Bg2 Pod

.@altcap just got the ultimate compliment on @theallinpodcast from @jason. "This would not have happened without you, Brad" — for turning Trump Accounts from an idea into reality and a hallmark achievement of the President’s main street agenda. “This is your legacy of everything you've done in your life. Lots of success and I've seen it up close and personal this is a million times everything you've done in your whole life. You'll be remembered for this.” 💐

Michael & Susan Dell, Brad Gerstner, and Ray Dalio are giving kids across America a jump start to build wealth! Get the facts about the historic week launching Trump Accounts: podcasts.apple.com/us/podcast/his…

Congrats to Brad Gerstner - @altcap - the Founder of the Invest America Act / @trumpaccounts & the relentless driving force for 4 years to ensure that every child in America gets a direct ownership stake in the American Dream! Founder Beast Mode! 🇺🇸🚀🇺🇸

President Trump says Elon Musk will likely donate SpaceX stock to TRUMP ACCOUNTS — and Trump will ring the opening bell for the New York Stock Exchange (and Nasdaq) from the Oval Office next week. I hope everyone got their kids or grandchildren signed up for Trump Accounts.. it is so good for their future. rumble.com/v7c5zco-elon-t…

Empowering families and communities. Micron is investing $250 million through @TrumpAccounts to support America’s future generations. Read what @altcap founder of @InvestAmerica24 has to say:

Thanks to @EdLudlow for having us on Bloomberg Tech yesterday to talk about our latest fundraise and the growing number of companies owning their open and specialized models.

Who remembers when Satya threw shade at CoreWeave on the @BG2Pod back in December 2024, when he referred to their relationship as a "one-time thing", and implied that Microsoft had managed to get caught up on AI infrastructure (after they were initially blindsided by the success of ChatGPT)? "That (the CoreWeave partnership) is a one-time thing, and then now it's all catching up.. so that was just more about trying to get caught up with demand." Fast forward to 2026 and Microsoft has not only increased the scope of its arrangement with CoreWeave, but also signed ~$50B+ in new deals across the entire neocloud spectrum ($17B with Nebius, $10B with IREN, tens of billions with nScale, multiple billions with Lambda, etc.). Absolutely diabolical that CoreWeave was seemingly better prepared for the growth in AI infrastructure demand than Microsoft, considering Microsoft funded the large majority of OpenAI's early research and had full exclusive access to its IP.. $CRWV $MSFT

"Lest we be overconfident in Silicon Valley, let's remember a small group of activists shut down supersonic technology, and all nuclear energy in this country. It's a disaster." @altcap explains why a data center moratorium would be "horrific" for America: "All of our GDP growth is coming from the fact that we are building data centers and driving productivity improvements in the economy." "A data center moratorium would thrust us straight into a recession and high unemployment." "Secondly, it would cede the entire global game to China. Overnight, we would lose to China in the global AI race. Which is not just about AI, it's about economic security, jobs, and national security." From his appearance on the show last month.

