In short: Cited as the example of a debt-and-equity-issuing tech giant, not a stance: he is top-down and doesn't know "whether I'm bullish or bearish on Oracle". But issuers of this kind are the likely "problem children" if the market gets an upset stomach, and Oracle "was having a tough go of it".
Oracle is a big database and cloud company that has been borrowing and issuing stock to fund its AI data-center build-out. Weniger says plainly he has no bottom-up view of the company; he is a "top-down" macro strategist.
He uses Oracle to illustrate a risk. When companies issue new shares ("diluters"), every existing share owns a smaller slice. When they borrow heavily, they compete with the government for lenders, which pushes interest rates up. Higher rates make profits expected years from now (say, Oracle's 2032 earnings) worth less today. If markets turn nervous, he expects the diluters to be the "problem children".
32:16have some sort of upset stomach, it will be those who are increasing share count. That would be the problem children. We've seen that a few times. I was referencing Oracle earlier. Oracle was having a tough go of it there for a spell. And I think that that's going to theoretically be one of those barbells in the coming months.
In short: Cited as evidence, not a stance — the proof the labs cannot slow down: "They both have hundreds of billions of commitments to hyperscalers. OpenAI alone represents 300 billion of Oracle's 600 billion plus backlog. If they were to slow down, they could not fulfill those commitments."
14:50Here's my theory. First of all, the claim that Dario Mode and Sam Alman are making about slowing down is just false on its face. The entire future, the entire future of anthropic and open AI depends on there not being any slowdown. They both have hundreds of billions of commitments to hyperscalers. Open AAI alone represents 300 billion of Oracle's 600 billion plus backlog.
In short: Up 5%+ as a relief trade; Ethridge flags the leverage (16:47–17:23). Ethridge: Oracle is "using their own equity as the thing that the debt is being offered against. And there's only so long that can go if the share price doesn't continue to appreciate." Santoli: a proxy for OpenAI's "very leveraged approach" to compute; is it more than "today's weather"?
In short: Neutral on the equity, positive on the credit: explicitly included with Meta, Google, Amazon and Microsoft as having "plenty of cash flow" to pay AI debt back over time — "the bonds are solid," even as the stock may fall.
Oracle is on his list of hyperscalers whose debt he considers "money good." It has enough cash coming in to service and eventually repay what it borrows for AI. He explicitly says that doesn't mean the stock is safe; the view is about the credit only.
16:17The other guys are borrowing the money to use Anthropic's products. So, can Anthropic go bust? Sure. But that's an equity problem, not a bond problem. And I think the hyperscalers, I think Meta, Google, Amazon, Microsoft, Oracle, they have plenty of cash flow coming on through to fund, to make the coupon payments and pay these things back over time.
In short: Named with OpenAI as an AI player "turning to the government" for support: "if it doesn't work, bail me out." It's his too-big-to-fail example, and he warns shareholders can be wiped out before any rescue.
Oracle has borrowed heavily to build AI data centers, much of it for OpenAI. Deluard groups it with OpenAI as a company effectively asking for a government backstop if the AI bet goes wrong. His warning is that a bailout protects the system, not necessarily the shareholders.
44:06Yeah. That's the argument. You see that on both sides, right? Both the big AI labs basically turning to the government and saying, okay, if it doesn't work, bail me out. You certainly see that with Oracle and OpenAI — hey, hand me 500 billion here and there.
In short: "The same thing happened with… Oracle, where… the bonds widened out tremendously right after it was issued. What the bond market is saying is these ratings don't make sense to us." (Auto-transcript garbles the rating detail — the point is market pricing below the agency rating.)
Oracle has borrowed heavily to build AI data centers. Gundlach cites its new bonds as another case where prices fell sharply right after they were sold — buyers demanding far more interest than the rating implied. It is a comment on the debt, not a stock recommendation.
6:55And I have a feeling that that rating was encouraged by some persuasion of the rating agencies — because the bond market isn't buying it at all. And the same thing happened with, there was another IPO, Oracle, where it got a junk bond rating but the bonds widened out tremendously right after it was issued. What the bond market is saying is these ratings don't make sense to us.
In short: The trigger for his work: in March–April the huge OpenAI commitment made Oracle's backlog "explode for multiple years," "and then I went and looked at the bonds. And the bonds said, no." Stocks and bonds disagreeing is "exactly what happened in 06" during the housing bubble.
Oracle signed an enormous cloud-computing commitment with OpenAI that made its order backlog soar for years ahead, and the stock loved it. Taylor looked at Oracle's bonds instead, and they sold off from March–April. Lenders were saying the deal adds risk, because it depends on a customer (OpenAI) that must raise hundreds of billions of dollars it doesn't have. That stock-up, bonds-down split launched his whole AI-bubble analysis.
1:13This is what I wanted to start talking about in May at your conference, but I had a modest medical issue. It had to be ushered out to the hospital. You're good now. You look great. So let's say March, April. It was right on that Oracle thing when the ARR, they got this huge commitment from OpenAI and their backlog just exploded for multiple years.
In short: A blow-out backlog financed by the shareholders — and a founder sale that was announced and withdrawn inside the call. "Oracle delivered more than 300,000 GPUs to its AI cloud customers since the end of Q4… RPO surged by 209 billion to 664 billion, but at the same time, Larry Ellison announced a new 10b5-1 plan to sell 7.5 billion of stock… Oracle also guided FY revenue to at least 90 billion and adjusted EPS to $8 a share while disclosing 28.5 billion of capex, negative 5 billion of free cash flow and a $20 billion ATM equity program, which means that it's going to be issuing stock. It's probably why Larry Ellison is trying to front run that and sell his own stock. First time in basically ever. So that's not going to be good when the market opens." Then, from a subscriber in Q&A: "Oracle chairman Larry Ellison just canceled the planned 7.5 billion stock sale without selling a single share, especially I think given this AI slowdown. I don't think he wanted to look like he was panicking." The overhang is removed; the ATM and negative free cash flow are not.
Oracle's cloud business is booming on AI demand: it has delivered more than 300,000 AI chips to customers in a quarter, and its signed-but-not-yet-delivered contracts jumped by $209 billion to $664 billion. The problem is paying for it. Oracle plans $28.5 billion of capital spending, expects free cash flow of about minus $5 billion, and set up a $20 billion programme to sell new shares into the market over time — which dilutes existing owners.
In the same week, founder Larry Ellison filed a plan to sell $7.5 billion of his own shares. Singh read that as the founder getting ahead of the company's own share sales, and expected the stock to react badly. Then, during the call, a subscriber pointed out that Ellison had cancelled the plan without selling a share — Singh thinks because he did not want to look like he was panicking during an AI-slowdown scare. The founder sale is off; the company's need to issue shares and its cash burn are not.
Full passage: premium transcript (PDF).
In short: 850 megawatts later — the backlog starts converting. "Oracle's AI story has been easy to summarize over the past year. It signed enormous contracts first, spent enormous sums building the capacity to serve them, and promised the revenue would follow. That promise finally started showing up in the August quarter." Oracle delivered 850 MW of new data center capacity in Q1 FY27, almost triple Q4 FY26, deploying 300,000+ GPUs. Revenue +30% Y/Y to $19.3B, ahead of its 27%–29% guide; adjusted EPS $1.92 vs $1.74 consensus. Cloud +62% to $11.6B, led by OCI +121% to $7.4B (accelerating from 93%); cloud applications a slower +10% to $4.2B; traditional software −3% to $5.6B. RPO $664B, +$209B Y/Y and +$26B sequentially, with >$30B of new AI cloud contracts booked even as more backlog converted. The CapEx "needs some context": reported $28.5B drove ~$5B of negative quarterly FCF (−$29B TTM), but customer prepayments covered $11B, netting the cash outlay to ~$18B; the previously announced $20B equity offering is complete, and the latest $30B+ of contracts "require no incremental increase to its planned capital raise." Margins: gross margin fell ~6pp (chart: 60%, −7pp) as lower-margin OCI grew in the mix, yet GAAP operating margin expanded ~6pp to 35% and net margin reached 25% — "so far, operating leverage elsewhere is absorbing the infrastructure mix shift." Q2 FY27 cloud growth guided to 65%–71%; ~$70B of net CapEx to fund this year. Stock jumped after hours but remains ~50% below its September 2025 peak. Takeaway: "Oracle no longer needs to prove that AI demand exists… The remaining question is whether the economics justify the enormous capital required to get there. This quarter was an important step toward proving they can."
Oracle spent decades selling database software. Over the last year it reinvented itself as a landlord for AI: it signs long contracts with AI companies to rent them computing power, then builds the data centres to deliver it. The worry was the order of events. Oracle signed hundreds of billions of dollars of contracts — its "backlog," the revenue customers have committed to but that hasn't been earned yet — long before it had the buildings, power and chips to serve them. Until the capacity exists, a backlog is only a promise.
This quarter the promise started being kept. Oracle switched on 850 megawatts of new data-centre capacity — nearly three times the previous quarter — and its cloud-infrastructure revenue more than doubled, with the growth rate speeding up from 93% to 121%. Total revenue grew 30%. And the backlog still grew, to $664 billion, because new contracts kept arriving faster than old ones were used up.
The cost is enormous. Oracle spent $28.5 billion on construction and equipment in three months, and over the past year it has burned $29 billion more cash than it generated. Two things soften that. Customers paid $11 billion upfront — effectively lending Oracle part of the money to build their own capacity — so Oracle's own outlay was closer to $18 billion. And it has already raised $20 billion by selling new shares, with management saying the latest contracts don't require raising more.
One trade-off to understand: renting out AI computing is a lower-margin business than selling software, so as it becomes a larger share of revenue, Oracle's gross margin falls — about 6 points this quarter. But its operating margin actually rose, because the rest of the company's costs grew far more slowly than revenue. The author's verdict: demand is no longer the question; whether the economics justify roughly $70 billion of spending this year is, and this quarter was "an important step toward proving they can." Analysis, not a recommendation.
In short: Named with SoftBank as the only public ways to express an OpenAI view — both too derivative for her; she would "rather sit around and wait for the IPO." No view on Oracle's own business.
In short: Lebenthal's buy-more after the print, with every number he asked for on Sep 8 delivered — and the stock red anyway. "They got the top line revenue growth that we were looking for, 30%… free cash flow a little better than expected, actually a lot better than expected… it was expected −10 billion, it was −5 billion… they did not announce any more CapEx plans than what are already in place, no new fundraising plans. And if I look at the credit default swaps on their five year paper, they're actually starting to come in… a few weeks ago they were up at 215 basis points, now they're at 181 basis points." On the drop: "the answer is simply that interest rates have gone up… people are looking at a company that has 80 odd billion going probably over the next two years to 150 billion in debt… To me, this is a buying opportunity… a name you can add new money to right now, unless you think the AI thing is all a bubble. I don't." Plus the backlog: "their continuing performance obligations went up by another $30 billion… this is coming up on $700 billion of backlog… they don't need all of that to come through for this to be a massively undervalued stock." Harrington dissents on management capability rather than valuation (see CoreWeave row).
Oracle rents computing power to companies building AI, which means it has to buy the data centres and chips years before the rent shows up. That is why its free cash flow — the cash left after building — is negative: it spent about $5 billion more than it took in, though Wall Street had penciled in twice that.
Lebenthal's checklist from three days earlier was: don't raise the building budget, and grow the top line. Both happened — revenue up 30%, no new capex plans, no new fundraising — and the insurance market agrees: a credit default swap is what investors pay to insure Oracle's debt against default, and that price fell from 215 to 181 basis points, meaning perceived default risk dropped. Signed-but-unbilled future work ("remaining performance obligations") grew another $30 billion to roughly $700 billion.
The stock fell anyway because interest rates rose that day, and Oracle's debt is heading from about $80 billion toward $150 billion — higher rates make that pile more expensive. His view is that this makes the shares cheaper, not riskier: "a name you can add new money to right now, unless you think the AI thing is all a bubble."
In short: "In some ways, it was a very good quarter" — revenue +30%, non-GAAP EPS +30% (non-GAAP excludes stock comp, "which I just think is wrong"). "On the somewhat negative side," RPO reached $664B vs $600B the prior quarter — "not that much greater" — and "approximately 50% of its RPO coming from OpenAI, a company that we all know has massive negative cash flow." Up 4% after hours but "way off its highs" on the over-reliance.
Oracle rents out computing power for AI in its cloud business. On the surface the quarter was strong: sales and adjusted profit both up about 30%. Eisman notes one accounting quibble — the "adjusted" profit leaves out the cost of paying employees in stock, which he thinks is wrong because that cost is real.
The bigger issue is the backlog. "RPO" (remaining performance obligations) is the total of signed contracts not yet delivered — future sales already agreed. It grew only from about $600 billion to $664 billion in a quarter, and roughly half of it is owed by one customer, OpenAI, which burns large amounts of cash. So half of Oracle's future depends on a single company that cannot yet pay its own way.
18:01On the somewhat negative side, remaining performance obligations, RPO, a form of backlog, reached 664 billion, which is not that much greater than the 600 billion of the prior quarter. And Oracle remains highly dependent on OpenAI, with approximately 50% of its RPO coming from OpenAI, a company that we all know has massive negative cash flow.
In short: Owned by Belski, with Lebenthal supplying the checklist for Thursday's print. Belski: "I think they got thrown out with the SaaS… the baby with the bathwater. With respect to this funding, I think the company has already bottomed… if you're looking for value in tech, this stock's providing some value in a tech space that obviously is very, very expensive. It's one of our core names in a couple portfolios," and he expects the report to "provide more clarity in terms of their funding mechanism." Lebenthal's conditions: "I would like to see them not raise their CapEx expenditures for this year. I think it's at 70 billion… the SaaS business is their cash cow, and they're using that to fund the CapEx along with external financing. Don't raise that, because they've been issuing a lot of debt. The five year credit default swaps are over 200 basis points… what you look for in this report on Thursday is the top line growth… we're expecting 33% year over year. If they can hit that, then we can worry about the margins," with the Abilene, TX data centre coming online as the margin read.
Oracle sells database and business software — a steady, cash-generating business — and is spending very heavily to build AI data centres, which is not steady at all. The market's worry is how it pays for that: the software profits do not cover it, so the company borrows the difference.
Belski owns it and thinks the worry has already done its damage: the stock was sold off with the rest of the software complex, the selling has stopped, and it is one of the few genuinely cheap ways to own technology. He expects this week's results to explain the funding plan more clearly.
Lebenthal's contribution is a checklist you can score on Thursday without any opinion at all. First, the company should not raise its roughly $70 billion capital-spending plan — more spending means more borrowing. Second, watch its credit default swaps, which are insurance contracts against the company defaulting; they are already above 200 basis points, meaning insuring the debt costs more than 2% a year, and rising would be a warning. Third, revenue for the year to May should grow 33%; hit that and the margin questions can wait.
In short: Flagged by Nathan as the week's read-through rather than a call: "this week we're going to have Oracle and Adobe earnings, which might be interesting. More on the Oracle side, Guy, cuz that's probably the clearest connection between what SaaS and where it sits as far as AI build out and all that sort of stuff that we've been talking about." No view on the stock — it is being watched as the cleanest public test of whether AI capex is converting into software revenue.
No view is offered on the stock. Oracle is flagged as a read-through — the number the hosts intend to use to judge something bigger.
The question hanging over the whole AI build-out is whether the enormous spending on data centres and chips turns into revenue for the companies selling software. Oracle sits unusually squarely on that seam: it sells traditional enterprise software and databases and rents AI cloud capacity, so its results report both sides of the trade at once. That is why Nathan calls it "the clearest connection between what SaaS and where it sits as far as AI build out."
Practically: watch the cloud/AI infrastructure backlog and bookings in the print, not the headline earnings per share. A strong number is evidence for the build-out converting; a weak one is the first place the software-de-rating argument would show up in real results.
1:08You saw some kind of movement one way or another. This week we're going to have Oracle and Adobe earnings, which might be interesting. More on the Oracle side, Guy, cuz that's probably the clearest connection between what SaaS and where it sits as far as AI build out and all that sort of stuff that we've been talking about.
In short: Jim Lebenthal's final trade, in four words: "Oracle's got good momentum." No elaboration was offered on the day — but it is the direct continuation of his position two sessions earlier, where he held the name through a bad stretch on the compute-constraint thesis ("we are compute constrained and Oracle will benefit") and read Dell's raised forecast straight through to it. Here the justification has changed from fundamental to technical: it is now the momentum, not the CapEx read-through, that he is naming.
Oracle rents out computing power to companies building AI systems. Lebenthal's final trade is four words — "Oracle's got good momentum" — and the change from two sessions earlier is worth noting, because it is a change in the kind of reason he gives.
Then, he held the stock through a painful stretch on a fundamental argument: the world is short of computing capacity, so Oracle will benefit whatever the share price does. Now he is naming price behaviour instead. Momentum arguments and value arguments have different exit rules — a value thesis survives a falling price, a momentum thesis does not — so it is worth knowing which one you are actually holding.
In short: The other unconditional base in the sweep: "Oracle looks like it's bottoming." Immediately qualified at the index level — "but I'm seeing a lot more charts that are rolling over than charts that are basing."
Oracle sells database software and, increasingly, rents out cloud computing capacity to companies training and running AI models.
Like Intel, it appears purely as a chart read: "Oracle looks like it's bottoming." No revenue, backlog or margin argument is offered — the entire claim is that the price pattern has stopped falling and started to base.
He caps it immediately with the count that matters more than either name: "I'm seeing a lot more charts that are rolling over than charts that are basing." In other words, treat these as individual exceptions, not as evidence that the market has turned.
9:18Intel looks like it's bottoming. Oracle looks like it's bottoming, but I'm seeing a lot more charts that are rolling over than charts that are basing. — Okay. Yeah, healthcare had been a laggard but has been recently catching a bid. Semis have been on fire for three years.
In short: Jim Lebenthal's final trade, and a position he is holding through real pain — he describes it as "a name… that has been in the AI trade in the CapEx build out and the stock has done terribly." He reads Dell's raised forecast straight through to it: "when I look at Dell's response today, I think to myself, hey, Oracle's going to be just OK. It's going to be fine. Don't worry about it." The final trade states the thesis in one line: "we are compute constrained and Oracle will benefit." No trim, no stop, no hedge disclosed — the position is being sat through on the CapEx call rather than the chart.
Oracle has spent heavily to turn itself into a supplier of rented AI computing power to companies that need it. It is Lebenthal's final trade, and he is holding it through a bad stretch — he describes it as a name in the AI CapEx build-out where "the stock has done terribly."
His thesis is one sentence: "we are compute constrained and Oracle will benefit." In plain terms, there is more demand for AI computing than there is hardware to run it on, so anyone who owns capacity can charge for it. The read-through he uses is Dell's raised forecast that morning: if the companies selling data-centre hardware are guiding higher, the spending that Oracle's plan depends on is still coming.
The risk is worth naming, because he does not: this is the most leveraged version of the AI-CapEx bet. The spending is Oracle's own, upfront and debt-funded, and the revenue arrives later. If the build-out slows for any reason — including the political backlash discussed at length elsewhere in the same episode — the cost is already sunk.
In short: Named in the roll-call of scaled compute providers — Oracle, SpaceX, the labs procuring directly, the neoclouds — that took the cloud market "from basically three vendors with thousands of customers" to nine or ten providers with fewer customers. Structurally why hyperscaler multiples don't re-rate on growth.
29:01You include Oracle and SpaceX and the labs who are procuring directly and the neoclouds and CoreWeave and Nebius at least. So you have a market that went from maybe three to 10. They still have scale advantages, but you have fewer customers. So I understand why those stocks are not trading at higher multiples given their growth.
In short: Named inside the same seven-name off-balance-sheet tally — $3.1 trillion of arrangements across Meta, Google, Microsoft, Amazon, Oracle, Nvidia and Broadcom — as hyperscaler CDS spreads hit all-time wides. His read on whether that unwinds: "these CDS spreads could calm down if they stop issuing more and more debt, but I don't think that that is the goal." The general form of the risk is the one he has been building for weeks — elevated AI capex "continues to erode free cash flow for the hyperscalers, which could lead to continued debt and equity issuance for the group in the future."
Oracle sits inside the same $3.1 trillion of off-balance-sheet commitments, and inside the same logic. Building AI data centres now costs more than these companies generate in spare cash, so the gap is funded with borrowing and with contractual obligations that sit outside the headline debt figures.
Singh's general formulation is the one to keep: elevated AI capital spending "continues to erode free cash flow for the hyperscalers, which could lead to continued debt and equity issuance for the group in the future." More issuance means wider credit spreads, and wider credit spreads eventually mean a higher cost of building — which is how a demand story becomes a financing story.
Full passage: premium transcript (PDF).
In short: Raised by Faber as the emblem of the regime change in equity supply — "it's even shifting from companies like Oracle, like buybacks versus share issuance and dilution… shifting pretty hard in the other direction." Kedrosky takes it as the setup for his issuance-supply argument rather than a view on the company.
Oracle comes up as a marker of a regime change in how big technology companies treat their own shares. For years the flow ran one way: companies used spare cash to buy back stock, shrinking the share count and supporting the price.
Now, with the data-center build-out consuming more cash than the business generates, that flow is reversing toward issuing shares and debt — dilution instead of buybacks. Kedrosky doesn't argue a view on Oracle itself; it's the listed-company illustration of the same supply argument he makes about the coming mega-IPOs. More paper for sale, from every direction at once.
23:24Now it's even shifting from companies like Oracle, like buybacks versus share issuance and dilution. Historically, all those things are shifting, and shifting pretty hard in the other direction. — And what's interesting is it goes back again to the early part of our conversation about scale and systems. Let's think about it.
In short: The single most concentrated public exposure in the chain, restated with the number: "if you look at just Oracle, 50% of its $600 billion backlog is just from OpenAI" — half the contracted future revenue owed by the company he calls the weak sister of the two.
Oracle has won enormous contracts to host AI workloads, and reports them as backlog — work sold but not yet delivered or billed. Investors normally treat a big backlog as a promise of future earnings.
Eisman reframes it as a credit exposure to one customer: "50% of its $600 billion backlog is just from OpenAI." Half of what Oracle expects to earn depends on a private, unprofitable company continuing to be able to pay. That makes Oracle the most concentrated public expression of the risk he has been describing — and, usefully for anyone watching, the place where trouble at OpenAI would become visible in a listed share price first.
5:14If you look at just Oracle, 50% of its $600 billion backlog is just from OpenAI. So, in a sense, the way I think about this is the entire AI ecosystem food chain is dependent upon the future health and success of Anthropic and OpenAI. And by the way, that also means that the US economy hinges upon the health and future success of Anthropic and Open AI.
In short: Named with Meta as an issuer of 6-8% long-dated corporate paper to fund AI capex, and as one of the large data-centre operators whose contract server builders "have already notified their customers of the forthcoming increases" from Nvidia. Part of the same crowding-out loop: the capex that is supposed to grow the economy out of the debt is being financed by taking money away from the long end of the curve.
Oracle sits in the same group: a company funding enormous AI data-centre construction by issuing long-dated bonds at 6-8% rather than out of profits, and one of the large operators whose server suppliers have already warned about Nvidia's price increases.
Singh's interest is structural rather than in the stock. Each of these bond issues competes directly with the US Treasury for the same pool of long-term savings, and the more the technology companies borrow to build, the higher the government's own borrowing costs go. Oracle is one exhibit in that argument.
Full passage: premium transcript (PDF).
In short: Guest: named in the legacy marketing-cloud group Zeta is taking wallet share from.
In short: Two roles, pulling opposite ways. Trennert uses it as evidence for the dilution thesis: "all the hyperscalers are issuing debt now. Oracle was just downgraded." Verrone uses it as the sentiment marker instead — the attitudes toward the hyperscalers "reached a fever pitch of negativity maybe two or three weeks ago… When Oracle got to like 120" — i.e. the low print is the emotional extreme, not the fundamental verdict.
Oracle is the most leveraged expression of the AI build-out among the big names — it is borrowing heavily to build capacity it has already contracted out, and it was "just downgraded," meaning a ratings agency judged its debt riskier than before.
Verrone uses the same stock for the opposite purpose. When sentiment toward the hyperscalers "reached a fever pitch of negativity" a few weeks earlier, Oracle at around $120 was the marker of the extreme. His framework treats that as an emotional low rather than a fundamental verdict — the point at which the marginal seller has finished selling.
Both readings can be true, which is why this lands neutral: the balance sheet is the most stretched in the group, and the price already reflects a great deal of that.
22:33I think what's interesting is this rotational market we speak of is rotational at the sector level as well and it's so sentiment driven. The attitudes towards some of the hyperscalers reached a fever pitch of negativity maybe two or three weeks ago and — When Oracle got to like 120 — And you think about this a little bit.
In short: The named exception to his "the big hyperscalers are probably okay" line. "There are a few of the hyperscalers that look a little bit dicier" — and asked whether he means Oracle, the most indebted relative to revenue: "Yeah, Oracle. In hyperscaler terms there's the rest and then Oracle. Oracle is not a tiny bad company or anything like that, but it doesn't have nearly the financial and corporate heft of the others. And it's clearly the weakest of the litter."
Wigglesworth's general verdict on the giants is reassuring: Meta, Alphabet and Amazon have enormous real businesses underneath the AI spending, so even a total write-off would be survivable. Oracle is where he stops saying that.
The problem is proportion. Oracle is carrying the most debt relative to its revenue of any of the hyperscalers, without the cash-generating ballast the others have. In his words, "in hyperscaler terms there's the rest and then Oracle" — he is careful to add it is "not a tiny bad company," but "it doesn't have nearly the financial and corporate heft of the others. And it's clearly the weakest of the litter."
The way to hold this view is as a ranking rather than a prediction: in a debt-financed capex cycle, the losses concentrate in whoever borrowed most against the least cushion. Oracle is his nomination for that position among the majors.
50:33There are a few of the hyperscalers that look a little bit dicier, but broadly speaking, they're probably okay. Yeah, you perhaps referring to Oracle, definitely the most indebted relative to its revenue. — Yeah, Oracle, in hyperscaler terms there's the rest and then Oracle. Oracle is not a tiny bad company or anything like that, but it's just it doesn't have nearly the financial and corporate heft of the others.
In short: Third of the host's software examples. Same group-level read — software bottomed first and is pushing off the lows; the sector's weakness rotated on to semi-cap equipment and then memory, and "one by one all of these areas have stabilized."
The third software name raised by the host as having been sold off hard and since recovered. Again, Newton answered at the group level rather than on Oracle specifically.
The takeaway is that software's decline came first and its repair came first — a rolling, one-group-at-a-time sequence through technology that, in his reading, is much healthier than everything breaking at once. No individual view on Oracle's chart was given.
13:27Initially we saw the software deterioration that started actually late last year. That has since bottomed. We've seen software start to push off the lows and then of course it shifted to more semiconductor semi-cap equipment weakness and then the final shoe to drop was really a lot of these memory names and one by one all of these areas have stabilized including the mag seven, the hyperscalers.
In short: Counted among the hyperscalers with "real businesses," but singled out as the most counterparty-concentrated of them: "of Oracle's 600-plus billion backlog, around half is from OpenAI." That makes it the cleanest public read on the LLM layer's health — and the most exposed if the layer cracks.
A "backlog" is contracted future revenue — work a customer has committed to buy but has not yet been billed for. Investors treat it as visibility into future earnings, and it is usually a comfort. Eisman turns it into the risk.
"Of Oracle's 600-plus billion backlog, around half is from OpenAI." Half of the promised future business depends on the financial health of one privately held company that is losing money. If that customer cannot pay, or renegotiates, the backlog shrinks — and with it the future revenue already reflected in the share price. He does not call Oracle a bad business; he counts it among the hyperscalers with real businesses. He simply identifies it as the place where the LLM layer's problems would show up first and most visibly in public markets, which makes it as much a monitoring instrument as a stock.
4:53This is Steve talking about the LLM providers, both OpenAI and Anthropic. — If the lack of moats begins to cause them problems, then the entire AI ecosystem could go through a correction phase because so much of the hyperscaler backlogs are from these two companies. For example, of Oracle's 600-plus billion backlog, around half is from OpenAI.
In short: "Oracle is questionable with its debt barely rated above junk" — the one hyperscaler he excludes from the winners. The concentration compounds it: "Oracle has a $600 billion backlog and half that backlog is from OpenAI alone," i.e. a leveraged balance sheet whose order book leans on a loss-making private counterparty.
Oracle is the hyperscaler he pointedly leaves out of the winners. Two problems stack on top of each other. First, the balance sheet: its debt is "barely rated above junk," meaning lenders already consider it near the low end of investment grade — thin room for error if conditions tighten.
Second, customer concentration: of Oracle's ~$600 billion backlog of promised future cloud revenue, roughly half is OpenAI. So the company with the least financial cushion has staked the largest share of its order book on a single loss-making private customer that funds itself by raising capital. A big backlog and a concentrated backlog look identical on a headline slide and are completely different risks.
12:49This bare case of entropic and open AI could occur but it might occur a year from now and for markets that is an eternity. Winners and losers are beginning to emerge amongst the big companies. Microsoft, Google and Amazon seem to be the hyperscaler winners. Oracle is questionable with its debt barely rated above junk. Meta is having problems.
In short: Market color: up 3.5% last week in the software complex — the laggard of the group Wapner ran through. No individual committee call this episode.
In short: The credit-market tell inside the capex boom: hyperscalers are set to spend more on capex than operating cash flow in 2027, so their ability to spend depends on debt and equity markets — "which is why CDS for companies like Oracle is blowing out, which is the credit risk the market is pricing for them raising debt."
Oracle is the warning light in the AI build-out. Because the big cloud companies will spend more on data centres in 2027 than they earn, they have to borrow the difference — and the bond market prices how risky that borrowing is through credit default swaps, essentially insurance against the company failing to pay.
Oracle's are "blowing out," meaning that insurance is getting rapidly more expensive. Capital markets are still open and funding the spending, Singh says, "but at an increasingly higher price." Oracle is where that price is rising fastest.
Full passage: premium transcript (PDF).
In short: Named by the host at the head of the hyperscaler list. The only view attaching to it is the cohort view: ~10× EV/sales on average with a step-function capex funded out of all their free cash flow and more — "a lot is going to have to go right for that multiple to be roughly correct."
3:44So Dean, you said that it's bullish for the companies that are not the frontier labs, but what about just the layer directly above the frontier lab? So the hyperscalers who are spending it. So that's Oracle, Microsoft, Google, Amazon and let's see who else am I missing? Meta, which announced that they actually are going to sell their excess compute.
In short: Named (with Microsoft and Anthropic) among the customers lined up for AMD's Helios full-rack AI systems after the first Meta and OpenAI deployments — another data point in the neocloud/hyperscaler demand queue AMD must convert into "tens of billions" of Data Center AI revenue by 2027. (Referenced; not a stance call.)
In short: A hyperscaler with "real businesses," but the exhibit for concentration risk: "so much of the hyperscaler backlogs are from these two companies. For example, of Oracle's 600 plus billion backlog, around half is from OpenAI." That is why Anthropic/OpenAI health is the catalyst he watches. (Gil Luria made the constructive case on the same backlog four days earlier — see Ep 70.)
Oracle is the fourth hyperscaler in his list, and the illustration of the systemic risk he now watches. Its order book — contracted future revenue — is over $600 billion, and "around half is from OpenAI."
That single fact links the whole chain: if the model companies get into trouble, the backlogs of the data-centre companies stop being assets. It is the reason his selloff trigger is not a stock or an economic number but the financial health of two private companies.
Worth reading alongside Gil Luria's argument on the same backlog four days earlier: Luria's point was that the market values that backlog at zero, making the shares attractive. Both can be true — a backlog valued at zero is cheap and concentrated in one customer.
7:50If the lack of moes begins to cause them problems, then the entire AI ecosystem could go through a correction phase because so much of the hyperscaler backlogs are from these two companies. For example, of Oracle's 600 plus billion backlog, around half is from open AI. On the other hand, AI is allowing the creation of software and other tech that is much cheaper than existing software and tech.
In short: Carried forward as the anchor of the credit thread rather than re-argued: this whole section is written "as a follow-up to last week's ORCL note" (the Jul-21 "Emerging Credit Risk with Contagion Potential Revisited"). The new-issue-vs-secondary regime change, the pulled data-center deals and the T+285 "Sopaipilla" print are offered as the market-structure evidence around the ORCL downgrade path — "I have been really concerned about credit — particularly IG credit."
Oracle isn't re-argued in this note — it is the reason the note exists. The whole credit section is written "as a follow-up to last week's ORCL note," where Paulo laid out how Oracle, having borrowed heavily to build AI data-centre capacity for OpenAI, was downgraded to the bottom rung of investment grade with its debt costs already trading like junk.
What this note adds is the surrounding weather. Rather than Oracle-specific news, he brings market-structure evidence that the whole investment-grade market has changed character: issuers are having to offer bigger discounts to get bonds sold, order books are thinner, planned July supply came in at less than half the original estimate, and data-centre bond deals are being pulled outright because the interest cost no longer clears the project's returns. One supposedly high-quality data-centre borrower had to pay 2.85 percentage points over Treasuries to get done, which reset pricing across the sector by roughly half a point.
That is the point for Oracle. Its downgrade path only becomes dangerous if the market that has to absorb its paper is deteriorating at the same time — and that is precisely what the new-issue statistics now say. Oracle's problem is being confirmed by the market it needs, not contradicted by it.
In short: Luria — who called the Sep-2025 top ("I'm always worried when everybody's on a bandwagon") — now takes the other side: OpenAI raised $122B, "the largest fund raise in history," went into "code red" and narrowed to compute, so "they are going to pay their Oracle bills… their entire backlog, $630 billion worth of backlog of compute revenue is valued by the market at zero." Ives: "or you could almost say negative."
Oracle rents out computing capacity to AI companies, above all OpenAI. Its share price has been a round trip — about 200, then 330 on a single enormous backlog announcement in September 2025, then down to about 140 when it emerged that most of that backlog was one deal with a customer that had almost no money.
Luria was the analyst who doubted the 330 and is now the analyst arguing the 140 overshot. OpenAI has since raised $122 billion — "the largest fund raise in history" — and cut its spending back to compute only, which means Oracle's bills get paid. The asymmetry he points to: the market currently assigns Oracle's entire $630 billion of contracted future compute revenue a value of zero. You are paying for the old Oracle and getting the AI backlog for free. (Eisman's own wrap four days later adds the caveat: roughly half that backlog is a single customer.)
31:14At that point it became clear wait a second they are going to pay their Oracle bills and therefore Oracle actually has a chance of being worth a lot more because their backlog is really being valued at zero. And as we sit here today, their entire backlog, $630 billion worth of backlog of compute revenue is valued by the market at zero zero.
In short: The AI-credit canary: "Oracle bonds are trading almost like junk" — remarkable for a company that "at one point got up close to a trillion dollar market cap" but "really leveraged up… trash balance sheet. The market's kind of unforgiving." He recalls it adding ~$200B of market cap in a day on its AI-spend announcement a year ago: "when you see that kind of stuff on the upside… almost invariably you're going to get the payback on the downside."
Oracle is the database and cloud giant that borrowed heavily to build AI capacity. The warning sign Hay points to isn't the share price — it's the bonds. "Oracle bonds are trading almost like junk," meaning lenders now demand nearly the yield they'd want from a speculative-grade borrower, which is remarkable for a company that briefly approached a trillion-dollar market value.
The reason is the balance sheet: it "really leveraged up… trashed balance sheet. The market's kind of unforgiving." Bond investors get paid only if they're repaid, so they tend to notice strain before equity investors do — which is why he treats the credit market as the early-warning system for a bubble.
He also draws a symmetry lesson. A year ago Oracle added roughly $200 billion of market value in a single day on an AI-spending announcement, back when spending was cheered. "When you see that kind of stuff on the upside, almost invariably you're going to get the payback on the downside."
47:00We are starting to see some concerns that the debt financing might not be there. Correct. — For sure. It's getting very costly. and Oracle bonds are trading — almost like junk and that's Oracle which has a huge market cap and but they really leverage up of all the I guess you wouldn't call it a magnificent seven maybe was a magnificent 10 but at one point it got up close to a trillion dollar market cap it's come way down of course but it's hard to believe that its bonds would be trading as as inexpensively as they are right now
In short: The 1990s version of his enabled-vs-enabler rule: "if you can find a non-tech company that understands how to use an Oracle database… pretty much just buy the stock" — the banks were among the first adopters and it added basis points to ROE. Framework analogue, not a view.
53:39that we do work on is so wide it's at one of the widest junctures I've ever in my career. And so I think they these companies when I grew up in the 1990s I would have told my bosses when I was an analyst if you can find a non- tech company that understands how to use an Oracle database or a cybase database pretty much just buy the stock right and the banks were first one of the first adopters of database technology and it added basis points to ROE and in my career if you can find stocks that have improving ROE
In short: The credit-contagion focus, now "tipping over": S&P downgraded ORCL to BBB- (one notch above junk); debt grew $100bn → $120bn (coupon 4.4% → 4.8%), mcap $500bn → $365bn, equity "collapsed below the March lows." Its 5Y CDS blew out to ~2× the BB CDX — bond ~200bps wide vs IG ~117 and wider than BB corp at 154, "on its own axis." A downgrade to junk = ~7% of the entire HY market + a 7.6yr-vs-2.9yr duration mismatch "air pocket." "The market as usual is running well ahead of the agencies."
This is a bond/credit call, not an equity trade — and the equity has already broken (Oracle's stock is below its March lows and its market value fell from over $500bn to $365bn). Oracle is borrowing enormous sums to build AI data centres, largely to serve OpenAI (its "Stargate" compute deal is worth $300bn+, and Oracle plans to spend at least $90bn next fiscal year). Its bond pile has grown from $100bn to $120bn and the interest it pays has crept up (4.4% to 4.8%).
Here's the tell Paulo is watching. A "credit default swap" (CDS) is insurance against a company defaulting — the higher its price, the more the market fears default. Oracle is officially rated investment-grade (safe), yet its CDS has blown out to about twice the level of the "BB" junk-bond index, and its bonds now trade wider (riskier) than the average junk-rated company. In other words, the market is already pricing Oracle as junk while the ratings agencies still call it investment-grade — "the market runs ahead of the agencies," and S&P has already cut it one notch to BBB- (the last rung before junk).
The real danger is what a downgrade to junk would do. Oracle's $120bn of debt would suddenly be ~7% of the entire junk-bond market — a flood of paper that junk-fund managers would have to "make room" for by selling other holdings, right when spreads are historically tight. Worse, Oracle's bonds are unusually long-dated (7.6-year average vs 2.9 years for the junk market), so they'd have to be dumped at a steep discount — an "air pocket" that crushes their price and drags spreads wider across all of high yield, into dealers who are already stretched. That's the contagion: one downgrade tipping the whole credit cycle into another leg down, and the "Who's Next?" fear spreading to other debt-heavy hyperscalers.
In short: Cautious/avoid: "Oracle's problem is they have too much exposure to OpenAI, and that's a much different problem" — with OpenAI "definitely losing market share, I'm a little bit more worried about Oracle... I wouldn't add that right away."
Singh is wary of Oracle and "wouldn't add it." Its specific problem is being too dependent on OpenAI — and with OpenAI "definitely losing market share" to cheaper rivals, that reliance is a risk. A name to avoid, not a short per se.
Full passage: premium transcript (PDF).
In short: "Oracle has the worst of those ROIC metrics I just talked about." Near-junk raises; stock down ~50% (65–70% off the OpenAI/AMD-deal spike) — the group's cautionary tale the other CEOs study.
Oracle went all-in on AI cloud deals (like the huge OpenAI/AMD arrangement) and its stock spiked, then collapsed — down around 50%, and 65–70% off the peak. Chanos says it has "the worst" return-on-new-spending numbers of the big cloud players.
He thinks the other cloud CEOs are watching Oracle as a warning: they don't want their own stocks punished like that, so they'll eventually pull back on spending. If even the giants are only earning ~10% on new investment, the smaller "neo-clouds" are "in a whole world of hurt."
36:34What do you think those CEOs are looking at Oracle? And is there anything that they can extrapolate a little bit because to me it makes — It just Oracle has the worst of those ROIC metrics I just talked about. And so those other guys, if they're doing their jobs and I think they're doing their jobs, they're looking at Oracle saying, we don't want to be down at those levels.
In short: The most stretched: spending 100% of revenues on data centers, debt is 2.5× last year's sales, stock −28% YTD and −60% off its highs. He believes some of these debt-laden names end up going bankrupt.
Oracle is the most dangerously stretched name in his view. It's plowing 100% of its revenue into building data centers and has piled on debt equal to 2.5× its annual sales. The stock is already down 28% this year and 60% from its highs. He thinks some of these heavily indebted AI-infrastructure companies actually go bankrupt if the revenue they're betting on doesn't arrive — and Oracle is his prime candidate.
18:19And yet, economically it's not working. Oracle are spending 100% of their revenues right now on their data centers. Amazon and Meta are spending 100% of their cash flows on these data centers. So if pricing is collapsing here and you're not going to get the revenues, and I worry very much about what OpenAI, I mean Anthropic, will end up doing here because they require lots of cash, that's why OpenAI and Anthropic want to have an IPO,
In short: At a new 52-week low. Sethi (owns it): sold about half after last year's news-driven pop, holding the rest — Oracle can still make the transition and has earnings power, but it's spending into the business like the Fab-7 and "gets no credit for it at all"; painful, but he's holding at these levels. Lebenthal ties the recent weakness to debt-market indigestion from monumental offerings (Amazon, SpaceX's ~$25B deal, 7%+ yield-to-worst on the long end) — Oracle has leaned on debt for its data-center buildout.
Oracle sells business software and, increasingly, cloud computing for AI. The stock just hit a new 52-week low. Sarat Sethi owns it: he sold about half after last year's big pop and is holding the rest, still believing Oracle can complete its transition into a cloud powerhouse — but the market is giving it "no credit" while it spends heavily to get there, like the other big AI spenders. It's painful, but he's holding.
Jim Lebenthal offers a technical reason for the recent slide: a flood of huge corporate bond sales (from Amazon and SpaceX, among others) is more than the debt market can easily absorb, pushing borrowing costs up across the board. Because Oracle has borrowed heavily to build its data centers, it gets caught in that "indigestion."
In short: "The first crack in the AI bull case" — an October backlog surge took the stock 230→330, then it corrected once analysts saw "50% of that backlog was solely from OpenAI"; now "even lower at 140." Also raised capital as its capex need ballooned.
Oracle was "the first crack" in the AI story. It reported a huge jump in backlog (future contracted revenue) last October and the stock ran from 230 to 330 — until analysts realized half of that backlog came from a single customer, OpenAI. Depending on one customer makes the revenue far riskier, and the stock corrected below where it started; it now trades around 140.
On top of that, Oracle had to raise capital as its data-center spending ballooned — the same "cash machine turns capital-hungry" pattern Eisman flags across the hyperscalers.
10:12Oracle reported thirdarter numbers in October and showed a massive increase in backlog. The stock went from 230 to 330 in a few days. But then analysts figured out that 50% of that backlog was solely from OpenAI and that made the market nervous and the stock corrected to below where it was when it reported.
In short: Bryn's enterprise-AI question mark: "if you think Oracle, that's a big question — they are tied to OpenAI." A name whose winner/loser status is hard to call in the murky enterprise layer.
In short: Weird/messy chart, like software — the low might be in but you have to respect it; "a lot of better opportunities out there" (hardware, semis). The 200-day is rolling over (downward to flat); messy is super frustrating in a market making new highs. (Neutral with a negative tilt.)
Oracle's chart looks "weird," much like software. The recent low might hold, but you'd have to "respect it" (use a tight stop), and Sohn would rather put money into clearer winners — hardware and semiconductors — where the trend is obviously up.
The specific red flag is that Oracle's 200-day moving average is "rolling over" — turning from up to flat or down — which signals the longer-term trend is weakening. A messy, sideways chart is especially frustrating, he notes, when the broad market is busy making new highs. Net: a neutral read with a negative lean.
8:43ORCL (Oracle) — a weird-looking chart, like software. The low might be in but you have to respect it; I'd take a flyer perhaps, but there are a lot of better opportunities (hardware, semis). The 200-day is also rolling over, downward to flat — messy. Messy is super frustrating in a market making new highs.
