Latest / Super Intelligence News and Report / Day 9
Transcript
- Super Intelligence News and Report — Day 9 Host 1: Welcome to Super Intelligence News and Report for Thursday, October 29th. Today's Today's deep dive is tailored specifically for you, the the listener who follows AI and business closely. We are jumping right past past the basic definitions today to unpack a really massive industry shift. Host 2: Yeah, we are looking at a 600 billion dollar infrastructure gamble and the intense fallout around who actually controls the hardware and the data so let's start right at the silicon layer. Host 1: Okay, let's do it. Host 2: Right. So on October 27th and 28th, 2026, the information and Reuters stated that OpenAI AI plans to buy Google TPUs. Now, I really need to stress that this is reported, not confirmed. Host 1: Right. Reported. Host 2: But still, the implications there are huge. Host 1: Oh, absolutely. Host 2: I mean, with hyperscalers projecting a combined $600 billion in 2026 AI infrastructure spend, the stakes here are just massive. And what makes this really credible is the is the hiring data. Host 1: Like who are they bringing on? Host 2: Well, OpenAI is actively recruiting a San Francisco TPU compiler engineer right now. Which directly follows them poaching ex-Google TPU director Richard Ho. Host 1: Okay, I want to make sure I understand the underlying strategy here. Is this basically like a tenant secretly building their own house using the landlord's power tools? Host 2: That is exactly it. Yeah. Host 1: I mean, they've they've been renting Google's compute, but now they want to own the actual processor. Host 2: Right. This escalates OpenAI's strategy from just temporarily renting Google capacity to making a permanent architecture commitment. They are diversifyingifying from NVIDIA to ensure they aren't trapped by a single vendor bottleneck. Host 1: And that desperation for hardware independence is clearly fueling some pretty brutal talent wars behind the scenes. Host 2: really is. Which brings us to our second headline. And this one is a confirmed legal development. On October 27th, 2026, a judge granted Apple Apple expanded expedited discovery in its trade secrets case against former employees who are now at OpenAI. Host 1: expedited discovery moves fast right like how much time do they have Host 2: super fast the defendants have just 14 days to produce records Host 1: wow just two weeks but wait expedited discovery is fairly standard in ip theft cases what makes this specific ruling so dangerous for those engineers Host 2: it is the unprecedented scope of what apple is allowed to demand personal ai tool history across platforms like chat gpt claude and signal is now being weaponized as standard discovery material in these hardware talent wars Host 1: that is wild i mean if courts are starting to treat every single personal prompt as a permanent permanent discoverable corporate record that completely reframes how internal ai tools should be managed Host 2: it changes everything about corporate privacy Host 1: which actually perfectly explains a very different kind of tracking crackdown over at Amazon for our third headline on October 27 2026 Amazon Amazon VP Turk Turkalitz posted a confirmed memo ordering software engineers to stop token maxing Host 2: token maxing I mean I have seen companies track AI usage to justify justify expensive licensing costs but why on earth would software engineers deliberately try to max out their token usage? Host 1: Well, because Amazon had set up an 80% weekly use leaderboard, you know, to encourage internal AI adoption. Host 2: Oh, I see where this is going. Host 1: Yeah. So instead of writing better code, staff simply gamed the metric by feeding an internal agent pointless infinite loop tasks just to burn through tokens and hit their KPI. Host 2: That is hilarious but also a massive waste of resources. Host 1: right so Amazon had to deprecate the leaderboard entirely as investors heavily scrutinized that that 600 billion dollar hyperscaler spend you mentioned earlier Amazon is writing the ROI rule book here they are judging AI success strictly by business outcomes not usage vanity metrics Host 2: that makes perfect sense I mean if you tie performance to raw usage you just incentivize waste but as soon as as you shift to outcome driven agents that actually execute complex business tasks you introduce a whole new massive security vulnerability Host 1: because those agents need actual access to your corporate databases to get things done right Host 2: exactly and and that brings us to our final confirmed move today also on October 27th 2026 in physical raised 40 million dollars at a 600 million dollar valuation Host 1: And what kind of volume are they are they handling to get that valuation? Host 2: They are currently securing 10 billion secrets a day for 1300 enterprise customers. Host 1: Let me push back on the back on the tech there for a second though. Why can't we just give an autonomous enterprise agent an API key, the exact same way we give a traditional software application one? Host 2: Well, because because traditional software doesn't randomly hallucinate or fall victim to prompt injection. Host 1: Ah, right. Host 2: If you hardcode a database key into an agent, and a malicious user tricks that agent into revealing its its system prompt, they just get the keys to the kingdom. Enterprise Enterprise agents cannot be trusted with raw credentials. Host 1: So, Infisical is building the essential least privileged privilege broker to stop agents from leaking keys by dynamically granting them them temporary access instead. Host 2: Exactly. Looking at all these developments together, the overarching theme for you today is pretty clear. The AI industry is rapidly maturing from raw capacity races to strict accountability. Host 1: Yeah, you see it across the board. Host 2: Right. Whether it's OpenAI aggressively owning its infrastructure, Apple tracing exact tool usage, or Amazon demanding demanding actual ROI. Host 1: It is a complete shift in how we audit intelligence. If every prompt prompt you enter and every token an enterprise agent processes becomes a permanent legally discoverable record, how will that change what we trust AI to do tomorrow? Host 1: We will leave you to mull that over as we hand off to the main story. Host 1: Can giants like Meta and Anthropic continue funding multi-gigawatt AI build outs when their own chip suppliers are literally the ones helping to finance the purchases? Host 2: Right. It's I mean, it's a wild dynamic. Host 1: It really is. And that is the central question driving the new cycle over the past 48 hours. We are doing a deep dive today into the massive interconnected funding web that was just revealed because, you know, this is is what's actually keeping the AI hardware hardware pipeline flowing right now. Host 2: Exactly. Host 1: You are watching the line between vendor, investor, and customer just completely dissolve. Host 2: Yeah, it is a phenomenal shift in the architecture of of infrastructure funding. I mean, to understand why this structure is materializing today, you really have to look at the immense, immediate pressure that Meta is under. Host 1: Oh, absolutely. Host 2: Mark Zuckerberg is pushing so aggressively for this idea of personal super intelligence. intelligence, and that ambition is is it's entirely centered around their Muse agent. Host 1: Right. Which if you're tracking the internal development, that's known as Hatch. Host 2: Yeah, exactly. Hatch. And to actually deliver that level of compute, Meta has guided a truly staggering capital spending target just for 2026 alone, we're talking between 135 and 145 billion dollars. Host 1: Which is just, I mean, it's hard to even wrap your head around that number. Host 2: It is. Host 1: And And when you look at their balance sheet, the math gets incredibly tight. Like Meta generated about 97 billion dollars billion dollars in operating cash flow over the last nine months. Host 2: Right. Host 1: So you have this 135 to 145 billion dollar capex bill looming for a single calendar year. And well, your operating cash flow simply cannot carry that weight on this own. Host 2: No, it it can't. Host 1: Outside equity is, you know, it's no longer optional at this point. It is structurally required to keep the build out on pace. Host 2: Which is exactly why we are seeing this supplier funded demand take center stage. Now, to be fair to the historical context, context, this isn't emerging from an absolute vacuum. Host 1: Right. We've seen hints of it. Host 2: Yeah. Back in in 2025, Nvidia agreed to over $100 billion in support tied to OpenAI. And we also saw Nvidia take, you know, continuous stakes in Anthropic, XAI, and Intel throughout 2026. Host 1: They were participating in what, over 100 funding rounds? Host 2: Exactly. Over 100. But But the deals that dropped between October 26th and 28th, they are a completely different animal. Host 1: They really are. Host 2: They take this somewhat abstract trend of ecosystem investing and they attach hard, undeniable balance sheet altering numbers numbers to it. Host 1: Okay, let's unpack this because to the outside observer, it feels a bit like a contractor lending you money just so you can turn around and hire them to build your build your house. Host 2: That is a great way to put it. Host 1: Right. It is a highly irregular closed capital loop. And to really grasp how this contractor financing is actively distorting normal market mechanics, we need to analyze the scale of the capital raises that just hit the wire. Host 2: The sheer volume is insane. Host 1: Let's start with Meta's strategy. They just unveiled a unveiled a $98 billion total planned equity program. Host 2: Wow. Host 1: I know. That is a seismic figure to drop on the market all at once. Host 2: It really is. And, you know, the mechanics of how they are deploying it tell you everything you need to know know about their urgency. They are pricing 37.5 billion of that that right now. Host 1: Right. Host 2: That breaks