I took some extra time today to really read the Q2 Earnings transcript of @BoostRunGPUs earnings and I think it is worth reiterating the key messages that I took away. It is rare to hear a CEO and CFO go this deep into the technical stack on an earnings call. Most management teams in AI infrastructure talk in megawatts, backlog, and slide decks. $BRUN spent its first ever public call talking about FP4 validation runs, NVIDIA committee approvals, and the ratio of contract value to capex. A mix of technicals and financial engineering. Management does not talk that way unless they built the thing themselves. Five moments from the Q2 call, and why each one is load bearing for the thesis: --- 1. "This isn't just, 'Hey, we passed this.' This is a pounding on the network, redlining it to the fullest, and ensuring 95% performance guarantee of the capabilities of the network and the entire stack." (Karos, on NVIDIA Exemplar certification, which he described as rigorous FP4 and FP8 performance testing) What it means. Exemplar functions as a statistically validated performance floor across the full stack, network included, at the two precisions inference actually runs on. Calling it a logo undersells what it took to get. Karos then named the reason it matters commercially: "there's some serious frustration about the quality of the compute that's being handed off." Why it is central. The dirty secret of the neocloud sector is that a large share of rented GPU capacity underdelivers against spec, and the customer only finds out after the handoff. $BRUN is selling a measured guarantee against that. At the time of certification on B300, the list was Oracle Cloud and Boost Run. A CEO who can describe the test protocol unprompted is a CEO whose company actually ran it. It clearly shows he is in the weeds of deals and the technicalities of deployment. --- 2. "Nothing goes into production at Boost Run, and I mean nothing on cluster size, unless it has passed that rigorous architecture review board, which is an actual committee at NVIDIA." What it means. Every cluster is gated by an external technical committee before a customer touches it. Karos tied this directly to why lenders get comfortable: "it's probably important to the primary banks and other institutions who are involved that there's fluidity and no problems." Why it is central. This is certification functioning as collateral quality. A credit desk underwriting a $200M equipment tranche is underwriting the probability that the hardware performs and the contract pays. NVIDIA's review board is a third party doing that diligence for free, every deployment. It is the mechanism connecting the technical story to the financing story, and Karos volunteered it without being asked. --- 3. "Say we have a $500 million contract with $250 million of CapEx. If a customer is putting down 15%, that's actually 30% because the TCV is two to one versus a CapEx." What it means. The CEO worked a capital structure example live on the call. Prepayment is quoted against total contract value, but it funds capex, so a 15% deposit on a 2:1 TCV to capex contract covers 30% of the hardware bill on day one. Reported average prepayment is 22% of TCV, on 100% of contracts. Why it is central. This single ratio is the answer to the loudest bear question on the name, which is how a company this size funds a $4B to $5B procurement. Roughly 30% arrives from customers before a loan draws. Most CEOs would have handed that question to the CFO. He did the math himself, in real time. --- 4. "A lot of times you will hit walls without having proper debt structure because you are not going to see things financed that are 'on-demand.'" What it means. Lenders will not write a loan against capacity with no contracted payment stream behind it. On demand GPU capacity is spot pricing and utilization hope, secured by an asset depreciating fast. To a credit committee that is an equity bet, so they decline it. Sector wide. Why it is central. On demand is the highest priced shelf in the store. $NBIS put short term capacity at roughly double the multi year contracted rate per MW. Because lenders will not fund it, only operators using their own cash can sell into it. $CRWV carries roughly $35B of debt with capacity pledged against contracts. $BRUN self funded from day one, collects a deposit on every deal, and already runs more than 12% of revenue through short term on demand. Karos manages capacity duration the way a rates trader manages a curve, because that is literally his prior job. The moat is the capital structure, and it does not appear on any screener. --- 5. "Four sites come online faster and in parallel, using a templatized design that has been battle tested across hardware generations and locations." What it means. Given the choice, management takes four 25MW sites over one 100MW site every time. Same capacity, delivered sooner, in parallel, with the failure of any one site contained. Why it is central. Every gigawatt campus story in this sector is a story about a date three years out. $BRUN is optimising for time to revenue instead of headline scale, and the new colo announced the morning of the call was already powered with no PPA or licensing risk outstanding. Six sites live, three more in six months, 253MW accessible. The open question is how many times a year they can run that playbook, not whether they can. --- Boost Run's C suite: Karos ran electronic trading at Galaxy Digital as an MD and exec committee member, and before that co founded and ran Blue Fire Capital, operating across six countries and thirteen data centers. CFO Erik Guckel holds a PhD in Chemical Engineering from Illinois alongside a Chicago MBA. COO Harry Georgakopoulos wrote a quantitative trading textbook for Palgrave and holds an MS in Financial Mathematics from Chicago. The CIO built ultra low latency trading networks across four continents. $CIFR: CEO Tyler Page began as an attorney at Davis Polk, then Lehman, Goldman, Guggenheim, Stone Ridge, NYDIG, and business development at Bitfury. COO Patrick Kelly came from Citadel, D.E. Shaw, and Magnetar, and co founded Stone Ridge. CFO Greg Mumford was a digital assets investment banker at KBW. That is an elite finance team with no engineering degree anywhere in the C suite. $DGXX: CEO Michel Amar holds a degree in accounting and business management, with a career spanning branded fashion, consumer electronics through a Brookstone partnership, and blockchain. The company appointed its first CTO in December 2025, Jag Jeyapaul, who is genuinely credentialed out of Oracle, Equinix, and VeriSign. Worth saying plainly: the technical depth is real, it is just eleven months old and it sits below the people who present on the call. BoostRun's management team comprises a healthy mix of technical experts and financial experts, with the CEO being a master of both. It's really quite different from what we see compared to other neoclouds, and this is central to the thesis. The strength of the company always stems from leadership. And BoostRun's leadership team is one of the best I have came across so far. I'm long af $BRUN