Inference
Microsoft announces the Maia [---] chip, a powerful new silicon workhorse designed for scaling AI inference. This development signifies a significant step in specialized hardware for AI, potentially impacting the landscape of AI computation.
About Inference
AI Inference is the process of deploying trained AI models to make predictions or decisions on new data.
Insights
- inference creators is up 35.2% from the previous month.
- inference engagements is up 99.24% from the previous week.
- inference mentions is up 21.03% from the previous month.
- inference posts created is up 32.98% from the previous week.
Engagements: [---------] (24h)
Engagements 24-Hour Time-Series Raw Data
Current Value: [---------]
Daily Average: [---------]
[--] Week: [----------] +99%
[--] Month: [-----------] +29%
[--] Months: [-------------] +320%
[--] Year: [-------------] +183%
1-Year High: [----------] on 2026-01-28
1-Year Low: [-------] on 2025-04-19
Engagements by network (24h): TikTok: [-----] News: [-----] Instagram: [---] YouTube: [-------] Reddit: [-----] X: [---------]
Mentions: [-----] (24h)
Mentions 24-Hour Time-Series Raw Data
Current Value: [-----]
Daily Average: [-----]
[--] Week: [------] +15%
[--] Month: [------] +21%
[--] Months: [-------] +29%
[--] Year: [-------] +410%
1-Year High: [------] on 2025-12-27
1-Year Low: [---] on 2025-03-16
Mentions by network (24h): TikTok: [--] News: [--] Instagram: [--] YouTube: [---] Reddit: [---] X: [-----]
Creators: [-----] (24h)
Creators 24-Hour Time-Series Raw Data
[-----] unique social accounts have posts mentioning Inference in the last [--] hours which is down 5.40% from [-----] in the previous [--] hours
Daily Average: [-----]
[--] Week: [-----] +11%
[--] Month: [------] +35%
[--] Months: [------] +48%
[--] Year: [-------] +201%
1-Year High: [-----] on 2025-12-27
1-Year Low: [---] on 2025-03-16
The most influential creators that mention Inference in the last [--] hours
| Creator | Rank | Followers | Posts | Engagements |
|---|---|---|---|---|
| @Supermicro | [--] | [------] | [--] | [---------] |
| @ordinarynullll | [--] | [---------] | [--] | [-------] |
| @nvidia | [--] | [---------] | [--] | [-------] |
| @CrusoeAI | [--] | [------] | [--] | [-------] |
| @CoreWeave | [--] | [------] | [--] | [------] |
| @MiniMax_AI | [--] | [------] | [--] | [------] |
| @tunemusicalmoments | [--] | [---------] | [--] | [------] |
| @firstadopter | [--] | [------] | [--] | [------] |
| @forloopcodes | [--] | [-----] | [--] | [------] |
| @rohanpaul_ai | [--] | [-------] | [--] | [------] |
Sentiment: 95%
Sentiment 24-Hour Time-Series Raw Data
Current Value: 95%
Daily Average: 87%
[--] Week: 84% -3%
[--] Month: 94% +10%
[--] Months: 94% +9%
[--] Year: 94% +10%
1-Year High: 99% on 2025-03-02
1-Year Low: 30% on 2025-11-08
Most Supportive Themes:
- Decentralized AI Clouds and Inference Engines: (40%) Focuses on making AI inference more accessible, affordable, and verifiable through new engines and decentralized platforms.
- AI Inference Cost-Effectiveness and Reliability: (30%) Companies are working on making AI inference more reliable and cost-effective, with a focus on verifiable AI outputs.
- AI Inference Hardware and Infrastructure: (30%) Discussions around specialized AI chips, distributed computing networks, and the infrastructure required for large-scale AI inference.
Most Critical Themes:
Top Inference News
Top news links shared on social in the last [--] hours
Showing a maximum of [--] news posts without a LunarCrush subscription.
"A flaw in using pretrained protein language models in proteinprotein interaction inference models Nature Machine Intelligence The usage of pretrained protein language models (pLMs) is rapidly growing. However Szymborski and Emad find that pretrained pLMs can be a source of data leakage in the task of proteinprotein interaction inference showing inflated performance scores"
News Link @SpringerNature 2026-02-13T19:00Z 2.1M followers, [----] engagements
"Microsoft announces powerful new chip for AI inference TechCrunch Maia comes equipped with over [---] billion transistors delivering over [--] petaflops in 4-bit precision and approximately [--] petaflops of 8-bit performance a substantial increase over its predecessor"
News Link @TechCrunch 2026-01-27T17:00Z 9.7M followers, [---] engagements
"The Gaping Hole In Todays AI Capabilities What if we could eliminate the rigid distinction in AI between training and inference enabling AI systems to continuously learn the way that humans do"
News Link @Forbes 2025-03-23T21:39Z 20.3M followers, [----] engagements
Top Inference Social Posts
Top posts by engagements in the last [--] hours
Showing a maximum of [--] top social posts without a LunarCrush subscription.
"Cango just had a $75.5 million capital injection Ive been following Cango for a while and things just got a whole lot more exciting - they just had a $75.5 million capital injection This comprised a $10.5 million investment from EWCL as well as personal investments from their chairman and director. Cango is strategically focusing on AI inference (not overcrowded model training) enabling small and mid-sized mining farms to transition into AI compute nodes to build a flexible distributed network. Cango has also set up U.S. subsidiary EcoHash in Dallas and brought in former Zoom infra lead Jack"
X Link @conorfkenny 2026-02-17T10:26Z 46.2K followers, [----] engagements
"Karpathys 243-line GPT is the "Mathematical Truth" of our era. Its a beautiful reminder that intelligence at its core is just a specific arrangement of matrix multiplications and attention heads. But theres a reason we don't run production on [---] lines of pure Python. The other [------] lines in a stack like SGLang exist to fight the physics of the hardwareFlashAttention KV cache management and CUDA kernels. You read Karpathy to understand the soul of the model; you use SGLang to give it a high-performance body. New art project. Train and inference GPT in [---] lines of pure dependency-free"
X Link @GenAI_is_real 2026-02-17T09:05Z [----] followers, 44.6K engagements
"GM Builders on CT 🏗 🤍 Happy Taco Tuesday 🌮 @0G_labs just expanded into the Ledger wallet app meaning OG holders can now move assets with hardware level security while still benefiting from a fast high performance L1. Stronger custody always strengthens confidence. Meanwhile @dgrid_ai is making decentralized AI inference practical. One unified API multiple models lower costs and reliable uptime. Builders get flexibility without depending on expensive centralized providers. And @permacastapp is preserving what Web3 often loses: context. By keeping content permanent and verifiable it protects"
X Link @0xZiggy_eth 2026-02-17T06:02Z 11.3K followers, 22K engagements
"GPT-OSS casually demonstrated that OpenAI can train on like 60-100T tokens. Ofc they can generate many T of highly enriched synthetics. Given their inference volume it makes zero sense to not overtrain. For all the "smell" [---] is likely smaller than open models that trail GPT-5. @Teknium @teortaxesTex Even the original GPT-4 wasn't 2T dense. I think GPT-5 is closer to 400B total and just inflates the price massively. @Teknium @teortaxesTex Even the original GPT-4 wasn't 2T dense. I think GPT-5 is closer to 400B total and just inflates the price massively"
X Link @teortaxesTex 2026-02-17T06:01Z 57.2K followers, 11.1K engagements
Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing