--- title: "Luxor's Compute Newsletter | Issue #2" description: "Luxor's Compute Newsletter: Weekly coverage of the AI compute market: the news, the numbers, and the infrastructure economics that matter." image: https://go.luxor.tech/hubfs/luxor-compute-newsletter-issue-002.png --- [Skip to content](https://go.luxor.tech/compute-newsletter/002#main-content) [![Luxor-1](https://go.luxor.tech/hs-fs/hubfs/Luxor-1.png?width=150&height=32&name=Luxor-1.png) ![Main-1](https://go.luxor.tech/hs-fs/hubfs/Main-1.png?width=150&height=32&name=Main-1.png)](https://luxor.tech/) Open main navigation Close main navigation Open search Close search This is a search field with an auto-suggest feature attached. - There are no suggestions because the search field is empty. Luxor's Compute NewsletterWEEKLY Weekly coverage of the AI compute market: the news, the numbers, and the infrastructure economics that matter. August 19, 2026  ·  4 min read Get every issue in your inbox No spam. Unsubscribe anytime. Compute Price Pulse 1-yr contract · median $/GPU-hr | H100 | $2.74   ▲   7.0% since last week | | --- | --- | Available on [Hashrate Index](https://data.hashrateindex.com/gpu-price-data/ornn-h100-index). Powered by [Ornn](https://data.ornn.com). The Lead ## Vera Rubin: What I'm Telling Operators Before They Buy Since GTC, Luxor's GPU hardware desk keeps fielding one question: should we buy Vera Rubin? Mike San Miguel's answer starts with the building, not the chip. NVIDIA approves your site before it allocates you the hardware, and the NVL72 ships as a single high-density rack drawing roughly 190 to 230 kW with direct-to-chip liquid cooling that is no longer optional. An air-to-liquid retrofit runs 12 to 18 months, and the racks are heavy enough that floor load ratings alone can stall otherwise-ready sites. The counterintuitive part: the expensive mistake isn't overbuying, it's waiting. Whatever you install today runs at close to full utilization, and a large share of inference doesn't need a facility-gated rack at all. If you can get your hands on a current-gen RTX PRO 6000, that gets you earning while competitors deliberate. The operators who win this generation won't be the ones who bought the fastest chip. They'll be the ones who got installed. [Read the full breakdown on Hashrate Index →](https://hashrateindex.com/blog/vera-rubin-what-to-know-before-you-buy/) --- The SPOTLIGHT ## Compute Becomes a Commodity: CME to List GPU Rental Futures CME Group is partnering with Silicon Data to launch the first regulated futures contracts tied to the price of AI compute, with trading set to begin October 5, pending regulatory approval. The two contracts settle against indexes tracking hourly rental rates for NVIDIA's H100 and B200 GPUs, and each represents a month's rent. The pitch: give buyers and sellers of GPU capacity what oil and power markets have had for decades, a public benchmark to hedge against, in a market where two companies renting identical hardware can pay wildly different prices with no way to know who got the better deal. --- LUXOR'S TAKE We watched this exact story play out in Bitcoin mining: once hashprice gave miners a public benchmark, hedging followed, and Luxor built a hashrate derivatives market on top of it. A tradable reference price for GPU rentals will do the same for compute, forcing unit economics into the open and rewarding the operators who lock in revenue instead of riding spot. [Read the full story on CNBC →](https://www.cnbc.com/2026/08/11/ai-computing-power-becomes-a-tradable-asset-class-as-cme-starts-futures.html) AI Hardware availability B300 and H200 in stock — start the revenue clock sooner When the neocloud math works, sourcing is the next question. Luxor has B300 and H200 GPU hardware available now. [Start a conversation →](https://luxor.tech/contact) Around the Industry *Throughline: the constraint on AI compute has moved from silicon to megawatts, and everyone from NVIDIA to Gartner to the space industry is repositioning around it.