Tremendous respect for @GavinSBaker @altcap @_clarktang & their teams. One of the best pods of the year. SpaceX becomes hyperscaler #4 in record time -Effectively went from not focused on it to the #4 in roughly ~30 days, past Oracle and others, on the Google and Anthropic cloud deals alone -Google deal implies ~$50B revenue/GW, Anthropic ~$25B/GW, against a bank base case of $14B/GW -Added ~$29B of run-rate in a month, taking the multiple from ~100x trailing revenue to ~39x [shows how savy Elon & Gwynne are at striking deals] -Core edge is build speed: Colossus 1 in 122 days, and speed compounds because idle construction is pure labor cost on already strained workforce -AWS analogy holds loosely, but they enter a market where AWS, Azure, GCP, CoreWeave and Oracle already fight for the same dollars Terrestrial data center math -Freda Duan bottom-up: 325k Nvidia GPUs across Colossus 1 and 2, blended $5.33/hr, ~590MW facility power at 1.2 PUE for the Anthropic deal -Implies ~$25B revenue/GW against ~$29B/GW to build (ARK estimate of Colossus costs) -Payback near 1.2 years before opex, 55% IRR if the contract renews [on Colossus 1] -Elon flagged the Anthropic deals short term, and the $8/hr GB200 rate for Colossus 2 compresses as supply expands, which the bull case underweights -The $50B+/GW Google figure has no equivalent bottom-up build, it sits as a residual on the scatter plot [one thing I thought was interesting, though, by their own admission Google is likely buying optionality to be first on the orbital DCs IF they work and come online fast enough so anchoring to the same revenue/GW they signed at doesn't seem like a usable datapoint and is rather an outlier given it's a call option] Orbital compute -Fox puts orbital capex at ~$5B/GW versus $20-25B/GW terrestrially, a 5x cut on roughly half the bill of materials -Built off 100 tons and ~5MW of compute per Starship launch at a $250/kg target, versus ~$1,500/kg on Falcon today -Requires two-stage reusability, never demonstrated, with a second-stage return attempt later this year -Ignores satellite failure rates, radiative cooling mass and orbital power intermittency in LEO Nvidia share -Altimeter's own chart steps Nvidia 2026 to 2027E down across all three lenses: revenue 79% to 72%, GW share 64% to 46%, unit 50% to 32%, with the 2027 column flagged with a question mark. -GW share falls harder, 64% to a projected 46%, and unit share drops 47% to 32% -Revenue share holds up better than units because Nvidia chips cost more, so ASIC unit gains don't convert to dollar losses one for one -Baker's "maintained share handsomely" line is softer than the Altimeter slide, and the 2027 step down assumes ASIC ramp at a pace with no precedent -Jensen's argument: tokens per watt is revenue, so a cheaper ASIC saves capex but yields less revenue in a watt-constrained world Model frontier economics -Inference ARR went $29B (Dec-25) to $83B (Apr-26), roughly 3x in five months -Internal estimates of $155B at year-end 2026, $300B in 2027, $500B in 2028, with the outer years top-down and provider mix unallocated -Morgan Stanley 2027 capex at $1.1T, Gerstner closer to $1.5T including SpaceX and CoreWeave -At 60-70% gross margin the capex against revenue maths, but the $300B and $500B targets carry no bottom-up build -Monetization per GW is rising from ~$20B to $30-40B, what they call accidental profitability from demand outrunning supply [think this is the biggest jump in assumptions] Frontier versus open source -Closed frontier models captured roughly 90% of AI economic value in H1 2026, against consensus that cheap open-source tokens would converge -Baker's split: frontier keeps the revenue, open source takes most of the tokens -Harvey example: an open-source base plus proprietary RL and a router beat Opus 4 at lower cost, while still consuming heavy Opus volume -Stronger open source is bullish for compute because it pushes margin out of models and into hardware -Jensen could ship a frontier open-source model whenever he chose, Nemotron is already efficient, so it reads as a business choice not a capability limit

🚀 Brad Gerstner on the historic SpaceX IPO — and why it's a win not just for @elonmusk, but for all of America: "What an incredible day this is for America. What an incredible day it is for the hard work of the thousands of people at Starbase who've worked to put us back on the top in space. Remember, when @SpaceX got started, we didn't even have a shuttle to go to space to pick up our own astronauts. We were paying the Russians to go get our astronauts. Today, we're back on the top in space." — @altcap on @CNBC

Why is Brad Gerstner calling Cursor the thing getting the least attention as SpaceX goes public? Because if Grok 4.3 reaches the coding frontier, SpaceX will combine: - The largest proprietary coding dataset (more than the public internet) - The largest available compute cluster - The fastest-growing agentic coding product $10B ARR exit rate from Cursor in 2026. 1.5T parameter base model. Data injected into pre-training. @altcap says it is the upside scenario the market hasn't priced. The full breakdown on what that means for the IPO: podcastalpha.substack.com/p/bg2-baker-sp… Source: BG2 Pod - youtube.com/watch?v=Tx9jT2…

Noam Brown (@polynoamial) posted something profound this week. Frontier models can solve most problems if you just let them run long enough. Nobody has ever run Mythos for a full year. We may never know how smart any given generation actually is. @GavinSBaker's takeaway: however bullish he was on compute before that post, he's more bullish now. @altcap

⚡ Gavin Baker on why Elon's data center strategy is printing money — and Brad Gerstner on why nobody saw it coming: "If you can borrow money at 6, 7, 8% and invest in something with a 55% ARR, I'm not the most sophisticated thinker, but that math maths... We do know from Jensen that @elonmusk brings data centers up faster than anyone, 122 days. Speed is literally cost." "This whole category of taking all of this compute, which he's uniquely good at standing up, and then reselling it in a way that's highly profitable was not in a lot of people's forecasts — now it's a major component of the forecast." — @GavinSBaker & @altcap on @BG2Pod / BGGB Pod As Jensen Huang said, “Elon is an N of 1.”

