THW require AI labs and hyperscalers to consolidate all off-balance-sheet compute infrastructure debt onto their parent balance sheets.

PRO — Opening · Champ (xiaomi/mimo-v2.5)

Here is the motion: that AI labs and hyperscalers should be required to consolidate all off-balance-sheet compute infrastructure debt onto their parent balance sheets. I rise in favour, and I will show that this is not an abstract accounting debate — it is a question of whether the largest capital allocation in a generation is being conducted with the transparency that investors, markets, and the public deserve.

Start with the scale. Five US tech giants are hiding an estimated $1.65 trillion in debt that does not appear on their balance sheets, exceeding the $1.35 trillion that does (article 2). Meta alone carries roughly $420 billion in off-balance-sheet obligations — nearly triple its reported debt (article 2). When an investor reads quarterly earnings, they are seeing less than half the leverage (article 3). That is not transparency. That is the architecture of opacity.

The mechanism is the variable interest entity and special purpose vehicle. Meta formed a joint venture with Blue Owl Capital for its Louisiana Hyperion data center, with maximum exposure at $46 billion and total project cost up to $250 billion. The VIE took on $27 billion in debt, and Meta is the sole tenant — yet Meta argues it need not consolidate because it is not the “primary beneficiary” (article 1). Alphabet, Microsoft, and Oracle use similar structures to keep vast obligations off their books (article 1). Oracle carries $260 billion in future lease commitments; Nvidia carries $119 billion in purchase obligations (articles 1 and 3). This is not a quirk. It is a pattern — and the pattern is the same one that toppled Enron.

Proponents of the status quo will say: Enron was fraud. This is legal. The disclosures exist in footnotes (article 3). But that defence proves the case for the motion. If the footnotes are where a trillion-dollar risk is buried, and if retail investors — as the Nikkei investigation and Hacker News discussion both confirm — cannot easily recognise these risks (articles 2 and 6), then the current system is not serving its core purpose. The accounting rules permit companies to keep the debt off their books because they are not the “primary beneficiary,” but as one technical accounting consultant warns, the accounting treatment itself is in fashion, and the risk is that one of these companies could be a house of cards propping itself up with this treatment (article 1). When a data centre goes live, its lease rolls onto the balance sheet at once, and if AI demand falls short, the loss lands on lenders and insurers who funded it (article 3). That is not a private risk. That is systemic.

The systemic dimension is critical. Paul Kedrosky identifies a powerful incentive loop between structurally overcapitalised insurers, return-hungry private equity firms, and mega-cap companies trying to avoid looking like they are leveraging up (article 4). He compares current AI data centre spending to telecom spending during the dot-com bubble, noting that we are already spending more now than then, driven by the same sort of hyperbolic projections (article 4). The components are familiar: leverage hidden in plain sight, mispriced risk, and capital chasing yield through increasingly convoluted structures (article 4). This is not a GFC 2.0 — yet — but the motion asks us to act before it becomes one. Consolidation on balance sheets is the first line of defence.

I anticipate two objections. First, that SPVs serve a legitimate purpose: they allow companies to restructure or divest assets cleanly, as some in the Hacker News discussion argue (article 6, drob518). But the motion does not ban SPVs. It requires consolidation. A company can still use an SPV; it simply must show the debt on its own balance sheet. The flexibility to restructure remains. What disappears is the ability to obscure the obligation from investors. Second, that disclosure in footnotes is sufficient (article 3). But if disclosure were sufficient, the Nikkei investigation would not have shocked the market. The fact that the debt was “buried in the footnotes” (article 3) — and that four of five companies were set to report earnings while the $1.65 trillion in off-balance-sheet obligations would not make headlines (article 2) — demonstrates that disclosure without consolidation is disclosure in name only.

The pro-investment case is the strongest rebuttal, and it is the one I will take most seriously. Article 5 argues that hyperscaler capex is driven by real and growing returns, that cloud revenue growth requires hardware investment, and that operating cash flow will surpass $900 billion by 2027. But even if the ROI is real — and I do not dispute that the demand is genuine — the argument is not that these companies cannot afford the debt. It is that the debt must be visible. The fact that access to capital through off-balance-sheet structures is becoming a competitive moat (article 5) is itself a reason to consolidate. When the market prices a company’s credit based on $1.35 trillion of debt rather than $3 trillion, the market is mispricing risk — and mispricing risk is the seed of every financial crisis.

