Observed Signal · Oct 20, 2026 · Podcast / Interview · Source: AINews swyx · Impact: 3/5 · Sentiment: Positive
SF Compute's Evan Conrad on Programmable Data Centers
In a podcast interview at AI Engineer NYC, Evan Conrad, CEO of SF Compute, discusses the evolution of compute trading markets. SF Compute has shifted from a simple GPU subleasing model to providing full 'physical settlement'—delivering usable compute rather than cash-settled derivatives. They've built the operational capacity of a cloud to handle the complexities of hardware, networking, and software. The company operates an order book that matches buy and sell orders with multiple constraints, despite low overall market liquidity. They have closed several hundred million dollars in volume and are onboarding new clusters. Key features include on-the-fly InfiniBand fabric subdivision, making data centers programmable. Conrad emphasizes the importance of neutrality and independence to avoid market manipulation, positioning SF Compute as a 'Switzerland' in the compute market, unlike larger players like NVIDIA or hyperscalers.
SF Compute is pioneering a new market for compute trading with physical settlement, which could lower financing costs and increase efficiency in the AI infrastructure ecosystem. This aligns with emerging trends in commoditizing compute and creating financial instruments for GPU resources, significant for the AI and cloud industries.
Track NVIDIA Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- SF Compute has closed several hundred million dollars in volume.
- SF Compute has moved 5% of its total capacity to the order book.
- SF Compute allows on-the-fly subdivision of InfiniBand fabric.
- SF Compute has shipped new features for platform-as-a-service companies.
- SF Compute is building a physical settlement model for compute trading.
Connected Companies & Entities
7 Entities mapped“NVIDIA is obviously like the 800-pound gorilla in the room. Why didn't they make SF Compute?”
“You met Oliver from SemiAnalysis, who is basically initiating coverage on compute markets is my read on this.”
“This is also why Crusoe and CoreWeave and Lambda, why the hyperscalers can't make it, why Base 10 and Fireworks and Modal can't make it beca...”
“This is also why Crusoe and CoreWeave and Lambda, why the hyperscalers can't make it, why Base 10 and Fireworks and Modal can't make it beca...”
“This is also why Crusoe and CoreWeave and Lambda, why the hyperscalers can't make it, why Base 10 and Fireworks and Modal can't make it beca...”
“This is also why Crusoe and CoreWeave and Lambda, why the hyperscalers can't make it, why Base 10 and Fireworks and Modal can't make it beca...”
“We're not in the data center business sort of, um, what we do work with is, so we offer supercomputing as a service.”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Startup Prices AI Compute, Eyes CME Compute Futures
Silicon Data, a startup building a reference price for GPU rental and an index for AI compute, closed a $30 million Series A. The company aims to serve as the pricing benchmark that a Wall Street futures contract could settle against and plans to launch compute futures trading on the CME on October 5, pending regulatory approval. The topic was discussed on TechCrunch’s Equity podcast with Steve Hou, Silicon Data’s head of research, about trends in the AI buildout and data center activity.
Tour of AI Data Center Highlights Infrastructure Supply Chain
CNBC toured CoreSite’s NY3 AI-optimized data center in Secaucus, New Jersey, and reported how hyperscaler demand is driving a surge in AI-focused data center infrastructure. CoreSite (owned by American Tower) and facility designers like Prime Data Centers described power sourcing, cooling, fiber connectivity, chips, and networking as the five critical buckets for modern AI centers. The story cites major vendors and customers — Amazon, Alphabet, Meta Platforms, Microsoft, OpenAI, Anthropic — and supplier companies benefiting from the buildout including Nvidia, Broadcom, Corning, GE Vernova, Eaton, Schneider and Dover. The article includes recent vendor financials and strategic deals (Corning fiber supply deals, Broadcom–Meta partnership, GE Vernova orders), and emphasizes solutions such as liquid cooling and fuel cell power as responses to rising energy and thermal demands.
CME and Silicon Data Launch Compute Futures Market
CME Group and Silicon Data announced a joint venture to launch a new compute futures market enabling traders and AI builders to hedge the rising cost of computing capacity. Contracts will be settled against Silicon Data’s GPU price indexes and daily GPU rental benchmarks, allowing buyers to lock in prices for on‑demand GPU capacity. Silicon Data also publishes RAM and memory indexes and projections for GPU rental prices. The move creates standardized reference pricing for GPU markets, which industry participants say have lacked reliable benchmarks amid surging demand and sharply higher memory prices in Q1 2026 driven by AI investment and hyperscaler spending.
Receive all future signals for NVIDIA & competitors in your briefing
Never miss material M&A deals, leadership changes, account wins, or product rollouts. Polaris7 monitors 20,000+ primary sources 24/7, delivering verified telemetry directly into your executive briefing.