Jesus Rodriguez
@jrdothoughts
Followers
5K
Following
5K
Media
290
Statuses
15K
Co-Founder, CTO-CPO of @SentoraHQ (fmr IntoTheBlock), Co-Founder of @layerlens_ai, @faktoryai, @neuralfabric and The Sequence AI, Teaching at Columbia-Wharton
Joined September 2011
The Sequence Opinion #710: The Inference Cloud Wars: Speed, Scale, and the Road to Commoditization
0
0
0
RT @Firelightfi: Get a closer look at Firelight’s analytics dashboard. Currently being built to deliver powerful on-chain insights. 🔥 http….
0
38
0
The Sequence Knowledge #709: Explainable-by-Design: An Intro to Intrinsic Interpretability in Generative AI
0
0
0
This is arguably the most pragmatic analysis of DAT companies. Worth reading: The Boom and Potential Bust of Digital Asset Treasury Companies: Which Tokens are Most Exposed? by @admff492
0
0
2
RT @drfeifei: A picture now is worth more than a thousand words in genAI; it can be turned into a full 3D world! And you can stroll in this….
0
321
0
Don’t ignore geopolitics. U.S. export controls restrict China’s access to top GPUs. EU pushes for “digital sovereignty.”. Middle East nations stockpile AI compute. The GPU cloud race is as strategic as it is technical.
0
0
1
Examples:. CoreWeave: largest independent fleet, NVIDIA-aligned, Kubernetes-native infra. Lambda: AI developer UX, hybrid on-prem + cloud. Crusoe: low-cost energy edge, sustainability story. Nebius: Europe-based, full-stack AI infra, sovereign positioning.
1
0
1
Examples:. CoreWeave: largest independent fleet, NVIDIA-aligned, Kubernetes-native infra. Lambda: AI developer UX, hybrid on-prem + cloud. Crusoe: low-cost energy edge, sustainability story. Nebius: Europe-based, full-stack AI infra, sovereign positioning.
1
0
0
So, how do neo-clouds differentiate? Two paths:. Up the stack → AI platforms, model fine-tuning, MLOps, vertical SaaS. Down the stack → faster interconnects, better schedulers, greener/cheaper power, first-to-market with bleeding-edge GPUs.
1
0
0
So, how do neo-clouds differentiate? Two paths:. Up the stack → AI platforms, model fine-tuning, MLOps, vertical SaaS. Down the stack → faster interconnects, better schedulers, greener/cheaper power, first-to-market with bleeding-edge GPUs.
1
0
0
Hyperscalers aren’t sitting still:. Google: TPU v5. AWS: Trainium & Inferentia. Microsoft: Project Athena & AMD partnership.They’re bypassing NVIDIA dependency with custom silicon while leveraging vast integrated ecosystems.
1
0
0
But the success story comes with a challenge: commoditization. A GPU-hour is a GPU-hour. As supply catches up, price competition intensifies. $8–10/hr H100s in 2023 now rent for <$3 on some platforms. Margins will collapse unless value moves up or down the stack.
1
0
0
Why now?.Explosive AI workloads → LLMs, diffusion, scientific compute. Hyperscalers had GPU supply gaps. Lower-cost, flexible options created a long tail of demand. Neo-clouds stepped in where AWS/GCP/Azure fell short.
1
0
0
The origins are fascinating:.CoreWeave pivoted from crypto mining to AI compute. Crusoe used stranded energy to power data centers. Lambda evolved from selling GPU workstations to running full GPU clouds. Specialization is the differentiator.
1
0
0
The GPU cloud market is at an inflection point:.Neo-clouds like CoreWeave or Lambda are rising fast, hyperscalers are striking back with custom chips and massive scale. The future will hinge on who moves up the stack with AI services or down the stack with infra excellence.
1
0
0
I just published a very long analysis about the present and future of gpu-clouds(neo-clouds ): Details are below:.
1
0
1
Some technical notes about @aave v4: At @SentoraHQ we couldnt be more excited about the possibilities.
medium.com
One of the most impressive DeFi architecture of the new wave.
1
9
53