down to 25 billion in common stock and 12.5 billion in preferred. Host 1: Okay. Host 2: The remaining piece is an up to 60 billion dollar at the market or ATM program that's slated to kick off in early 2027. Host 1: And for a company of Meta's maturity, relying on a $60 billion ATM program, where you are dynamically dribbling shares shares into the open market rather than doing a massive block sale, that's a huge lever to pull. Host 2: It's massive. But if they manage to sell all of it, it covers roughly two thirds of their 2026 capex requirement. Host 1: Yeah. And the market's reaction to that immediate $25 billion common stock piece is what I find so fascinating. They priced it at $625 a share. Host 2: which is what, a 2.7% discount? Host 1: Exactly. A surprisingly tight 2.7% discount to the Friday close. Normally, I mean, if you flood the market with 25 billion in fresh equity, you expect to offer a much steeper discount to absorb that kind of kind of supply. Host 2: Right. To entice buyers. Host 1: Exactly. But the But the institutional appetite here was ravenous. It was massively oversubscribed, like five times over. Host 2: Five times over on 25 25 billion? That's wild. Host 1: It is. And what should really really catch your attention is the concentration here. The 15 largest buyers, took 70% of the allocation. Host 2: Right. And that level of concentration means the smartest money in the room is highly convicted on this closed loop strategy. Host 1: Yes. Exactly. And when you look at who is actually injecting capital, you see exactly how close that loop really is. Out of that equity offering, $4 billion is a combined supplier buy-in. Host 2: 4 billion. Host 1: NVIDIA alone is putting in $3 in $3 billion. And we should be really clear here, this is not passive portfolio management by NVIDIA. Host 2: Oh, not at all. That 3 billion is explicitly structured to deepen their their integration with Meta's infrastructure. It is tied directly to accelerating work on Muse models, deploying the Vera Rubin platform, integrating Vera CPUs, and scaling the Open Shell framework with Century Monitoring. Host 1: So they are locking themselves in. Host 2: Literally, NVIDIA is essentially buying a permanent structural moat inside Meta's data centers. Host 1: Wow. Host 2: And AMD is doing the exact same thing with the other $1 billion of that supplier buy-in. Host 1: Right. Yeah. AMD's capital is tied directly to ensuring their EPIC processors support the Muse ecosystem they specifically highlighted day zero support for Muse Spark open and Glimmer models Host 2: yeah Host 1: AMD even noted that initial Helios rack scale systems are already installed and undergoing active testing at Meta Host 2: Right now, Host 1: yeah. You have hardware vendors literally utilizing their own balance sheets to buy equity in their largest customer just to ensure their specific proprietary architectures are physically bolted into the data center floors. Host 2: But equity is only one side of the coin here because if equity is how you fund a cash generating behemoth like Meta, complex debt is how how you fuel a pure play frontier lab like Anthropic. Host 1: Ah, right. Host 2: which is why we need to examine this massive new $63 billion financing vehicle that was just lined up for Anthropic. Host 1: Lined up by heavy hitters too. I mean, you've got Bank of America, Citi, Morgan Stanley, and Blackstone all in on in on this. Host 2: Yep. Host 1: And the entire purpose of this vehicle is to finance Broadcom AI chips and networking gear. Host 2: Yeah. And the structural engineering of that 63 billion dollars is arguably the most aggressive financial maneuver we have seen this entire cycle. Host 1: Really? How so? Host 2: well 43.5 billion of it takes the form of first lien senior debt Host 1: Okay Host 2: But Frontier AI labs do not have the hard assets or predictable decade long cash flows to secure 43.5 billion in traditional first lien debt. Host 1: Right. They just don't have the history for that. Host 2: Exactly. The linchpin making this palatable to conservative banking institutions is Broadcom itself. Broadcom is providing what is called residual value support. Host 1: Okay. Let's pause there because the mechanics of that are wild. Residual value support effectively means Broadcom is acting as a final guarantor guarantor to the lenders. Host 2: Right. Host 1: So if the value of of these AI clusters depreciates faster than expected or if the special purpose vehicle defaults, Broadcom Com covers the gap. Host 2: Yes. Host 1: The chip maker is backstopping the financial risk of their own their own rapidly depreciating hardware hardware just to facilitate the sale. Host 2: Exactly how it functions. They are writing an insurance policy on their own technology to unlock bank liquidity for their customer. Host 1: Wow. Host 2: And when you look at Anthropic's financials, you immediately see why this level of synthetic support is required. We just got a look at Anthropic's reported first half revenue, which was revealed in their October 26th S-1 filing, for a NASDAQ IPO under the ticker ANT. Host 1: ANT. Host 2: Right. They pulled in 16.3 billion dollars in the first half. Host 1: Which, I mean, that is a phenomenal revenue number for a for a company of their age. Host 2: Oh, absolutely. Host 1: But when you place a a 16.3 billion dollar half year revenue against a 63 billion dollar financing vehicle just to acquire Broadcom gear, well, the math snaps into focus. Host 2: It really does. Host 1: Even hyper growth revenue is nearly enough to organically fund the bleeding edge of this hardware race. Host 2: And that shortfall is driving extreme behavior across the entire ecosystem. largely because the underlying cost basis is shifting so rapidly. Host 1: Right. Host 2: Now, I need to state clearly that this next data point is reported, not confirmed, but based on trend force supply chain reports, chain reports, TSMC is weighing a 10 to 20% price hike on on co-os advanced packaging. Host 1: 10 to 20%. Host 1: Yeah. Host 2: Yeah. When your core bottleneck gets a 20% markup, every financial model you built yesterday is suddenly underfunded. Host 3: today. Host 4: Here's where it gets really interesting because you have all these billions flying around these complex interdependent webs and it is is completely fractured Wall Street's perspective on what is actually healthy for the market. Host 4: market. Host 3: Oh, totally. Host 4: I'm going to argue the the bull case here because frankly, this structure is brilliant. And the market is validating it in real time. Host 3: Okay, let's hear it. Host 4: Look at the institutional demand we just talked about. You do not get a 5x oversubscription on a a 25 billion dollar equity sale if the broader market thinks this strategy is flawed. Host 3: Fair point. Host 4: And more importantly, more than half of that allocation went to long only holders, holders that proves long term conservative investors are deeply comfortable funding builders they know even even if the suppliers are holding their hands. Host 3: But you're assuming that demand is an accurate reflection of fundamental health rather than just, you know, you know, momentum chasing. Host 4: Well, I'm assuming it reflects trust in in the balance sheets involved. Look Look at the immediate market reaction. AMD jumped 5.45% to $560.26 on this news. Host 3: Right. Host 4: Right now, 47 analysts have a strong buy consensus on AMD with price targets stretching all the way up to $800 from Citi. Host 3: Sure. Host 4: And as for the risk, Broadcom is taking on that debt, Broadcom is an absolute fortress. Moody's explicitly views Broadcom as one of the strongest large tech companies in existence. Host 3: They are strong, yes. Host 4: They're perfectly perfectly capable of utilizing that fortress balance sheet to unlock debt for Anthropic. It's just a highly efficient necessary utilization of capital across an ecosystem that needs to move incredibly fast. Host 3: If we connect this to the bigger picture, I have to take the bear case here. because relying on the strength of Broadcom's balance sheet ignores ignores the deep structural fragility of the system itself. Host 4: How so? Host 3: Well, consider what Nigel Green, the CEO of DeVere Group is pointing out. He calls that $3 billion NVIDIA purchase into Meta a massive red flag for what he terms a circular AI economy. Host 4: I understand the skepticism but how is it a red flag if both companies are aggressively aggressively growing their capabilities? Host 3: because it fundamentally distorts price discovery and demand signals. Think about it. about it. Money goes out of NVIDIA's treasury as an investment, and it comes right back onto NVIDIA's income statement as a booked sale. Host 5: Right. I see that. Host 3: That severely blurs the line of true organic market demand. We have to remember that outside of this insular Silicon Valley loop, everyday consumers are still incredibly reluctant to pay a premium for AI wrappers. Host 5: True. Host 3: And And enterprise businesses are notoriously notoriously disciplined you know they only pay for clear measurable outcomes that reduce headcount or increase efficiency. if the primary source of capital funding your multi gigawatt hardware purchase is the hardware company itself you are obscuring the the actual end user demand curve from the broader market. Host 5: But again, you have institutions like Morgan Stanley and Blackstone signing off on the debt side. I mean a $43.5 billion first lien debt package for Anthropic? Doesn't clear the credit committees if the internal risk models are flashing red? Host 3: Individually, those models might look green, but Moody's issued a specific vital warning about exactly exactly this dynamic. Host 5: What do they say? Host 3: They noted that while Broadcom's individual balance sheet is remarkably strong today, the risk profile skyrockets if multiple deals structured exactly like this stack up in a short period of time to meet the insatiable funding needs of these frontier labs. Host 3: Individually, they look like secure, well collateralized deals, but together, aggregated across the sector, they create a highly correlated systemic risk. Host 4: Meaning if one one domino falls, the entire chain reacts. Host 3: Exactly. Just think about the mechanics. If organic end user demand falters, the frontier lab, say Anthropic, can't generate the cash flow to pay the lease on the hardware. Host 4: Right. Host 3: The special purpose vehicle defaults. The banks instantly call in the chip makers residual value guarantee. Host 4: Oh, wow. Host 3: Yeah. And if this happens across multiple labs simultaneously, you have chip makers suddenly forced to pay out tens of billions in cash guarantees for rapidly depreciating hardware that is just sitting in data centers that no one is paying to use. Host 4: Okay, weighing both sides here, the bull case is undeniably right about the current market appetite. I mean, the cash is readily available, the institutions are eager to deploy it, and the balance sheets of these incumbent chip makers are currently pristine enough to act as financial anchors. Host 3: Yep. Host 4: But But the bear case exposes a reality you simply cannot ignore. It is is a closed loop. And sooner or later, someone completely outside outside this interdependent financing circle has to pick up the bill, by actually paying for the AI outputs at scale. Host 3: Exactly. Host 4: Sooner or later, the music stops and the end user has to provide the actual liquidity that justifies all justifies all this infrastructure. Host 3: So what does this all mean? Host 5: Why are we seeing Meta and Anthropic scramble to arrange these incredibly complex highly correlated financial bridges right this exact second? Host 3: That's the real question. Host 5: Right. Because if this is a multi-year build out why not wait for the the revenue to scale naturally and fund fund it organically? Host 3: Because they are trapped in a vicious cost pressure cooker the urgency is not being dictated by their own product roadmaps maps it is being dictated entirely by the physical supply chain Host 5: The hardware reality Host 3: yes we mentioned the potential 10 to 20% COAS pricing increase earlier but the silicon itself is also moving Host 4: okay Host 3: and again this specific figure is reported not confirmed but Liberty Times reporting indicates that TSMC has finalized a three to 6% wafer out price rise starting in January 2027. Host 5: A three to 6% hike. Host 3: Yeah. And crucially, that price hike is going to be be the steepest on their two nanometer and three nanometer nodes. Host 5: Which are exactly the cutting edge nodes these these next generation AI chips rely on. So the foundational material is literally getting more expensive. Host 3: Yes. And it is not just the financial cost of the silicon. It It is the physical time it takes to get it into a rack. Equipment lead times have stretched significantly across the board. Host 5: How much? Host 3: Well, we remodeling three to six six months not too long ago and now we are looking at six to eight months and and TSMC is reportedly just ordering at existing prices without even attempting to push for bulk discounts because securing securing hard capacity is vastly more important than protecting margin right now. Host 3: now. Host 5: I want to dig into the testing bottlenecks too because that physically limits how fast these chips can ship regardless of how much capital you raise. Host 3: Oh the testing bottleneck is becoming a critical choke point. Current estimates from the supply chain show that final testing for Nvidia's Blackwell architectures takes roughly four times as long as it did for the previous Hopper generation. Host 5: Four times as long. Host 3: Yeah and this isn't just plugging it in to see if it turns on this is complex burn in testing for massive multi-die packages where you know a single microscopic flaw ruins tens of tens of thousands of dollars of silicon. Host 3: silicon. Host 5: That's incredible. Host 3: And when we look ahead to the upcoming Rubin architecture estimates suggest it could take seven times as as long to test as Hopper. Wait, Host 5: Wait, seven times? Host 3: Seven times. Combine those testing times with COAS advanced packaging packaging capacity which is already incredibly tight and heading toward a maximum of 200,000 wafers a month in 2027 and you have a perfect storm. Host 5: Wow. Host 3: The unit cost of AI compute is getting vastly more expensive and time consuming to build up front. Host 4: It is exactly like trying to build a massive fleet of rockets but the titanium titanium you need gets more expensive every single day and the factory suddenly takes takes twice as long just to test the engines. Host 3: That's exactly it. Host 4: If you decide to wait for the organic ticket revenue from your first launch to fund the construction of your second rocket you will either be priced completely out of the raw materials or you will lose your spot in the manufacturing line to a competitor. Host 4: You have to lock in your