens out.
More than 30% of the land in the US is suitable for NanoDc, which is over 100 Terawatts of solar power.
Just sayin
We built a new type of datacenter for AI inference. We call it NanoDC.
NanoDC is a solar-powered autonomous datacenter that can be deployed anywhere on the planet in 4 weeks.
11 Trillion will be invested in AI compute over the next 4 years. Most of this will be for inference and most of it will be severely delayed: stuck in grid interconnect queues, site permits, or supply chain issues. NanoDC sidesteps these issues completely.
NanoDC is built rapidly at a factory and shipped to the site ready to be turned on in 30 minutes. It generates its own power with solar and batteries. You open the door and the solar panels fold-out like an accordion. The compute is cooled with patent-pending high-density air cooling. It connects to the internet via Starlink.
No construction. No grid interconnect. No Water supply.
It never waits for transformers and turbine generators either.
Getting more compute is as easy as dropping more units and turning them on.
The first units will be live this month. Thousands will be shipping starting 2027.
Sunlight in. Tokens out.
az retweeted
We built a new type of datacenter for AI inference. We call it NanoDC.
NanoDC is a solar-powered autonomous datacenter that can be deployed anywhere on the planet in 4 weeks.
11 Trillion will be invested in AI compute over the next 4 years. Most of this will be for inference and most of it will be severely delayed: stuck in grid interconnect queues, site permits, or supply chain issues. NanoDC sidesteps these issues completely.
NanoDC is built rapidly at a factory and shipped to the site ready to be turned on in 30 minutes. It generates its own power with solar and batteries. You open the door and the solar panels fold-out like an accordion. The compute is cooled with patent-pending high-density air cooling. It connects to the internet via Starlink.
No construction. No grid interconnect. No Water supply.
It never waits for transformers and turbine generators either.
Getting more compute is as easy as dropping more units and turning them on.
The first units will be live this month. Thousands will be shipping starting 2027.
Sunlight in. Tokens out.
Welcome to the future
We built a new type of datacenter for AI inference. We call it NanoDC.
NanoDC is a solar-powered autonomous datacenter that can be deployed anywhere on the planet in 4 weeks.
11 Trillion will be invested in AI compute over the next 4 years. Most of this will be for inference and most of it will be severely delayed: stuck in grid interconnect queues, site permits, or supply chain issues. NanoDC sidesteps these issues completely.
NanoDC is built rapidly at a factory and shipped to the site ready to be turned on in 30 minutes. It generates its own power with solar and batteries. You open the door and the solar panels fold-out like an accordion. The compute is cooled with patent-pending high-density air cooling. It connects to the internet via Starlink.
No construction. No grid interconnect. No Water supply.
It never waits for transformers and turbine generators either.
Getting more compute is as easy as dropping more units and turning them on.
The first units will be live this month. Thousands will be shipping starting 2027.
Sunlight in. Tokens out.
az retweeted
Incredibly excited to come out of stealth and launch Liquid Energy.
We are building the fastest path to abundance with energy and intelligence for everyone.
Intelligence is Liquid Energy.
We built a new type of datacenter for AI inference. We call it NanoDC.
NanoDC is a solar-powered autonomous datacenter that can be deployed anywhere on the planet in 4 weeks.
11 Trillion will be invested in AI compute over the next 4 years. Most of this will be for inference and most of it will be severely delayed: stuck in grid interconnect queues, site permits, or supply chain issues. NanoDC sidesteps these issues completely.
NanoDC is built rapidly at a factory and shipped to the site ready to be turned on in 30 minutes. It generates its own power with solar and batteries. You open the door and the solar panels fold-out like an accordion. The compute is cooled with patent-pending high-density air cooling. It connects to the internet via Starlink.
No construction. No grid interconnect. No Water supply.
It never waits for transformers and turbine generators either.
Getting more compute is as easy as dropping more units and turning them on.
The first units will be live this month. Thousands will be shipping starting 2027.
Sunlight in. Tokens out.
az retweeted
dawg im building a small lifestyle brand for scaling compute across the solar system hmu if you’re into casual shit like that
az retweeted
Anyone renting RTX GPUs?
I’m trying to understand the use cases for which people are renting lots of RTX GPUs such as RTX PRO 6000, 5090, 4090, etc.
I am building an off-grid modular compute unit that hosts 2 RTX PRO 6000s per node. I have lots of offers from marketplace/aggregator platforms for renting them. But I would really love to understand end users and their use cases and workflows.
The ideal use case for the compute I am building:
- Does not need more than 96GB VRAM per GPU
- Does not need high bandwidth clustering with NVLink and Infiniband
- Does not need super high bandwidth or super low latency internet
I am trying to understand scalable use cases for which people rent these consumer/workstation grade GPUs as opposed to clusters of B300/H200.
Some examples would be:
- Batch/flex tier inference on smaller LLMs (30-70B)
- Inference for diffusion models
- Training small robotics models with simulation environments that need RTX GPUs
- Fine tuning
- Domain specific models (incl for biotech, pharma, materials)
- Rendering heavy workloads with agents in the loop
Please post comments if you have insights or dm me if you can make some intros. If you fit in the above categories, I would love to chat and lunch/coffee is on me!
Every since I found out how to use a web browser, I've never seen a problem i thought I couldn't solve eventually.
In 2022-2023 that gap went from eventually to soon
And now it's a week max
I will soon get the solution before I propose the problem
when the contractor doesn't want to make a blast shield for your explosive computer fridge so you do it yourself
the strong do not explain themselves
maybe it's just me, but this venezuela thing feels different from the usual US military adventurism. at least for admirers of the US like me. in a way it never felt during the past few decades.
like, they just casually invaded a sovereign country, exfiltrated their president and paraded him on the streets of new york. obv maduro was no saint, but he wasn't a noteworthy antagonist either. this whole thing was just a naked exercise of power, nothing else. that just hits a bit different.
