Building @Botanika_sol and the data infrastructure behind AI.
Seoul, Korea
- Tweets 3,314
- Following 576
- Followers 43,181
- Likes 4,393
Two AI infrastructure updates caught my eye this week.
NVIDIA is putting more attention on power and cooling for AI data centers. Google Cloud is giving customers better visibility into how their storage is actually used.
Both point to a practical question I think we’ll hear more often. Once the hardware is deployed, how well does the whole system run?
That’s the question I keep coming back to as we build Botanika.
21
1,001
Glad to have Metrix Capital on board 🚀
Metrix Capital joined Botanika’s $1.5M round.
@botanika_sol is building at the point where three narratives converge:
→ DePIN supplies the hardware
→ RWA shapes the ownership
→ AI creates the demand
NIMBUS is the first output: hardware that actually runs, a network that puts it to work, and a token model funded by usage, not inflation.
Infrastructure ownership, distributed instead of concentrated in five companies.
1
2
27
1,438
Siwon Kim retweeted
$2.2M. 24 months. A live RATP / Riyadh Metro pilot.
Botanika’s NURA data indexing and structuring engine is already processing data as part of the program, supporting AI agent and data services across a real-world enterprise workload.
Commercial launch readiness isn’t just NIMBUS hardware entering production or $BONSAI approaching TGE.
It also means the data infrastructure already has work to do.
4
4
20
423
thank you @SuperteamKorea for the shoutout and hoping to catch everyone at the @solana @SolanaSummitOrg in Korea & Singapore!
excited to share more about what we are building at @botanika_sol
Mark your calendars for Oct 6! Right after Solana Summit Korea (Sep 30), Solana Summit Singapore takes the stage.
🇰🇷 Check how Korea’s next wave is already shipping on @solana 's AI and RWA stack.👇
@botanika_sol is building AI infrastructure you can actually own, native hardware paired with RWA-backed data storage, so compute and storage aren’t just rented from the cloud.
@vezta_io turns prediction markets into a single trading flow, bundling venues into one basket and one click.
@YumiFinance is issuing onchain debt instruments for the teams building robotics and AI, underwriting the work instead of waiting for traditional credit.
@RewardyJapan is the consumer layer: a wallet where 1M+ users earn rewards and spend crypto like cash. And
@idolly_AI lets fans create and collect their own AI idols on Solana.
Different products. Different bets. Same ecosystem.
Catch them at @SolanaSummitOrg on Oct 6 in SG 🫡
3
2
18
1,270
Western Digital's CEO said something wild on their earnings call: they're "pretty much sold out for calendar 2026." Every hard drive. Spoken for. And it's September.
Been sitting with that for a few days. everyone's still watching @nvidia and GPU allocations like that's the whole game. meanwhile the company that makes the drives is telling you outright there's nothing left to sell for the rest of the year.
And that's before you even count how many new data centers are planned to get built next year.
16
1,218
Siwon Kim retweeted
Compute is becoming abundant. Proprietary data isn’t.
As AI infrastructure scales, the harder question becomes who stores, controls and exposes the data models depend on.
Our latest deep dive: Why Data Control Becomes AI Infrastructure
1
2
12
561
17.1 petabytes per second
That's where enterprise data creation is expected to be by 2029, roughly 2.5x 2025 levels.
Every new agent, model and application adds another layer of data that has to be stored, moved and accessed somewhere.
This is pretty much the problem space we've been obsessing over at @botanika_sol.
And imo we're still very early in figuring out what the infrastructure for that scale of data should actually look like.
17
1,008
Ok so nobody's talking about this enough
@nebiusai is up 87% this year. just locked in $19.4B with @Microsoft and $27B with @Meta.
Microsoft. The company that could build its own infra if it wanted to. Instead it's writing a $19B check to a company that didn't matter 3 years ago.
@CoreWeave's doing the same thing >> $6B data center campus, and somehow ended up as a vendor to Meta, @OpenAI AND @nvidia at the same time.
This is the tell. AI demand outran what @awscloud /@Azure /@Google can physically build. so the market just... made new infra companies to fill the gap.
Been thinking about this a lot because we're building into that exact gap at Botanika. different bet though >> not another GPU farm, a distributed network of independent nodes, AI doing the routing.
Closed $1.5M this month. Building...
1
20
1,107
$497b is expected to go into ai infrastructure this year alone.
And how that capital is being distributed ?
In q1, 97.6% of ai infra spending went into servers. Storage took 2.4%.
That's a lot of compute being built around a very small piece of the stack responsible for keeping the data around.
3
657
Siwon Kim retweeted
NIMBUS is where Botanika becomes physical.
Storage capacity leaves the data center and moves into infrastructure owned by the network.
Data comes in. Nodes take on the workload. Operators keep the infrastructure running.
That’s the network we’re building.
1
3
18
665
AI workloads are getting insanely data hungry.
Models are larger, context windows keep growing and agents are constantly reading + writing data.
Feels like storage and data movement are still getting a fraction of the attention they deserve in the ai infra conversation. Imo there is a lot of value sitting in this part of the stack over the next few years.
2
11
1,045
Part of our mission at @botanika_sol is building a rewarding network that can scale consistently and sustainably.
Proud of our team for getting a step closer to figuring it out
1
1
12
654