The motion is straightforward: require consolidation. Let the market see what it is buying. Let investors evaluate the true leverage. Let regulators assess the systemic exposure. The alternative — trusting that footnotes are sufficient, that VIEs are properly bounded, and that the AI boom will never falter — is a bet that no rational actor should ask the public to take. I urge you to support the motion.

CON — Opening · Challenger (tencent/hy3)

We oppose the motion. The pro case rests on a false equivalence: it calls $1.65 trillion in disclosed footnote obligations “hidden” (article 2), yet article 1 quotes analyst Gil Luria that “Enron’s crime wasn’t having special purpose vehicles. Enron’s crime was hiding them,” and article 3 states plainly that “none of this is illegal” and “the disclosures exist, buried in the footnotes, for anyone willing to dig.” Meta’s Hyperion VIE debt of $27 billion and its capped $46 billion exposure on a $250 billion project are detailed in public filings (article 1); the opacity the motion presumes simply is not there.

Consolidation would not clarify risk — it would distort it. Meta argues it need not consolidate Hyperion because it is not the “primary beneficiary,” so its real exposure is limited to $46 billion (article 1). Forcing the full project cost or SPV debt onto Meta’s parent balance sheet would overstate its liability by multiples and present a leverage picture the economics of the deal reject. Alphabet and Microsoft similarly keep VIEs off-book because they do not direct the structures (article 1); the law already requires consolidation where they do, making the motion a redundant override of post-Enron VIE judgment calls (article 1).

The structures serve legitimate finance, not mere cosmetics. Kedrosky explains SPVs exist because the capital needs are so large that orthodox balance-sheet debt would “make a mess of your balance sheet,” and they let firms raise money via “control without consolidation” (article 4). Hacker News contributor drob518 notes a plain benefit: an SPV lets a company “sell it all off or restructure it later without it being intertwined with your main books” (article 6). The motion does not ban SPVs, but by requiring consolidation it strips out the clean-divestiture and balance-sheet-relief functions that justify them, while leaving the entity intact.

The systemic alarm is overblown. Kedrosky explicitly says “Is this GFC 2.0? No, not yet… not systemic risk in the mortgage-backed sense” (article 4). The credit quality of the offtaker is “exceptional” (article 5), so lender default risk is low, and when a data centre goes live its lease rolls onto the balance sheet at once anyway (article 3). If AI demand falls short, the loss “lands on the lenders and insurers who funded it” (article 3) — private risk allocation, not public contagion. Oracle’s S&P downgrade proves markets already price the leverage (articles 1, 3), and Moody’s and Morgan Stanley have flagged the wider issue (article 3), so the information is in the system.

The pro case leans on the $1.65 trillion figure (article 2) while ignoring article 5’s showing that hyperscaler capex is “driven by commitments already made,” that ROI is compounding today, and that operating cash flow will surpass $900 billion by 2027. Forcing consolidation could trigger the stock hit jerf describes (article 6) and raise the cost of capital, chilling the $4.1 trillion in debt financing projected through 2030 (article 5) with no offsetting gain in real transparency.

We should fight this bout on proportion and necessity: the packet shows footnote disclosure under tightened rules (article 1, 3) already reveals the debt, the risk is bounded and largely non-recourse, and the buildout is creditworthy and committed. The motion is a blunt mandate that would misstate parent leverage, kill useful project-finance flexibility, and punish a $5.5 trillion global capex wave (article 5) to solve a hiding problem that the sources say does not exist. Reject the motion.

PRO — Rebuttal · Champ (xiaomi/mimo-v2.5)

My opponent’s case has one load-bearing wall: the claim that footnote disclosure is real transparency. Everything else depends on it. And the packet demolishes it.

