The AI factory company. We are on a mission to accelerate the abundance of energy and intelligence.
Joined January 2019
- Tweets 1,045
- Following 272
- Followers 15,355
- Likes 999
362 Photos and videos
A major breakthrough from @cartesia: Sonic-3.5 and Ink-2 are now the #1 streaming models for TTS and STT. ๐ฃ๏ธ
Trained on our infrastructure, these are exactly the kinds of category-defining AI models we built Crusoe to power.
Try it out: ๐ play.cartesia.ai
#VoiceAI #AIinfrastructure
Jun 15
We released Sonic-3.5 and Ink-2, the #1 streaming models for text to speech and speech to text you can use in your voice agents today.
New architectures enable new frontiers for speed and quality.
We're now the only provider to have #1 models for both speaking and listening.
8
1,095
New @BrookingsInst research on counties receiving their first large data center:
๐ Information-sector employment 22%
๐ Construction employment 11%
๐ Private employment 4%โ5%
Read more: ๐ brookings.edu/articles/new-eโฆ
4
789
Weโre teaming up with @HPE and @omni_ventures to host From Pilot to Production: How to Scale in Robotics โ a @deeptechweek gathering for builders in robotics and physical AI.
๐
June 24
๐ 6:00 p.m. PDT
๐ San Francisco
RSVP: ๐ luma.com/deep-l66l
#DeepTechWeek #PhysicalAI
6
561
As we continue building in Abilene, we love seeing neighbors win.
โOur guest countโs up 20% YoY.โ โ @PeriniRanch restaurateurs on how AI infrastructure is transforming their corner of West Texas.
โถ๏ธ More from @OpenAI: youtu.be/6QoMzZ8FdTA?si=r3VNโฆ
#AI #Stargate #AIinfrastructure
1
8
1,659
Crusoe retweeted
Jun 12
On y va ! ๐ซ๐ท
@CrusoeAI is a track sponsor at the RAISE Summit Hackathon โ Paris, July 4โ5.
Build with open models on Crusoe infra, compete for exciting prizes ๐
Devs, you'll want in on our track ๐
cerebralvalley.ai/e/raise-suโฆ
2
4
622
Inference challenges at scale? Probably not a GPU problem. ๐
Most inference bottlenecks aren't what they look like.
New guide on designing inference for scale, from memory architecture to KV-aware routing: crusoe.ai/resources/learn/deโฆ
#AIinference #AIinfrastructure
1
5
922
We joined @HOFCapital, @CloudflareDev @ArklexAI to host SGLangโs #NYTechWeek Happy Hour Lightning Talk exploring AI inference in finance.
Our AI Research Manager Omri Berkovitch spoke on efficiently running large-scale inference in production. ๐ธ
#AIinference #FinanceAI
1
10
1,244
What does it take to build an AI infrastructure company from the ground up?
Our Co-Founder CEO @ChaseLochmiller joined @gracegongGG to share lessons learned from nearly 8 years of scaling Crusoe.
Full conversation: ๐งyoutube.com/live/dEi6m_Q1Y3g
Jun 2
@ChaseLochmiller โ Co-Founder & CEO of @CrusoeAI โ on Venture with Grace, sharing what could become AIโs biggest constraint.
As AI models continue to scale, the next bottleneck may not be compute โ it may be energy.
Chase discussed how the path to increasingly powerful AI systems depends on access to abundant, reliable power and the infrastructure required to use it efficiently. This shift is already reshaping how data centers are designed, built, and operated.
His key point: the future of AI will be determined not just by advances in models, but by the energy systems that power them.
Episode is live now โ a fascinating conversation on AI infrastructure, energy, and the next phase of scale. Link in bio.
#AI #Energy #DataCenters #ComputeInfrastructure #ArtificialIntelligence #TechTrends
3
10
6,544
Crusoe is approaching 5 GW of contracted AI infrastructure โ with a 40 GW pipeline behind it.
That rivals the entire data center capacity of Northern Virginia, one of the world's largest data center markets, in 2025.
Grateful to every customer who has trusted us to build alongside them. ๐ crusoe.ai/resources/newsroomโฆ
10
15
185
104,464
For the second consecutive year, Crusoe partnered with Protect Our Rivers for their annual Van Bibber Creek cleanup.
