Come to @TomorrowXSummit in Austin November 17-18. We have Jensen Huang, Lisa Su, Lip-Bu Tan, Marc Rowan, David Solomon, Jon Grey and so many more great speakers - I think we are approaching 100 CEOs across three stages at this point. I get emails every day asking for an invitation - this is your invitation! Anyone can sign up but note there will be a security screening and super-tight security at the stadium.

This is my conversation with Michael Kagan, the founder of Mellanox and current CTO of NVIDIA We speak about his life, the progression of networking over the span of his career, and how the work that NVIDIA is now doing to build "Planet scale" computing I hope you enjoy the conversation as much as I did 00:00 Highlights 01:16 Growing Up in Russia and Moving to Israel 03:42 Intel, the PC Revolution, and a Passion for Technology 06:14 Sponsor: Liquid 06:33 Lessons from Intel: Software and Networking 10:10 Founding Mellanox and the Vision for Cloud Computing 14:34 From Supercomputers to the Cloud 17:46 How GPUs Sparked the AI Revolution 20:52 Sponsor: Compute Exchange 21:14 NVLink and the Rise of Rack-Scale Computing 24:57 The Engineering Challenges Behind NVL72 27:39 Designing Hardware for Future AI Workloads 29:43 BlueField DPUs and the AI Factory 32:55 Making AI Agents Safer 35:37 What Mellanox Could Have Done Differently 37:03 NVLink, InfiniBand, and Spectrum-X 39:24 Connecting AI Factories at Planetary Scale 42:42 Keeping Massive AI Systems Running Through Failures

Gavin Baker says the compute market is now in contango per Silicon Data and his portfolio companies are shocked at how much more they have to pay for clusters in 2027 “You know, Silicon Data, which I do think is, my firm is an investor in them, but I do think they’re the most authoritative data source. Carmen Li, the founder, stood up at GTC and said that the compute market is now in contango. It’s not just that the spot market is tight. If you want a cluster a year from now, you have to pay a lot of money." "And then this is maybe another discontinuity between public and private markets. But just the number of companies we have in our portfolio who are like, okay, we got a cluster of 4,000 Blackwells, and we got it in May or June ‘26, and we wanna bring on another cluster of 4,000 in September 2027, and they’re shocked by how much more they have to pay in the future." "And it’s not like everybody is unaware of this. Everybody’s aware of the supply that’s coming on. But even with all of that supply coming on, it is hard for the companies in our portfolio to find compute even well into ‘27. So for, you know, all the commodity experts who are focused on backwardation and what that implied, I got news for you. We’re not in backwardation anymore.”

An AI cloud operator can sell years of compute capacity and still struggle to finance delivery. The squeeze I’m watching: customer prices fixed before funding is secured. A thread on the financing gap behind AI compute contracts.

Renting Nvidia’s H100 costs 45% more than it did in December. It went on sale in 2022, and two newer generations have come out since. Yet, demand for compute is so far ahead of supply that older GPUs are getting more expensive.

@carmenli , CEO of @silicondata and @computeexchange , joins the show to explain how compute is becoming a brand-new asset class. A former energy options market maker at DRW who later ran global data integrations at @Bloomberg , Carmen was pre-seeded by DRW founder Don Wilson to build the GPU industry benchmarks. Silicon Data is the referee "the Bloomberg for compute" while Compute Exchange is the liquidity layer where GPUs, tokens and servers change hands. We get into why token price indexes are falling, recent token & model consumption trends, why GPU residual values are above accounting depreciation schedules, and how neoclouds, their lenders, and compute users can hedge GPU exposure to reduce business volatility. • "The A100 has been around for six years at this point, and we still see three-year contracts signed on A100s." Residual values are still hovering above five-year straight-line depreciation. • “we'd definitely be looking at 30 trillion of [GPU] derivatives every year." Topics: Compute as an asset class, GPU indices, CME compute futures, GPU depreciation and useful life, neoclouds, token pricing, AI chip supply chain, 💡 This episode is presented by: @OxfordDataPlan –The Home of Alternative Data *Not Investment Advice. Highlights: (0:00) Doug introduces Carmen Li, CEO of Silicon Data and Compute Exchange (0:20) From energy options market maker at DRW to Bloomberg Data, and why AI startups' swinging GPU costs made compute an industry-wide hedging problem (4:09) How DRW founder Don Wilson pre-seeded Silicon Data, and building products with no product or sales team (6:10) Two sides of the coin: Silicon Data as "the Bloomberg for compute" and Compute Exchange as "the Nasdaq for compute" (7:49) Licensing indices to CME for H100 and B200 compute futures, and four ETFs filed before the futures even trade (10:22) What makes an index the source of truth: tracking supply-demand, resisting manipulation, staying neutral (12:57) Normalizing 50 B200 server configurations, and what earns a premium or a discount to the benchmark (16:04) Sizing the market: from 1–3 trillion in spot compute to a potential 30 trillion in derivatives notional (18:13) Who hedges first: neoclouds, hyperscalers and the financiers holding GPUs on their balance sheets (21:06) Michael Burry's depreciation bear case vs. the physical market: A100s still signing three-year contracts after six years (22:49) Why older chips stay productive: cooling, floor weight, wait times and the second-hand market (26:20) 300+ neoclouds and why more data sources make an index harder to manipulate (28:48) B200s at ~$5 per GPU hour, the implied forward curve and the shift into contango (31:18) Why an index provider can't give forward guidance, or own the assets it measures (33:59) Token indices on the Bloomberg Terminal, and a leading indicator for AI revenue (37:03) Why token prices are falling: workflows moving to open-source and open-weight models (38:51) Token pricing has no transparency, and why customers will demand control (41:49) Where to invest in the AI stack, lessons from the internet era and why "AI is a gigantic baby" (45:03) Homegrown silicon, fab capacity and why a hundred chip designers can't create a bubble (48:56) Why compute futures enable price discovery, not a bubble (49:29) Running a data company and a marketplace at the same time (52:47) Carmen's lesson for entrepreneurs: pick something you love (54:08) The vision: the liquidity provider and the referee for every layer of the cake Research Sponsors: @Fiscal_ai - Delivering Modern Financial Data Infrastructure. 15% discount with code PITCH @AlphaSenseInc Grade AI @CarbonArcAI Infrastructure for the AI Economy Pitch The PM Links: 📩Early episode drops episode feedback on Substack Daily industry chatter and market color on LinkedIn Carmen Li Links: Carmen LI on LinkedIn: Silicon Data: Compute Exchange: