AI-related financing could grow to about $10 trillion over the next five years and eventually reach excessive levels, Pimco chief executive Emmanuel Roman said, Bloomberg reported on 3 November. Roman put the yearly need at about $2 trillion, against a US bond market he sized at about $12 trillion. That yearly figure is about 1.3 to 2 times the pace implied by Morgan Stanley chief executive Ted Pick’s estimate of $10 trillion to $15 trillion over a decade, by our arithmetic.
Update, 3 Nov: the other side. Goldman’s David Solomon told Bloomberg Television that deal demand remains robust, State Street’s Ron O’Hanley said it is “hard to see” AI as a bubble yet, and the White House’s Kevin Hassett called AI “a really well-financed ecosystem”, Bloomberg reported. Our piece on what the bulls said, and what their firms manage.
Roman is one of four senior figures in finance who warned about AI money within two days. The ECB’s vice-president named AI stock valuations as a main financial-stability risk, a senior Chinese regulator called AI investment “overly concentrated”, and the credit chief of Canada Pension Plan Investment Board told investors to watch their exposure. Below, their dollar estimates sit on one axis, without adding them up.
Business Pill · Concentration Explained
An evergreen Business Engineer explainer on concentration: when a few big pieces make up most of a total. It teaches the general idea only and does not discuss Pimco, the ECB or the people quoted here.
The key insight: Pimco chief executive Emmanuel Roman put AI-related financing at about $10 trillion over five years, about $2 trillion a year, Bloomberg reported. That is about 1.3 to 2 times the yearly pace implied by Morgan Stanley chief executive Ted Pick’s $10 trillion to $15 trillion over a decade, by our arithmetic, while the ECB, China’s financial regulator and CPPIB’s credit chief flagged valuations, concentration and exposure.
According to Bloomberg, Roman estimated that about $2 trillion of financing needs could be created each year over the next five years to fund the AI industry’s capital spending. Bloomberg’s headline says he warned AI financing is on a path to become “too much”. The roughly $2 trillion a year is about 17% of the $12 trillion bond market Roman compared it with, by our arithmetic.
Pimco is also one of the lenders. In our 1 November piece on Meta’s Hyperion bond, IFR reported that people familiar with the deal said Pimco was the main anchor investor. In late October, Bloomberg reported that Pimco had cut AI debt positions, as noted in an update to our 28 October piece on record bond sales.
Three large AI money estimates on one per-year axis, in trillions of dollars: Pimco chief executive Emmanuel Roman’s about $2 trillion a year of AI-related financing needs over the next five years (about $10 trillion in total, Bloomberg), SemiAnalysis’s about $11.1 trillion of cumulative AI capex from 2024 to 2029 (about $1.85 trillion a year on average) and Morgan Stanley chief executive Ted Pick’s $10 trillion to $15 trillion of capital over the next decade ($1 trillion to $1.5 trillion a year, The Standard). Roman’s per-year figure is his own; the other per-year figures are our arithmetic. The rows measure different things (financing, capex, capital) and are not added.
The ECB, China’s Regulator and Canada’s Biggest Pension
ECB vice-president Boris Vujฤiฤ said on 2 November that the central bank has identified high valuations for AI stocks, bond-market stress and potential losses at shadow banks as the main risks to financial stability, Reuters reported. “As the FSR has been prepared, we saw the correction in the markets in AI (stocks),” he told an event in Ireland, adding: “Obviously, everybody’s exposed to that…including Europe”.
The Financial Stability Review itself is due on 18 November, Reuters reported. The prepared speech the ECB posted for the Dublin conference does not discuss AI stocks, so these remarks rest on Reuters’ account.
On 3 November, Xiao Yuanqi spoke at the Hong Kong Monetary Authority’s Global Financial Leaders’ Summit. He is vice head of the China National Financial Regulatory Administration, and he said that “a massive influx of capital into the technology sector and leading firms has driven up leverage levels within these areas—potentially fueling excessively rapid expansion,” Reuters reported. He said the concentration raises borrowing costs for traditional industries while “simultaneously heightening credit risks”.
David Colla, global head of credit investments at Canada Pension Plan Investment Board, sees signs of “bubble-like” exuberance in the rush to finance AI, Bloomberg reported on 2 November. “You cannot wake up one day and just say, ‘Well, we just had this theme that AI was great, and we woke up and it was 25% of our portfolio,’” he said. Benzinga, citing the interview, reported that AI deals make up roughly 3.5% of the $618 billion fund’s credit portfolio, and that CPPIB is watching off-balance-sheet structures, where firms set up separate companies to carry data-centre debt.
Three Estimates, Three Clocks
Three of the numbers in circulation land near $10 trillion, but over different periods and for different things. Roman’s is financing over five years. Pick’s, given at the 5th Global Financial Leaders’ Investment Summit according to The Standard, is capital for AI development over the next decade. SemiAnalysis’s 6 July report says cumulative AI capex from 2024 to 2029 “will reach ~$11.1T”.
Put on a per-year basis, Roman’s is about $2 trillion, his own figure. By our arithmetic, Pick’s range works out to $1 trillion to $1.5 trillion a year and SemiAnalysis’s to about $1.85 trillion a year on average over six years. The chart shows them on one axis. They are not added together, because financing, capex and capital are different measures.
SemiAnalysis also forecasts “over $7T of debt outstanding by 2029”, the figure we set beside Apple’s and Oracle’s funding choices in our 2 November piece Self-Funded vs Market-Funded AI. That is a stock of debt at one date, not a yearly flow, so it stays off the per-year axis. It is about 64% of the firm’s capex total, by our arithmetic.
What is new here is that, within two days, a bond investor, a central bank, a Chinese regulator and a pension fund described the same build-out as a risk, and that three of the dollar estimates can be read on one per-year scale.
The Structural Read
As we read the four warnings, each comes from a different seat: a bond investor that lends to AI projects, a central bank watching European funds, a regulator watching credit costs for older industries, and a pension fund sizing its own exposure.
As we read the numbers, the headline totals look alike, near $10 trillion, but the clocks differ. Over five years, Roman’s financing estimate implies a faster yearly pace than Pick’s decade-long capital estimate, and the per-year view is what separates them.
David Colla, CPPIB, to Bloomberg, 2 Nov 2026
“You cannot wake up one day and just say, ‘Well, we just had this theme that AI was great, and we woke up and it was 25% of our portfolio,’”
Three Implications
FOR READING THE ESTIMATES Financing, capex and capital are different measures, so the three per-year figures sit side by side and are not added: about $2 trillion (Roman), about $1.85 trillion (SemiAnalysis, by our arithmetic) and $1 trillion to $1.5 trillion (Pick, by our arithmetic).
FOR SINGLE DEALS Oracle’s $10 billion bond sale on 2 November equals 0.5% of one year of Roman’s $2 trillion, by our arithmetic.
FOR EXPOSURE Benzinga reported that AI deals are roughly 3.5% of CPPIB’s credit portfolio, about one-seventh of the 25% Colla used as his cautionary example, by our arithmetic.
The framework reads: “It is a sovereign-scale financing problem dressed in tech-company clothing”.
As we read it, the week’s warnings come from the institutions that size and supervise that financing: a bond investor, a central bank, a banking regulator and a pension fund. The framework was published in May 2026 and does not discuss these remarks.
We read the opening paragraphs of Bloomberg’s Pimco and CPPIB articles, not the full texts, and the CPPIB portfolio figures come from Benzinga’s account. Bloomberg describes the $12 trillion as the US bond market but, in the paragraphs we read, does not say which part of that market it covers. The ECB’s review is not yet published, so its wording on AI is not known. The estimates are forecasts by their authors, and their definitions differ. We did not contact the people quoted.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Pimco’s chief executive put AI financing at about $10 trillion over five years, about $2 trillion a year, while the ECB, China’s financial regulator and CPPIB’s credit chief flagged valuations, concentration and exposure. On a per-year basis, Roman’s figure is the highest of the three large estimates, by our arithmetic.
A note on sourcing. Roman’s and Colla’s remarks are from the opening paragraphs of Bloomberg’s articles of 3 and 2 November; CPPIB portfolio figures are as reported by Benzinga. Vujฤiฤ’s and Xiao’s remarks are as reported by Reuters; Pick’s estimate is as reported by The Standard. SemiAnalysis figures are from its 6 July report. Per-year figures, ratios and shares are our arithmetic; the estimates use different definitions and are not added.
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 (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