cash now even before you have sold a single ticket to space. Host 3: That analogy captures the dynamic perfectly. We are in this bizarre paradigm where compute is theoretically getting cheaper to query on the software user side but vastly more expensive and complex to build on the physical infrastructure side. Host 4: Right. Host 3: Meta and Anthropic are securing this synthetic capital today because well if they wait until 2026 to figure out how to organically pay for pay for that 145 billion dollar capex the physical capacity at TSMC will already be sold to to someone else who was willing to leverage their suppliers balance sheet. Host 4: Which brings us to the immediate future because the true test of of this circular economy isn't going to be debated in academic papers or theoretical models. It is going to happen violently in the public markets over the next few weeks. Host 3: Absolutely is. Host 4: we are heading into an intense gauntlet of catalysts that will prove whether this structure actually holds Host 3: The earnings wire is going to be the first major stress test Host 5: and it starts immediately Microsoft and Alphabet report after the close on October 28th the market is going to be hyper focused on Microsoft specifically whether Azure can maintain growth numbers near that 45% 45% mark which you know is what Host 6: their own massive spend. Host 7: Right. Host 6: And everyone will be parsing Alphabet's commentary on their Google Cloud TPUs, looking for signs of internal internal efficiencies versus external revenue generation. Host 7: Definitely. Host 6: Then you have Amazon reporting on October 29th and AMD on November 3rd for AMD, we will really need to watch their data center numbers closely. Along with any forward guidance on those Helios rack scale systems they just mentioned with Meta. with Meta. Host 7: And looking beyond the established legacy players, the Anthropic IPO is shaping up to be a massive market defining litmus test. Host 6: Oh, for sure. Host 7: They filed their S-1 on October 26th 26th and they are targeting a mid to late November listing. If we track the standard 15 day rule after a public filing, they could officially launch their investor road shows in early November. Host 6: Okay. Host 7: The critical existential question for the the market is whether that public offering will will raise enough truly organic external capital to meaningfully reduce their reliance on Broadcom's synthetic lease vehicles. Host 6: Then looking looking slightly further out, you have Meta's $60 billion at the market program slated for early 2027. Host 7: Right, the ATM program. Host 6: We touched on how an ATM program works, slowly bleeding shares into the market. The big unknown there is whether whether an active sale of that unprecedented unprecedented magnitude will severely pressure Meta's underlying share price over a a prolonged period. Host 7: That's a huge risk. Host 6: But there is also a secondary arguably more dangerous dangerous question lurking here regarding the accounting standards. Will auditors start demanding that revenue be reported net of these supplier investments? Host 6: Oh, Host 7: Oh, that would change everything. Host 6: Exactly. If you If you are taking $3 billion from Nvidia while paying them $10 billion for chips, and the accountants force you to net that out, the top line optics of the entire industry change dramatically. Host 7: But ultimately, all of these complex financial maneuvers, the IPOs, the synthetic debt vehicles, the ATM equity sales, the the residual value guarantees, they all eventually bow to one ultimate unavoidable metric. Host 6: End user demand? Host 7: Yes. Do Muse and Claude actually deliver measurable, undeniable business outcomes that justify this historic multi-gigawatt capex? If enterprise adoption stalls or if the productivity gains plateau, no amount of creative supplier financing will save these balance sheets from the realities of depreciation. Host 6: It all comes back to the rockets. You can secure the titanium, you can build the engines, and your suppliers can even buy a few of the tickets themselves to make the launch look successful. But eventually eventually, you need a paying public that actually wants to go to space or you just have a very expensive silo. Host 7: This raises an important question one that I think every institutional investor needs to be asking themselves as they model out the next five years of this sector. We are watching traditional silicon hardware hardware companies functionally evolve into investment investment banks for their own customers, Host 6: Yeah. Host 7: they are underwriting specialized debt, they are taking massive concentrated equity stakes, and they are backstopping residual asset value. If this circular financing loop becomes the permanent standard for building frontier technology, how will we ever accurately measure true organic market demand again? Host 6: We'll see you back here here tomorrow, the same time and please check the check the show notes for all sources