Attacking the critical path - medium voltage switchgear is one of the longest lead times in the AI infrastructure supply chain, so @CrusoeAI has vertically integrated the capability to manufacture this gear in-house through Crusoe Industries. This brings down our lead time by 40%. Crusoe is now one of only three approved manufacturers in the U.S. for this kind of ABB gas insulated switchgear.

Thoughtworks' frontier API bill hit $450K a month, up about 27% monthly. Their fix: rent frontier models for frontier work, run open-weight models for the rest, and benchmark real traffic on Crusoe Cloud. Worst-case 2028 forecast: down 72%. Details from @thoughtworks:

We're partnering with @Entirl on @StartupIreland's new AI Accelerator Programme, delivered by @dogpatchlabs. 📸 Crusoe's Paul Murtagh with leaders from @Amazon, @Anthropic, @Google, @Microsoft, @NVIDIA, @OpenAI + @Stripe at #AiSummitIE# in Dublin for the launch.

Crusoe and @Lancium hosted a Childress, TX, job fair near our 1.0 GW, 270-acre AI campus-in-progress. 👥 450 job seekers 💼 30+ positions filled Thanks @WorkforceSP + @DPRConstruction for supporting! More on how we’re investing in the future of Childress County: 🔗

Energy first, then compute. @ChaseLochmiller, co-founder and CEO of @CrusoeAI is building a lot of the data centers that power the age of intelligence. They just closed a round at a $31B valuation, about 8 years into a company that started as two guys burning flared gas in the Bakken oil field to mine Bitcoin. That wasn’t a pivot, btw. Chase was AI-pilled early. The bet was: if data and compute are the bottleneck, compute becomes an energy problem. So instead of bringing stranded energy to a market, bring the market to the energy. Bitcoin was the easiest way to turn electrons into money while he waited for AI demand to show up. It showed up on November 30th, 2022. He’s a learning machine. On track to be a professor, then a quant, then Everest, now running a 2,000+ person vertically integrated business with welders and AI researchers in the same company. A few things we get into: •Why Crusoe is an AI supermajor: upstream (power), midstream (data centers), downstream (AI cloud). Exxon famously doesn’t hedge oil because the stack hedges itself. Margins move around the stack; own the whole thing.

 •“Most startups die from suicide, not homicide.” We said the exact same thing at HubSpot. If someone else is chasing your opportunity, that’s a compliment.

 •Coopetition. He’ll partner with companies he competes with elsewhere.  Useful when you’re building infrastructure the entire industry needs.

 •The CEO journey isn’t linear. “As soon as I’m sick of hearing myself say something, we’re just scratching the surface.”

 •How he scales himself: a domain expert owns each vertical. Energy, data centers, manufacturing, AI cloud. He goes deep enough to get aligned on direction and understand how it connects to the rest of the business. Then he holds them accountable to execute. Alignment first, then trust with accountability.

 •Should every CEO hire a coach? Yes. His coach was as much therapist as coach, and it was critical for him and the company. We get into that more. •People scale linearly, companies scale exponentially (h/t First Round). When you start thinking about letting someone go, that was probably the right time.

 •Actively hunt bad news. Good news is nice, but the CEO’s job is to “run into the burning buildings,” find what’s broken, and fix it.

 •Selling the Bitcoin business to NYDIG while it was making money. One word: focus. •The data center backlash. “The industry has a marketing problem. The truth is on our side.”
 Timestamps: 0:00 – Cold open 1:10 – Intro 1:35 – Professor, quant, Everest, energy, data centers: is there a throughline? 2:43 – What people get wrong about the CEO journey 3:53 – Live on the infinite growth loop 4:44 – “My biz is becoming dizzyingly complex beneath me” 6:44 – Scaling yourself with domain experts 7:36 – Trust, DRIs, and not being hands off 8:33 – Cross-org execution and the weekly staff meeting 10:11 – “As soon as I’m sick of hearing myself say it” 11:37 – Keeping urgency at 2,000 employees; the Elon tension 13:12 – Books, mentors, and a coach 15:12 – Most startups die from suicide, not homicide 16:50 – Paranoia and coopetition 17:26 – The AI supermajor: upstream, midstream, downstream 19:47 – Propeller, Blue Energy, and energy first 23:24 – Flared gas, Bitcoin, and bringing the market to the energy 24:45 – Wait But Why, Andrew Ng, and compute as an energy problem 28:52 – November 30, 2022 30:19 – Selling the Bitcoin business: focus 32:43 – Polish is learned: public speaking and Shane Metcalfe 34:31 – Should every CEO hire a coach? Yes. 37:54 – Hire slow, fire fast, and reading the mosaic 41:09 – Ali Ghodsi vs. Dev Ittycheria on exec turnover 42:51 – AI researchers and welders: the collective genius 44:15 – The data center backlash and the marketing problem 48:53 – Thick skin and speaking truth to power 51:09 – Everest: the 2014 avalanche and the Khumbu Icefall 55:12 – The Balcony and the summit 57:36 – My takeaways