In short: "The worst off" of the group: weakest free cash flow, credit-default swaps back on the rise, and laying off staff (headcount ~141k vs ~160-162k a year ago, "going to fire more") to fund the AI buildout.
Oracle is "the worst off" of the group: it has the weakest cash flow, its debt-insurance cost (credit-default swaps) is rising again, and it's cutting staff — headcount down to ~141,000 from ~160,000 a year ago, with more layoffs coming — all to fund its AI data-center push. A name on the wrong side of the capex-burn trade.
Full passage: premium transcript (PDF).
In short: Also among the hyperscalers down Thursday despite the chip/power rally — part of the tech-leadership rollover.
9:46However, and perhaps more importantly, the rally on Thursday was confined largely to chips and power related companies like GE Vernova. The hyperscalers, Amazon, Meta, Oracle, Google, and Microsoft were all down on Thursday. Apple also was down as it raised prices on its MacBook and iPad because of higher memory chip prices.
In short: Lebenthal's high-conviction believer — doubling down; "taking a risk" (negative FCF, indebted) but the hyperscaler CapEx wave that propels Micron also benefits Oracle. "I might be a little early."
Oracle is a database and cloud-computing company that has reinvented itself as a major provider of AI cloud capacity. Lebenthal is a high-conviction believer and is adding even though he admits "I might be a little early." The bull case: the same enormous AI build-out that is driving Micron also fills Oracle's order book, and the hyperscalers all insist that AI compute is profitable.
The risk he openly acknowledges is the balance sheet — Oracle is spending so heavily it has negative free cash flow (it's burning more cash than it generates after investment) and carries a lot of debt. His view: to earn an outsized return you sometimes have to take that kind of risk, and he believes the spending pays off.
In short: "Increased its capital raising plans by 20 billion recently" — another data point that non-capital-intensive software has become capital-intensive hardware.
7:54In 2026, Google will be spending 180 to 190 billion on AI capex and that is just too much for its cash flow. There are also stories that Meta and Microsoft will be doing similar transactions soon. Oracle increased its capital raising plans by 20 billion recently. This all goes to show that certain non-capital-intensive large software companies have now become capital intensive hardware companies.
In short: Reference/comp, not a stance: SPCX's expected $20-25bn IG bond is seen pricing ~25bps tighter than ORCL ("I would rather lend to SPCX than Larry Ellison"); separately, MSFT walking from a $3bn ORCL cloud-leasing deal is offered as datacenter-capex-cracking evidence.
Oracle is a big enterprise-software and cloud-datacenter company. Paulo doesn't give it a buy/sell view — he uses it two ways as a reference point. First, as a yardstick for how cheap SpaceX's coming bond deal is: SpaceX is expected to borrow about 0.25% cheaper than Oracle does, which says more about credit-market froth than about either company. Second, the news that Microsoft walked away from a $3 billion deal to lease Oracle cloud capacity is, to him, a small crack in the AI-datacenter building boom.
In short: "The Mag 7 plus Oracle… those stocks are actually down 2% since October" — the digestion problem; and if the bankers are right, SpaceX's $1.4T of future revenue has to come out of this same group's cash flow.
13:28It's predicted — the share of PE ratio suggests when you get to these valuations, the expected forward returns for 10 years are in fact that. You think that's going to play out? — Right. No, 100% and guess what? You look at the CapEx of the Mag 7 plus Oracle, those big stocks, those stocks to your point are actually down 2% since October.
In short: Sold off hard: capex ~$5.7B above plan ($15.9B in the quarter; $56B annualized vs $50B) and it needs to raise ~$40B of equity and debt — "surprised Ellison is diluting and raising so much." Decent earnings ignored.
Full passage: premium transcript (PDF).
In short: Beat on revenue/EPS (Q4 FY26 rev +21% to $19.2B, Cloud +47%, OCI +93%) and grew RPO +363% Y/Y to $638B — yet the stock fell 12%: the market "stopped paying for backlog and started pricing the cost to deliver it." FY27 guides ~$90B rev but ~$70B capex + gross margins stepping down + a fresh ~$40B raise (net debt $124B). FCF was −$24B in FY26. "Now a bet on execution."
Oracle sells database software and, increasingly, cloud computing (renting out servers, especially for AI). The headline number everyone watched was "RPO" — remaining performance obligations, which is the dollar value of contracts customers have signed but Oracle hasn't billed or delivered yet. Think of it as the order book. That order book jumped 363% in a year to $638 billion, an enormous signal of future demand.
So why did the stock drop 12%? Because the market shifted from celebrating the order book to worrying about the cost of filling it. To deliver all that AI cloud capacity, Oracle has to spend roughly $70 billion building data centers, raise about $40 billion more in debt/equity, and accept that its profit margins will "step down" because renting raw computing power is a lower-margin business than selling software licenses. Its free cash flow was actually negative $24 billion last year as construction outran the cash coming in. App Economy's read: the backlog is no longer the question — execution is. Can Oracle actually build and operate all this profitably without the financing strain biting? "Now a bet on execution."
In short: EPS $2.11 (+24%, beat) and an "incredible" $638B RPO backlog (+363% y/y, +$85B in 3 months) — "all that's good." But capex hit $55.7B (above its $50B plan) and it added $20B to capital-raising plans (FY27: $40B equity+debt). Down 10% after hours. The capital-intensity story.
Oracle's results were genuinely strong: profit up 24%, and a huge "backlog" (signed-but-not-yet-delivered cloud contracts) of $638 billion, up more than fourfold from a year ago. That's the good news.
The catch is the cost of fulfilling it. Oracle's capital spending ran to $55.7 billion — above its own plan — and it just added $20 billion to how much it intends to raise. So even a company with a spectacular order book is being forced into the same "insatiable need for capital," which is why the stock fell 10% after hours. Eisman uses it as Exhibit B (after Google) that AI is turning tech "capital-light into capital-heavy."
6:26Oracle added to the capital intensity story when it reported this week. Now on the positive side, the company reported earnings per share of 2.11 which was up 24% versus last year and a beat and revenue beat as well. Also, the company's remaining performance obligations which is a form of backlog reached a pretty incredible $638 billion which is up 363% versus last year and up 85 billion in just 3 months. So all that's good.
In short: Centerpiece of the circular-funding graphic (Bloomberg/Burry): OpenAI invests in Oracle while Oracle funds OpenAI's data centers.
39:46OpenAI's data centers and AMD is part of this circular funding as well and you have Micron and Nebius and Cororeweave which effectively is just a conduit that buys Nvidia GPUs and builds its own data centers., their names like Figure AAI which are private where you don't even know what the financials are and Microsoft and Honey AI.
In short: His "industry standard" database-moat example; separately cancelled a big AI compute deal that Meta then took.
33:13You've got the network effect that was so powerful for LinkedIn, Facebook, Alibaba, you name it. Then you can become an industry standard. Oracle and Bloomberg are the industry standard. Oracle, you know, they charge a lot and, you know, there's free versions, there's open- source Oracle, but they had all the database administrators.
In short: "Incredibly meaningful to the AI story" (selling cloud infrastructure). Reports this week — great headline numbers are a given, but it must raise guidance; merely reiterating would make it one more name decelerating vs expectations.
Oracle is a big enterprise-software and cloud-infrastructure company that has become central to the AI story. It reports results this week, and he says strong headline numbers are basically guaranteed — so they won't be enough. The real test is its forecast ("guidance"): Oracle needs to raise its outlook. If it merely repeats the same forecast as last time, the market will treat it as one more AI name whose growth is slowing, which is exactly what spooked investors about Broadcom.
