TuringData

AI storage for training and inference. Start lean. Scale freely. Make every GPU count.

[TuringData @TuringData_IO]

541.5 GiB/s from just 3 nodes.

Proud of our first MLPerf® Storage submission: high AI throughput from a lean TuringData architecture, supporting 96 simulated accelerators on 3D U-Net.

Read more from @MLCommons here: https://t.co/EprPXzlPZk

[TuringData @TuringData_IO]

Proudly, TuringData ranked #2 in 3D U-Net training throughput, while also showing strong performance across checkpointing and KV cache workloads.

Thank you, @MLCommons, for establishing such an open and rigorous benchmark! #MLPerf

[MLCommons @MLCommons]

MLPerf Storage v3.0 is out.

144 results. 19 orgs. 11 first-time submitters.

4 workloads: training, checkpointing, vector DB, KV cache.
New efficiency metrics: rack-unit density and provisioned power.
S3 API support. 8B to 1250B parameter checkpointing.
https://t.co/yEiumHDd8d

[TuringData @TuringData_IO]

Impressive progress from AMD. A 4× gain in rack-scale energy efficiency raises the ceiling, but scheduling, memory, caching, and serving architecture determine how much of that gain becomes useful token output. AI efficiency must be measured end to end.

[AMD @AMD]

More AI compute from every watt.​

AMD has reached an estimated 4x increase in rack-scale energy efficiency for AI training and inference from 2024 to 2026.​

See how innovation across the full AI stack is driving progress: https://t.co/107n5YPYDP

[TuringData @TuringData_IO]

Your #GPUs may be busy, but are they productive?Utilisation measures activity. Infrastructure yield measures results: cost per token, tokens per GPU-hour, and delivery within SLOs.
https://t.co/bLLNI7Eclw
via @FE_APAC

[TuringData @TuringData_IO]

How fast data reaches your GPUs decides how much of their capability turns into output. Same infrastructure. Different economics. #turingdata
💠Enterprise AI: more value from the GPUs you already own
💠AI Factories: more tokens per GPU, lower cost per token

[TuringData @TuringData_IO]

No Layer Works Alone: Where the Data Layer Fits in AI Infrastructure
https://t.co/cH95gv8PYT

[TuringData @TuringData_IO]

Happy 61st National Day, #Singapore 🇸🇬 #SG61

TuringData is headquartered here — building the data layer for #AIinfrastructure: higher #GPU utilisation across training and inference, better token economics as AI scales.

[TuringData @TuringData_IO]

One half of our logo gets all the attention. The other half is usually where the bottleneck is. The industry has done remarkable work making compute faster. We're focused on the path into it. #turingdata #AIstorage

[TuringData @TuringData_IO]

As #inference workloads scale, KV-cache management, memory orchestration and the underlying data infrastructure increasingly determine #AI performance, #GPU utilization and cost per token. A nice conversation during @superai_conf

https://t.co/Vb0D6vqhyc

[TuringData @TuringData_IO]

Better AI Token Economics starts with smarter storage. In our conversation with CRN Asia, we shared how organizations can maximize token throughput while minimizing infrastructure waste through AI-native storage.
Read the full interview 👇 https://t.co/7W0e57MbHx

[TuringData @TuringData_IO]

"Tokenmaxxing" is dead. Welcome to #Tokenminimizing. 📉 AI budgets are burning, but optimizing the model layer (cheaper LLMs, lean prompts) is only half the battle. The hidden TCO killer? Infrastructure. Slow data stacks stall GPUs, quietly driving up your true token costs.

[TuringData @TuringData_IO]

"Is TuringData a cloud storage provider like AWS or Google Cloud?"
Short answer: No.
And the difference matters for your AI ROI. See why:
https://t.co/rO1itU5kB9
#AIInfrastructure #AIStorage #GPUStorage #LLM

[TuringData @TuringData_IO]

Scaling AI is stuck on legacy data, high CAPEX, and idle GPUs. #TuringData fixes the data layer with software-defined storage:
Maximize GPU Use: 480 GB/s & 7.5M IOPS
Start with only 3 nodes, and unify distributed data without the high CAPEX
Watch: https://t.co/3XlsXPjRIx

[TuringData @TuringData_IO]

Building AI without solving your data layer is like building a Ferrari without fuel. 🏎️💨 Buying GPUs isn't enough—if the data feed is slow, your ROI plummets.
👉 Learn how to fix AI's biggest hidden bottleneck: https://t.co/3XlsXPjRIx
#AIInfrastructure #GPUs #AIStorage

[TuringData @TuringData_IO]

Stop counting GPUs. Start counting tokens. 📊
At #SuperAI 2026, #TuringData VP Nikhil Madan shared how to beat the "HBM Tax" & scale throughput.
Our Cache Fabric delivers:
⚡ 10x better TTFT
🚀 7x higher token throughput
Read how: https://t.co/OgsrEDfflV
#KVCache #Token

[TuringData @TuringData_IO]

Singapore isn’t just our HQ — it’s where we help AI teams unlock speed. At #ATxSG2026, our VP Nikhil Madan shared why Singapore is becoming a key AI hub in Southeast Asia and how we maximize GPU utilization.
🎥 Watch: https://t.co/3XlsXPjRIx
#AIInfrastructure #GPU #TuringData

[TuringData @TuringData_IO]

100% GPU utilization. 0% idle team members. 😄 That's a wrap on #SuperAI.
From booth conversations to technical deep dives, live podcast appearances, and everything in between — the #TuringData team showed up and delivered at the highest throughput. Onward. 🚀

[TuringData @TuringData_IO]

🔥 Final day of #SuperAI2026. Make it count.
Come find us at Booth PB16 to talk AI data infrastructure, inference at scale, agent systems, and yes — a small surprise for visitors. 👋
📍 Booth PB16 | Sands Expo & Convention Centre
#TokenEconomics #AIInfrastructure #TuringData

[TuringData @TuringData_IO]

🎤 Nikhil just wrapped his Forum Stage session at #SuperAI2026. Thank you to everyone who came out this morning!

⏰ He's back on in 20 min — live on Asia Tech Podcast at 2:40 PM SGT 👇
https://t.co/3YfOfWPpfZ

📍 Still here all day today and tomorrow - stop by booth PB16

[TuringData @TuringData_IO]

Tune in live in a few hours 🎙️ #TuringData VP Nikhil Madan joins Asia Tech Podcast today at 2:40 PM SGT.
We’ll explore where modern data infrastructure is headed and why it’s becoming a core driver of competitive advantage in the AI era.
👉 Live stream: https://t.co/3YfOfWPpfZ