Congrats to the winners of our 10/17 Recursive Agents Hackathon in SF, and thanks to @cerebral_valley, @nvidia and @youdotcom for helping make it happen! 📸 Attendees built agents that can take a goal and run with it on Crusoe Managed Inference.

📸 Our SF team volunteered at Redding Elementary, Mission Creek Park and @FleetWeekSF's Career Mentorship Workshop with @handsonbayarea, @Googleorg, and @RecParkSF. Thanks to our volunteers and to the Downtown Volunteer Coalition for organizing!

Crusoe visited a Warrenton High School ag construction class to discuss job opportunities and career paths in Missouri’s construction industry. Thanks to our team and volunteers from @DigCastle and McCarthy Holdings for mentoring the next generation of builders. 🦺

Slurm: 2002, the bare metal era. Kubernetes: 2014, the container era. Different eras, different jobs. Mike Petersen on why, and how to run Slurm on Kubernetes. New AI Lab →

TOMORROW: What will it take to build the infrastructure powering the next era of AI? At #BloombergInvest# Dallas @ChaseLochmiller, Co-Founder & CEO of @CrusoeAI, joins @dinabass to discuss the surging demand for computing capacity, the Abilene-based Stargate Project and how Texas is emerging as a major hub for AI infrastructure. Tune in tomorrow at 10:45 AM CDT ⏯️

After 20 years exploring for and producing energy, nothing has impressed me as much as ITAIPU. As a young geologist I was fortunate to be awarded a Thomas J Watson Fellowship, which enabled me to travel the world studying the energy resources and systems of many countries. I have subsequently spent my career as an operator, explorer and student of energy - from geothermal in Iceland to coal in China to solar in Spain to wind in Argentina to oil and gas in Texas, I’ve seen a lot of different energy systems up close. The scale, history and potential of ITAIPU is singularly compelling. This is why the @CrusoeAI team is excited to partner in Paraguay to add value to this amazing energy resource through AI infrastructure development. A shared vision of investment, employment, technology, innovation, clean power and community benefit make this project extremely exciting and I look forward to realizing this vision together with our local partners.

Congrats to @GradiumAI on open-sourcing Phonon, a suite of text-to-speech models under 50 million parameters. Natural-sounding speech in five languages that runs locally on phones and laptops, no network needed. Proud that Phonon was trained on Crusoe Cloud.

Thanks to everyone we met at #NVIDIAGTC# Berlin this week! 📸 Moments from the conference, including our SVP of Product Management Erwan Menard speaking at Wednesday’s Build in Europe reception alongside panelists from @NVIDIA and Treble Technologies. #Berlin#

I’m glad to announce that @CrusoeAI is partnering with Araico to utilize excess clean, reliable and 100% renewable hydroelectric power to build cutting AI datacenter infrastructure in Paraguay. Supported by the U.S. government and Paraguayan government, this is a great opportunity to create value for the people of Paraguay by adding value to their existing resources, and an exciting new chapter of Crusoe’s energy-first international expansion strategy.

Fine-tuning takes a dataset. RL takes a judgment. On Crusoe Cloud, both paths produce LoRA adapters on the same open models and endpoints. Start with Serverless Fine-Tuning, graduate to Managed RL, and track every run in Managed MLflow. The full loop:

At today’s White House Science: A New Golden Age Summit, @ChaseLochmiller pledged $50M in Crusoe Cloud credits for compute and inference to the Genesis Mission Consortium, supporting researchers across 15+ agencies including DOE, NASA, NIH, and NSF. We’re proud to take part in supporting the scientific breakthroughs AI will make possible. Details here:

We’ll be in San Jose for @PyTorch Conference North America Oct 20 - 21. 🎤 Talks on Triton, MoE fine-tuning + model customization 🤖 Booth demos + challenges 🍻 AI Infra Night with @dstackai, @SGLang (@lmsysorg) Register: 🔗 #PyTorchCon# ​​#PyTorchFoundation#