5:52And unfortunately, it's going to get more volatile later this week. Oracle's reporting earnings this week, and this is another important one. Oracle needs to not only post great headline numbers, which I'm sure they will, but they need to have those great headline numbers in addition to guidance.
In short: The fifth hyperscaler in the combined-CapEx tally that declined vs Q4; memory-chip cost rises imply volumes fell even more.
Oracle rounds out Woo's five cloud giants. Beyond the group's spending dropping versus last quarter, he notes memory-chip prices jumped — so the amount of actual hardware bought probably fell even more than the dollar figures suggest. Again, evidence for the cooling-AI thesis rather than a stock view.
6:52The AI trade has always been driven by AI-related capital expenditures by the hyperscalers. Exactly what they're buying more of at any given moment may change, but the underlying driver remains the same. To me, the most important takeaway from the first quarter earning season is the fact that the combined CapEx spending of Microsoft, Amazon, Alphabet, Meta, and Oracle fell compared to the fourth quarter.
In short: The extreme of the off-balance-sheet commitment problem (Morgan Stanley, "AI: Off Balance Sheet, On the Hook"): lease + purchase commitments running over 700% of forward operating cash flow (vs >100% for GOOG/META/AMZN/MSFT). "Most obligations remain off balance sheet." Consistent with his standing datacenter-credit-posterchild view; cited as evidence here.
Oracle is the most stretched example of a related problem: it has signed contracts to buy and lease AI infrastructure worth more than seven times the cash its operations generate in a year. Most of those obligations don't even show up as debt on the balance sheet yet, because accounting lets companies delay recording them until the gear is delivered or a payment becomes "probable." Paulo treats Oracle as the clearest sign that the AI build-out is being financed with promises the cash flow can't currently cover — and notes those very contracts are being used to back private-credit loans, so the risk quietly leaks into the credit market too.
In short: Named once, in the customer list — "Microsoft, Meta, Google, Oracle,…". No view.
In short: Lumped with the Mag 7 as "technology basically 50% of the S&P" — extreme passive concentration (vs financials ~27% in 2007).
Oracle is a big software and cloud company. He lumps it with the Mag 7 to make a concentration point: together, technology is now roughly 50% of the entire S&P 500.
For comparison, financial stocks were only ~27% of the index right before the 2007 crash. So today's tech weighting is an even more extreme, dangerous concentration — and if you own the index, you're heavily exposed to it whether you realize it or not.
7:59with artificial intelligence. You know Gary let's get into your point. Around this IPO cycle. So You've got the mag seven and Oracle and technology are basically 50% of the S&P 500's composition roughly okay like That's two maybe one and a half to two standard deviations outside the norm like if you look to look I went I wrote the Lehman book our first book our by the way our two books colossal failure common sense and when markets speak about a million readers our new one is called when markets speak and we get
In short: "Look at the drawdown at Oracle" — overinvested; the data-center buildout is going slower than the market thinks.
Oracle is the database/cloud company spending big on AI data centers. He points to its sharp stock drawdown as evidence the data-center buildout is happening more slowly than the market assumes — leaving heavily-invested names exposed.
44:29Mag 7 is not at record highs. You know, that's because Microsoft's unchanged I think for almost 2 years. Meta is unchanged for a year and a half. These companies have over Look at Look at Look at the Look at the drawdown at Oracle. These companies have overinvested and now China's picking them apart. — Well, China's Foreign Ministry is denying this.
In short: "Problem child ORCL's CDS continued to tick higher despite Friday's bounce in the stock" — the credit market pricing rising default risk on the name he's called a significant, misunderstood datacenter-capex contagion risk.
In short: The credit-contagion focus (not an equity stance): ORCL is rated BBB / negative outlook at S&P and Moody's, with ~$101bn of bonds (+$3bn loans). Its 5Y trades 162bps and CDS 154 vs BB cash at 174bps / BB CDX ~115, and the new 30Y is ~275bps over Treasuries — "definitely getting into BB territory." A downgrade to BB+ would add $101bn to the ~$1.7T HY market (+6% overnight), forcing HY managers to "make room" — a flow shock that could break record-tight spreads.
Junk-bond ("high yield") interest rates over safe bonds are near the lowest they've ever been — which normally signals calm — yet company bankruptcies are quietly surging. Paulo's argument is that the calm is fake: it isn't because companies are healthy, it's a plumbing quirk. Because private-credit funds now do the lending that used to happen in the junk-bond market, there are very few junk bonds to buy, and lots of money chasing them — so their price is bid up and their yield squeezed down. Scarcity, not safety.
His insight is about what could break that. Everyone watches for more defaults. Paulo says watch for a supply shock instead: if a big, investment-grade company gets downgraded to junk, its bonds suddenly flood the junk market. Oracle is his candidate — it's rated one notch above junk with a "negative" warning, its newest bonds already trade like junk, and it has about $101 billion of debt. The entire junk market is roughly $1.7 trillion, so an Oracle downgrade would grow it about 6% in a single day.
Junk-bond fund managers would suddenly have to "make room" — sell other bonds to absorb the flood — right as nervous investors stop adding money. That reverses the scarcity that was holding everything together, and yields could blow out. He also flags a rumored $12 billion hedge fund that's borrowed heavily to bet the calm continues; a modest move against it would wipe it out — a 2007-style accident waiting to happen. Note this is a view on Oracle's debt and on credit markets, not a call to short the stock.
In short: A blue chip whose bonds got repriced after the AI-CapEx hike — credit markets flashing before equities.
Oracle is the database/cloud giant now spending heavily on AI data centers. His warning isn't about the stock price — it's the bonds. After Oracle hiked its AI spending plans, lenders demanded higher returns to hold its debt (the debt "repriced"). He reads that as the credit market flashing caution about the AI buildout before the stock market does.
8:08I want to turn to the credit markets. We've discussed Oracle's bonds sliding after the CapEx hike, but you've also flagged a major divergence in the financial sector — while Goldman Sachs is up sharply, private credit players like Blue Owl are down significantly. Is private credit the subprime of the cycle? Well, I was at Lehman Brothers. I sat down in Palm Beach with a legendary credit investor who made over half a billion dollars during the subprime crisis. And he sees a massive crisis next year.
In short: Named only as the data-center customer Energy Transfer is building natural-gas pipelines to serve (three sites, two in Texas) — a demand-side reference, not an investment view on Oracle.
In short: The datacenter-credit posterchild: CDS blown out for weeks; debt pushed off-balance-sheet into ABS ("opaque disintermediation"); Deutsche shows ORCL as a huge outlier. "May yet end up as the WorldCom" of this cycle's datacenter-SPV insanity — "the numbers will not math."
Oracle went all-in on building AI data centers, largely financed with debt. Paulo sees it as the poster child for the riskiest part of this cycle. The cost to insure Oracle's debt against default (its "CDS") has spiked, and a lot of the borrowing has been moved off Oracle's own books into complex, hard-to-see financing structures. He compares it to WorldCom — a company whose aggressive accounting and debt eventually blew up — and says "the numbers will not math": the promised revenue doesn't justify the spending. A name to avoid, and a warning for the whole data-center-credit complex.
Nothing matches this filter.
Verbatim excerpts from the public transcripts (auto-pulled at each mention's timestamp, lightly cleaned). Timestamps link into the video; "source page" opens that commentator's full analysis at this row. Click a mention's header line to expand it (one open at a time). For personal study — not investment advice.