A richness of great insight on the @BG2Pod from @GavinSBaker ...too many to choose from but here's a few good ones: “Something very important on open source: there’s this belief that it’s bearish for AI. It’s actually — maybe bearish for the frontier models — but really bullish for compute and hardware. If the frontier models are capturing less of the margin, then you’re going to spend more on compute. The better open source does, the better it is for compute providers..." “Two things can be true. The majority of economic value may continue to accrue to the frontier — and man, has it ever accrued to the frontier thus far. And the majority of tokens consumed in the world may be open source — and they are today. I think this current state is likely to persist.” "There is a belief that these data centers are commodities. I do not share that belief. In the same way that Elon was able to re-engineer a rocket from first principles and make it reusable, he engineered an electric car from first principles — I think he looked at data center design from first principles and designed something fundamentally different." "My understanding is that Cursor and Anthropic have more tokens of proprietary coding data than anyone else — and each have more tokens of proprietary coding data than exist on the public internet. Cursor used Kimi K2.5, used their own private data, did some RL, some supervised fine-tuning, and got a really good model. Then they spent three weeks in the Colossus 2 cluster and got a model that 12 days ago was Pareto dominant with Composer 2.5. It suggests the Cursor data is very valuable for coding, and that XAI/SpaceX has a shot at being a real player in coding.” “Nobody has run Mythos for a year continuously, and we may never know how smart each generation of models actually is or was — because we don’t have time to appropriately evaluate their intelligence before the next model comes out. This is a profound statement...Imagine Albert Einstein had just thought about fundamental physics 24 hours a day. He doesn’t have to eat, doesn’t have to sleep, never gets old, never has diminution of intelligence — and he thought for one year. We might already have solved a lot of these intractable problems. My takeaway was: however bullish I was on compute before, I’m just a lot more bullish.” On $NVDA: If all of his customers are going to compete with him, then why not compete with his customers? He has his own models that are really, really good — Nemotron 3.1 was really cool from a compute-efficiency perspective, and he’s always careful to release small models so as to not tread on Anthropic, OpenAI, Google’s toes. But that is a choice he is making. If the economics change, I think Nvidia can join the frontier and become one of the world’s largest cloud computing companies much faster than people think.” “Clark’s analysis shows that XAI’s deal with Google for cloud computing generates more operating profit per gigawatt than Anthropic, than Meta, than Google, than OpenAI. Freda calculated a 55% IRR on Colossus 1. If you can borrow money at 6, 7, 8% and invest in something with a 55% IRR — I’m not the most sophisticated thinker, but that math maths.” “I always imagine stocks as runners. In ‘22 that runner had gone downhill — it had a lot of energy. The market, particularly in the last two months, has run up a very steep hill. A lot of these stocks — forget climbing a mountain — they’ve gone straight up a cliff. They’re tired. They need to rest. Do they just rest at the top of the cliff, hang out in their harness, or do they need to go downhill for a bit? We’ll see. I do see a lot on X about finding the next bottleneck — I think that was the last game. That game is over.”

The AI market rally in 2026 has one anchor. It's private. podcastalpha.substack.com/p/brad-gerstne… @altcap Gerstner holds Anthropic equity with non-public financial access. His claim: without Anthropic's 2026 revenue delivery, the market would be down 10-15% right now. Not NVIDIA. Not OpenAI launches. Anthropic's profit path. High gross margins and a path to positive FCF in Q2 2026 triggered the market's ascent out of a deep correction. One private company validated that AI revenue was real and profitable. The anchor is private. The companies depending on it are public. If Anthropic's next revenue update disappoints, every AI-correlated position loses its primary justification. Source: TBPN (@tbpn) - youtube.com/watch?v=HRYysS…