Reuters reported on 13 October 2026 that a selloff in AI supplier shares was “the strongest sign to date that the mood on the trade of the decade is shifting,” with worries rising interest rates could scupper the spending spree. SK Hynix dropped 10% and Samsung fell 8% in Seoul, according to Reuters’ Asia markets report of the same day.
On the same day, the International Monetary Fund’s Global Financial Stability Report put numbers on the money behind the trade: it says fourteen publicly disclosed transactions since 2023 account for approximately $300 billion of equity investments led by hyperscalers in financially weaker AI firms, supplemented by $650 billion in debt, in exchange for $1.4 trillion in future revenue commitments.
Business Pill · FINANCIAL CONDITIONS
A one-minute explainer of financial conditions: how the cost and availability of money change what gets built. It teaches the general idea only and says nothing about any company in this story.
The key insight: As we read it, the IMF’s numbers show where the stress sits: about $320 billion of the $650 billion of debt it counts in AI circular-financing deals is taken on by the investees, the financially weaker firms, not by the hyperscalers that lead the equity.
The figures moved during the day and differ by outlet and time. Reuters’ Asia report said MSCI’s broadest index of Asia-Pacific shares outside Japan fell more than 2% to its lowest level since late July, with Kioxia Holdings down 14%, SoftBank 8% lower and Taiwan’s TSMC down 1.8%. It said Nasdaq futures fell 0.77%.
The Associated Press reported the Kospi down 7.4% to 5,914.35 during Tuesday’s session, its lowest level since late July, with Samsung Electronics down 7.7% and SK Hynix down 8.8%. CNBC reported that SK Hynix plunged more than 10% and that SoftBank Group sank 8.5%, after the Nasdaq Composite dropped 0.44% on Monday.
Share and index moves on 13 October 2026 as each outlet reported them during the session (Reuters’ Asia markets report and the AP). Figures moved through the day and differ by outlet and time.
What Was Posted on X
โTo operate a gigawatt cluster for a year, if you're talking about building the CapEx it's probably closer to, you pay for it upfront, so it's probably $50 or $60 billion. But then that's just one cluster for one part of training.โ
Stripe posted these quotes and tagged @finkd and @alexandr_wang; the post does not say who speaks each line. On the upfront CapEx for a gigawatt cluster, the post quotes: “you pay for it upfront, so it’s probably $50 or $60 billion. But then that’s just one cluster for one part of training.” Another line: “we were definitely building out this compute for inference ahead of knowing that the product was going to be ready.” These are the speakers’ own estimates and views, not disclosed figures; we include them as context on what a single cluster costs.
What Zuckerberg Said on Cheeky Pint
On Stripe’s Cheeky Pint podcast, in an episode uploaded on 12 October, Meta chief executive Mark Zuckerberg recalled the end of last year: “We haven’t seen the models yet. You know, we’re going to write this check for like, I don’t know, like a hundred billion dollars or $200 billion.” He added that he asked Alexandr Wang, “Alex, you sure this is going to work?”
That is Zuckerberg’s own recollection on the podcast, not a disclosed budget figure. The quotes are our transcription of the episode’s captions. The per-gigawatt cost lines in Stripe’s post above come from the same episode; the captions do not show which guest says them, so we leave them unattributed.
Cheeky Pint (Stripe), “Muse Wonder: Zuckerberg and Wang’s $100B AI,” from 53:12.
The Trigger the Outlets Named
CNBC and Reuters both pointed to China’s ChangXin Memory Technologies (CXMT). CNBC reported that over the weekend CXMT announced significant progress in developing a new chip architecture designed to increase memory density. “The CXMT news appears to have been a clear trigger for the selloff in memory stocks, but I think the bigger story is that investors are becoming much less forgiving of the AI trade,” Saxo’s Charu Chanana told Reuters.
Not everyone read the CXMT news as a fundamental shift. Phillip Wool of Rayliant Global Advisors told CNBC: “I read that more as a sign of how vulnerable AI sentiment is at this moment.”
The IMF’s count of fourteen publicly disclosed AI circular-financing transactions since 2023: approximately $300 billion of equity, $650 billion of debt and $1.4 trillion of future revenue commitments (Global Financial Stability Report, October 2026).
What Investors Are Saying
Investor Brad Gerstner of Altimeter on the BG2 Pod (with Gavin Baker, Andrew Fox and Clark Tang), from about 64:18 in the episode: โWe’re spending 1.5 trillion of CapEx on 300 billion of inference revenue.โ The chart on screen as he says it is Altimeter’s estimate of inference revenue, labelled Dec-27 Est. $300B. That is his framing and Altimeter’s estimate, not an IMF or company figure.
What the IMF Says About the Money
The IMF’s report says hyperscalers, which have accounted for the bulk of capex within the AI value chain, are issuing debt at a substantial pace, lifting their share to almost 10 percent in total investment-grade corporate issuance so far in 2026, compared with just 0.5 percent in 2024.