* [NVIDIA Starts Locking Up the Power Layer](https://blogs.nvidia.com/blog/securing-the-infrastructure-of-intelligence/) NVIDIA announced a partnership with SB Energy to secure land, power, and shell capacity at the roughly 4 GW PORTS-Pike campus in Portsmouth, Ohio. They've locked in a 20-year term, with OpenAI as the anchor tenant and an opportunity NVIDIA sizes at roughly $600 billion of its own compute through 2030. It follows reporting that the chipmaker has committed up to $3 billion to Lancium, the power developer behind the Stargate campus in Texas [(MarketScale)](https://www.marketscale.com/industries/software-and-technology/nvidia-bets-3-billion-on-power-infrastructure-as-ais-energy-crunch-reshapes-enterprise-procurement). When the company selling the chips starts writing checks for the megawatts, the message is hard to miss. [(NVIDIA)](https://blogs.nvidia.com/blog/securing-the-infrastructure-of-intelligence/) [Gartner: Data Center Power Demand Up 26% in 2026](https://www.gartner.com/en/newsroom/press-releases/2026-06-10-gartner-says-data-center-electricity-demand-to-grow-26-percent-in-2026) For site operators Gartner forecasts global data center electricity consumption reaching 565 TWh in 2026, up from 447 TWh last year, with capacity demand climbing from 104 GW to 132 GW. AI-optimized servers drive the growth, hitting 31% of data center power consumption this year and overtaking conventional servers by 2027. For anyone holding energized capacity, the takeaway runs in your favor: locking down megawatts is now the fight that decides who gets to scale. [(Gartner)](https://www.gartner.com/en/newsroom/press-releases/2026-06-10-gartner-says-data-center-electricity-demand-to-grow-26-percent-in-2026) [Orbital Data Centers: Frontier or Distraction?](https://www.datacenterdynamics.com/en/opinions/are-orbital-ai-data-centers-the-next-frontier-for-compute-infrastructure/) A DCD opinion piece argues the sudden interest in orbital compute says less about space than about Earth: power, grid connections, water, and permitting now gate deployments, and more than a quarter of the data center projects expected to go live in 2025 were delayed. Orbit is what an industry starts pricing when terrestrial constraints bind. [(Data Center Dynamics)](https://www.datacenterdynamics.com/en/opinions/are-orbital-ai-data-centers-the-next-frontier-for-compute-infrastructure/) --- From the HRI Vault [![GPU-as-a-Service Explained: The Business Model Behind AI](https://storage.ghost.io/c/5f/b9/5fb98cf7-67da-41b5-a184-b1b39752525f/content/images/size/w1200/2026/02/Hashrate-Index-GPUaaS-Blog-Banner.jpg)](https://hashrateindex.com/blog/gpu-as-a-service-business-model-ai/) [GPU-as-a-Service Explained: The Business Model Behind AI](https://hashrateindex.com/blog/gpu-as-a-service-business-model-ai/) How renting GPUs connects AI demand with infrastructure supply: the pricing models, what customers are actually buying, and the economics a provider needs to make the model work. A useful companion to this week's Lead if you're weighing owning versus renting capacity. [Read now →](https://hashrateindex.com/blog/gpu-as-a-service-business-model-ai/) AI Hardware availability Offtake is becoming a financed product. Luxor’s compute offtake listings show real-time availability — fill capacity or secure it. [Explore available compute →](https://luxor.tech/ai/compute) [![Luxor](https://go.luxor.tech/hs-fs/hubfs/Luxor-2.png?width=150&height=34&name=Luxor-2.png)](https://luxor.tech) [The full-stack compute company — serving Bitcoin mining and AI infrastructure.](https://luxor.tech) [luxor.tech](https://luxor.tech)   [![Luxor logo](https://go.luxor.tech/hs-fs/hubfs/Luxor-1.png?width=121&height=25&name=Luxor-1.png "Luxor logo")](https://luxor.tech) [![Luxor Technology](https://go.luxor.tech/hs-fs/hubfs/Main.png?width=1808&height=383&name=Main.png "Luxor Technology")](https://luxor.tech) The full-stack compute company — serving Bitcoin mining and AI infrastructure. Luxor Technology 2026 © is a US-based compute power company focused on Proof-of-Work Bitcoin mining and compute.