Our Arvada volunteers removed nearly 800 lbs of trash! ๐๏ธ
More on how we support local water stewardship:
๐ crusoe.ai/about/impact?utm_cโฆ
8
1,353
You've watched a multi-week training run die because one node went sideways.
On June 30, two Crusoe engineers are getting into how the largest AI teams actually run Slurm and Kubernetes at scale โ the failure modes, the architecture decisions, the trade-offs that don't make it into the docs.
No fluff. Real scenarios.
๐
June 30 | 10 a.m. PT
Sign up here: bit.ly/4aiGCbt
#MLOps #DistributedTraining #AIorchestration #HPC
7
1,581
NIXL workloads on our @nvidia HGX H200 and HGX B200 instances were driving unexpected memory issues.
Root cause: NIXL registers every GPU buffer on all 8 NICs by default. With UCX's relaxed ordering override, that's a 16x firmware page multiplier vs. NCCL.
We surfaced it to NVIDIA. It landed as an upstream fix in NIXL 1.2 UCX 1.21. Up to 75% less host memory overhead.๐คฏ
Full writeup: crusoe.ai/resources/blog/howโฆ
#NIXL #NVIDIA #InferenceInfrastructure #MLInfra #GPU
1
2
17
2,377
Building the future of AI starts with great people.
New hires and interns joined our Co-Founder and CEO @ChaseLochmiller in SF for a Q&A on scaling Crusoe during a historic moment for AI infrastructure.
Join our team of innovators: โก๏ธ
crusoe.ai/careers?utm_campaiโฆ
#CrusoeCulture #AI #AIInfrastructure
1
8
942
Denver: join us tonight at 6 p.m. MDT for the unofficial @CVPR 2026 afterparty, co-hosted with @lmsysorg, @PhiloLabs_AI, and @reactorworld.
RSVP: ๐ luma.com/sglang-1d8k?utm_camโฆ
#CVPR2026 #SGLangCommunity #VLM
1
8
817
London didnโt disappoint.
Thanks to our speakers from @SpAItial_AI, @poolsideai, and @jetbrains for an evening of insights on taking research from prototype to production.
Proud to support the builders shaping the future of AI across the UK. ๐ฌ๐ง
#CrusoeAITalks #ML #AI
3
9
695
NVIDIA Nemotron 3 Ultra now in Crusoe Intelligence Foundry.
550B params, 55B active. Hybrid Transformer-Mamba MoE. Built for long-running agents โ planning, tool use, orchestration across hundreds of turns.
5x faster inference. Up to 30% lower cost. Up to 1M context.
Try the model โ bit.ly/3RPZDM7
@NVIDIA #Nemotron #AgenticAI #OpenWeights #AIInference
2
14
827
Crusoe has signed a ~750MW power agreement with Bergen Engines to bring reliable, on-site baseload power to our AI data center campuses across the United States.
Read the announcement: ๐ bergenengines.com/bergen-engโฆ
2
20
1,726
Loading: another summer of growth, innovation, and leadership.
Crusoeโs Summer 2026 internship class by the numbers:
๐ 73 interns
๐ซ 51 universities and colleges
๐ก 7 different teams
Stay tuned for more snapshots from our #Classof26.
#CrusoeCulture #CrusoeInterns
2
7
971
A 1,024-GPU cluster fails every 8 hours on average.
When one node goes down during distributed training, the other 1,023 sit idle. Every Crusoe node runs up to 30 burn-in tests before it ever reaches you โ DCGM diagnostics, NVLink bandwidth, NCCL collectives over IB, storage latency under real checkpoint patterns.
What we test and why: bit.ly/4u5CqTp
#MLOps #DistributedTraining #GPU
5
724
Computex 2026 is a reminder that Crusoe builds in lockstep with NVIDIA's roadmap at every layer of the stack.
โ Crusoe is planning to deploy NVIDIA Vera CPUs โ 88 Olympus cores, 1.2TB/s memory bandwidth, purpose-built for agentic AI at factory scale
โ Crusoe is adopting NVIDIA DSX OS components for AI factory lifecycle management, health automation, and multi-tenant operations
Full stack. NVIDIA-native. Built for what's next.
โก๏ธ Check it out: blogs.nvidia.com/blog/ai-cloโฆ
14
991