On the circular deals, the report says there is a significant concentration of financing and revenue promises in a handful of investees, and a note to its Figure 1.2 says about $320 billion of the $650 billion debt is taken on by investees, including debt backed by investor guarantees, contingent financing, and other off-balance-sheet debt. “The consequences of an AI correction could be global,” the IMF warned, according to Reuters.
The Financing Signals Reuters Listed
Reuters’ analysis said cracks in AI bets are widening, with a data centre IPO collapsing last week in Australia and credit investors becoming picky on AI debt. It said Firmus, a data centre operator backed by Nvidia, shelved a $5 billion IPO in Australia, and that Tencent’s dollar and dim sum bond sale had guidance for 10-year dollar notes about 1.15 percentage points on top of Treasury yields.
It added that SoftBank last month sold the biggest high-yield debt pile ever and promised to pay as much as 9.75% on $4.5 billion, that BlackRock’s research arm downgraded its US stocks outlook to neutral on Monday, and that Broadcom’s CEO said capital was the biggest constraint on growth.
The Structural Reading
As we read it, the day’s sources describe a repricing of the money behind AI more than of the demand for it: the outlets cite bond yields, a pulled IPO, picky credit investors and bond pricing, while Lucerne Asset Management’s Marc Velan told Reuters the “brutal session” in Korea said more about positioning and the cost of capital than about any fundamental deterioration in AI demand.
As we read it, the day’s financing signals point the same way: Reuters listed a pulled data-centre IPO, credit investors becoming picky on AI debt, Tencent’s bond guidance and SoftBank paying as much as 9.75% on $4.5 billion.
The IMF puts the debt in context: it says hyperscalers’ share of total investment-grade corporate issuance rose to almost 10 percent so far in 2026, compared with just 0.5 percent in 2024.
As we read it, the people the outlets quoted mostly framed the day as a question of returns: Saxo’s Charu Chanana told Reuters investors are becoming much less forgiving of the AI trade, and Lucerne’s Marc Velan pointed to monetisation and returns on invested capital.
Charu Chanana, chief investment strategist at Saxo, to Reuters, 13 October 2026
“They are questioning not just how much AI demand can grow, but how long today’s exceptional profits and margins can last as competition increases.”
Three Implications
AI CHIPMAKERS Reuters, CNBC and the AP reported sharp falls in SK Hynix, Samsung and Kioxia, and CNBC and Reuters pointed to CXMT’s reported memory progress.
AI INFRASTRUCTURE BORROWERS Reuters reported that credit investors are becoming picky on AI debt and that a data-centre IPO collapsed in Australia last week.
HYPERSCALERS AND THEIR COUNTERPARTIES The IMF says fourteen disclosed circular-financing deals carry $650 billion of debt against $1.4 trillion of future revenue commitments, concentrated in a handful of investees.
The framework puts it this way: “This piece is about the financing. How the money flows, where it cracks, what it buys, who can absorb it, and what happens when revenue can’t catch up.”
As we read it, Tuesday was a day on which markets asked those questions out loud: who holds the $650 billion of debt the IMF counts, and whether the revenue behind $1.4 trillion of commitments arrives.
What Is Not Established
We read the Reuters, CNBC and AP reports and the IMF report in full; the Wall Street Journal’s and Bloomberg’s market reports on the selloff were paywalled and not read. Share moves are intraday figures from different times and outlets, and we did not read the Korea Exchange’s official closing data. None of the texts establishes that any single cause drove the selloff, and the IMF’s debt figures describe disclosed deals since 2023, not Tuesday’s trading.
Business Engineer Framework
The AI Capex Map
A Business Engineer framework on the trillion-dollar AI capex cycle and who underwrites it.
Chip and AI shares fell sharply across Asia on 13 October 2026, with SK Hynix down 8.8% to more than 10% depending on the outlet and time. The same day, the IMF reported $650 billion of debt in fourteen disclosed AI circular-financing deals and hyperscalers at almost 10 percent of investment-grade issuance, up from 0.5 percent in 2024.
A note on sourcing. This piece is based on two Reuters reports of 13 October 2026 (Tom Westbrook’s analysis and Ankur Banerjee’s Asia markets report), read in full through a full-text syndication, CNBC’s report of the same day and The Associated Press’s Asian markets report, both read in full, and the IMF’s October 2026 Global Financial Stability Report, read in the full PDF from the IMF’s eLibrary on 13 October 2026.
The Wall Street Journal’s and Bloomberg’s market reports were paywalled and are not used; the Korea Exchange’s closing data was not read. Share moves are intraday figures as each outlet reported them. Nothing here is a forecast, and nothing here is financial or investment advice.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur, HackerNoon and Search Engine People. Find him on LinkedIn and Substack. See how we source.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