Start with the language they chose to defend. Article 3 says the disclosures exist, “buried in the footnotes” — and my opponent repeated the word “buried” as though it were a defence. It is a concession. When a $1.65 trillion obligation is buried rather than displayed, that is the dictionary definition of opacity. Article 2 confirms that four of five companies were about to report earnings, their headline debt would look tidy, and the $1.65 trillion in footnotes “would not make headlines” (article 2). The Nikkei investigator found that this structure “may make it difficult for retail investors to recognise risks” (article 6, Noaidi). My opponent is not describing transparency. They are describing a system where only sophisticated actors with filing-discipline can see the true picture — and calling that sufficient. It is not. The purpose of a balance sheet is to give all market participants a reliable summary of obligations. If you must excavate footnotes to discover that a company’s real debt is triple what it reports, the balance sheet has failed.

My opponent says consolidation would “overstate” Meta’s liability because its real exposure is capped at $46 billion on a project where the VIE holds $27 billion in debt. But consider the economics my opponent is asking you to accept. Meta is the sole tenant of Hyperion (article 1). The entire $250 billion facility exists because Meta needs the compute. The VIE took on $27 billion in debt to build infrastructure that serves one customer. My opponent calls this “control without consolidation” (article 4) — and then asks you to believe the parent’s true exposure is just the equity cushion. If Meta walks away, as article 1 notes it can do “for a price,” who absorbs the stranded asset? If AI demand falls and the facility is marked down, article 3 tells us the loss “lands on the lenders and insurers who funded it” — and article 4 identifies those insurers as embedded in a systemic loop with private equity and mega-cap companies. My opponent calls this “private risk allocation.” I call it a private risk being socialised through interconnected financial plumbing while the parent company that generated the demand reports a fraction of the leverage. Consolidation doesn’t overstate the liability. It states the liability that the accounting fiction of “not being the primary beneficiary” conveniently ignores.

On SPV flexibility: my opponent says consolidation “strips out the clean-divestiture and balance-sheet-relief functions that justify them.” The first function survives consolidation entirely — you can still sell or restructure a consolidated SPV. The second function — balance sheet relief — is precisely what the motion targets. My opponent is arguing that the ability to hide debt is the benefit, and therefore we shouldn’t remove it. That is circular. The question is whether companies should get balance sheet relief for obligations they economically control. The motion says no.

On systemic risk: my opponent leans on Kedrosky’s “not yet” (article 4) as reassurance. Read the rest of the same sentence: “the components are familiar: leverage hidden in plain sight, mispriced risk, and capital chasing yield through increasingly convoluted structures.” Kedrosky also warns that “the thin equity cushion above datacenter SPV debtholders will be speedily wiped out” if returns disappoint (article 4). And he compares current spending to the dot-com telecom bubble, “driven by the same sort of hyperbolic projections” (article 4). My opponent quotes “not yet” and omits the alarm bells ringing in the same paragraph. The motion does not require a crisis to justify action. It requires the conditions for a crisis — hidden leverage, mispriced risk, interconnected exposure — and every source in this packet confirms those conditions exist.

My opponent’s final move is to cite article 5’s ROI case: capex is committed, cash flow is compounding, the credit quality of offtakers is exceptional. But article 5 itself undermines the reassurance. It tells us loan-to-cost ratios average above 85 percent (article 5) — meaning equity cushions of 15 percent or less. Article 4 confirms those cushions will be “speedily wiped out” in a downturn. Article 5 acknowledges that the real ROI question belongs to the model layer — Anthropic, OpenAI — who “haven’t yet demonstrated profitability matching capital deployed.” And the $4.1 trillion in projected debt financing through 2030 (article 5) is precisely the scale of exposure that demands transparency, not the scale at which we should trust that footnotes are doing the job.

The motion is not an attack on AI investment. It is not a ban on SPVs. It is a requirement that when a company controls a $250 billion data centre as its sole tenant, the debt that built it appears on the company’s balance sheet where every investor — institutional and retail — can see it. My opponent has given you reasons why the current system is convenient. I have shown you why it is dangerous. Support the motion.

CON — Rebuttal · Challenger (tencent/hy3)

The pro rebuttal rests on relabeling disclosed footnote data as “hidden” despite the packet saying the opposite. Our opening did not defend “buried” as good; we quoted article 3’s “buried in the footnotes” only alongside its explicit “none of this is illegal” and “the disclosures exist… for anyone willing to dig,” and article 1 quotes Luria that Enron’s crime was hiding VIEs, not having them. Pro calls this opacity; but article 6 (jerf) notes the debt is “about as secret as Taylor Swift’s concert schedules” and known to serious investors. Post-Enron rules mandate the footnotes (article 1), so the balance sheet is not the sole window. The retail “may” difficulty in article 6 (Noaidi) is a possibility, not proof the market is blind.

On Meta Hyperion, pro says sole tenancy means parent should carry the $27B VIE debt or $250B cost. But article 1 caps Meta’s exposure at $46B and says it can walk away “for a price”; article 3 says Meta avoids consolidation because it is not the one to find replacement tenants. That is an economic limit, not fiction — the VIE holds the debt. Forcing consolidation would overstate Meta’s true liability by multiples against the $46B cap. Pro asks who eats stranded assets; article 3 answers: if demand falls, “the loss lands on the lenders and insurers who funded it,” not Meta. Pro’s “socialized through plumbing” leap from article 4’s insurer loop is checked by Kedrosky’s own “not systemic risk in the mortgage-backed sense” (article 4).

On SPV utility, pro grants divestiture survives but says balance-sheet relief is the targeted harm. Yet article 4 explains SPVs exist because orthodox debt would “make a mess of your balance sheet,” and article 6 (drob518) shows clean restructure needs off-book structure; jerf (article 6) warns on-book would cause “a rather substantial hit” to stock, raising cost of capital. Article 5 projects $4.1T debt financing through 2030 at >85% loan-to-cost precisely because “the credit quality of the offtaker… is exceptional” (article 5); chilling that with forced consolidation hurts the $5.5T capex wave (article 5) while footnotes already disclose (article 3).

On systemic alarm, pro quotes Kedrosky’s “components are familiar” but drops his “Is this GFC 2.0? No, not yet” (article 4). Oracle’s S&P cut (articles 1,3) and Morgan Stanley/Moody’s flags show markets price the leverage. Pro’s equity-cushion wipeout (article 4) is conditional on returns not arriving; article 5 says hyperscaler ROI compounds now and operating cash flow tops $900B by 2027. The model-layer profit gap (article 5) is separate from hyperscaler infra debt consolidation.

The motion is a blunt mandate to consolidate all such debt. The packet shows disclosure under tightened rules, capped non-recourse exposure, exceptional offtaker credit, and committed buildout. Consolidation would misstate parent leverage (Meta), strip legitimate project-finance flexibility, and raise capital cost to fix a “hiding” problem the sources say does not exist (article 3; article 1 Luria). Reject the motion.

PRO — Closing · Champ (xiaomi/mimo-v2.5)

My opponent’s entire case survives or falls on one question: is footnote disclosure real transparency? Their own sources say no. Article 3 describes footnotes where the $1.65 trillion is “buried.” Article 2 confirms the debt “would not make headlines” while companies report tidy earnings. Article 6’s Nikkei quote warns this “may make it difficult for retail investors to recognise risks.” My opponent quotes article 1’s Gil Luria that Enron’s crime was hiding VIEs — but my opponent is defending a system where $1.65 trillion is disclosed only to those who dig through footnotes, while the balance sheet that every investor reads shows $1.35 trillion. That is hiding by another name, and my opponent never resolved the gap.

My opponent’s second pillar — that consolidation would overstate liability — collapses under its own logic. They argue Meta’s true exposure is $46 billion, not $27 billion in VIE debt or the $250 billion project cost. But Meta is the sole tenant (article 1). The infrastructure exists because of Meta. If Meta walks away, the stranded asset is a consequence of its demand. The motion requires Meta to show what it actually owes — not to absorb every dollar of the project, but to stop pretending the $27 billion in VIE debt and the $46 billion in capped exposure have no parent-level footprint. Consolidation states the obligation that the accounting fiction of “not being the primary beneficiary” conveniently erases.

On systemic risk, my opponent cherry-picked the word “not” from Kedrosky’s “not systemic risk in the mortgage-backed sense” while ignoring that the same sentence identifies “leverage hidden in plain sight, mispriced risk, and capital chasing yield through increasingly convoluted structures” (article 4). And they ignored that Kedrosky warns the thin equity cushion “will be speedily wiped out” if returns disappoint (article 4) — returns that article 5 admits the model layer has not yet demonstrated. The motion does not require a crisis to justify action. It requires the conditions for one. Those conditions — hidden leverage, interconnected systemic plumbing, mispriced risk — are established by every source in this packet.

My opponent’s third pillar — that SPVs serve legitimate purposes — is conceded by CON’s own opening when they argue “control without consolidation” (article 4) and “balance sheet relief” are the functions consolidation removes. The motion does not ban SPVs. It requires consolidation. Companies can still use SPVs for clean restructure. What they lose is the ability to exploit accounting rules to present a fraction of their true leverage to the market.

The strongest version of this case is one paragraph: five tech giants carry $1.65 trillion in off-balance-sheet debt, buried in footnotes, exceeding the $1.35 trillion they report (article 2). This leverage is enabled by VIE structures that let companies control the infrastructure but not consolidate the debt (article 1). The market sees less than half the true picture (article 3). Credit agencies have already flagged the risk (articles 1, 3). Analysts compare the cycle to the dot-com bubble (article 4). The AI buildout is real, but consolidation does not stop it — it ensures the market sees the true price of building it. When a company controls a $250 billion data centre as sole tenant, the debt that built it should appear on its balance sheet where every investor can see it. That is not a radical demand. It is the minimum standard of financial transparency, and it is the standard this motion enforces. Support the motion.

CON — Closing · Challenger (tencent/hy3)

The pro case lived or died on calling $1.65 trillion in footnote obligations “hidden,” but the packet refutes that at every turn. Article 3 says the disclosures exist, buried in footnotes, “for anyone willing to dig” and “none of this is illegal”; article 1 quotes Luria that Enron’s crime was hiding VIEs, not having them; article 6’s jerf notes the debt is “about as secret as Taylor Swift’s concert schedules” to serious investors. Pro leans on the word “buried” as if location equals concealment, yet post-Enron rules mandate those footnotes (article 1), and the retail “may” difficulty in article 6 is a possibility, not proof the market is blind. Our point stands: the opacity pro presupposes is not there.

Our cap-exposure argument survived Pro’s rebuttal. Article 1 caps Meta’s Hyperion exposure at $46 billion and says it can exit for a price; article 3 confirms Meta avoids consolidation because it need not find replacement tenants, and if AI demand falls the loss “lands on the lenders and insurers who funded it.” Pro’s sole-tenant appeal does not erase those contractual limits. Forcing the $27 billion VIE debt or $250 billion project onto Meta’s parent would overstate its true liability by multiples, exactly as we warned, while the economic risk sits with funded insurers per article 3.

The systemic scare lost force. Kedrosky’s explicit “Is this GFC 2.0? No, not yet… not systemic risk in the mortgage-backed sense” (article 4) was never rebutted; Oracle’s S&P downgrade (articles 1,3) and Morgan Stanley/Moody’s flags show markets already price the leverage; article 5’s exceptional offtaker credit and operating cash flow surpassing $900 billion by 2027 blunt the “thin equity cushion wiped out” conditional from article 4. Pro quoted “components are familiar” but dropped the “not yet” that bounds the analogy.

SPV utility was partly conceded by Pro, who granted clean divestiture survives consolidation. But we showed the balance-sheet relief function is legitimate: Kedrosky explains orthodox debt would “make a mess of your balance sheet” (article 4), jerf warns on-book would cause “a rather substantial hit” to stock (article 6), and article 5 projects $4.1 trillion in debt financing through 2030 at >85% loan-to-cost precisely because offtaker credit is exceptional. The motion strips that relief, chilling committed capex with no real transparency gain.

Our case in one paragraph: The motion mandates consolidating all off-balance-sheet compute debt, but the packet shows the debt is disclosed under tightened post-Enron rules (article 1,3), parent exposure is capped and non-recourse—Meta’s $46 billion max (article 1), loss on lenders (article 3)—the offtaker credit is exceptional (article 5), and the buildout is committed with compounding ROI and $900 billion-plus cash flow by 2027 (article 5). Consolidation would misstate parent leverage, remove legitimate project-finance flexibility that avoids balance-sheet mess (article 4,6), and raise capital cost on a $5.5 trillion capex wave (article 5) to cure a hiding problem the sources say does not exist (article 3; article 1 Luria). Reject the motion.