Bloomberg: "Early signs suggest that the credit market is approaching the limits of how much more debt it can absorb. One after another, yields are soaring into the double digits on bonds and loans funding the AI infrastructure buildout. Bankers are finding it more difficult to offload risk to insurers and other investors...With those jitters growing louder each week — and spurring, in turn, angst about the potential economic fallout — Citadel Securities is cautioning clients that even the biggest names in tech could get ensnared as they put their balance sheets behind scores of projects critical to their AI ambitions." Years ago, I was debating the implications of the COVID outbreak with a client at 13D and I was expressed concerns about debt sustainability when he told me: "Never underestimate how much money there is in the world." That's stuck with me since and puts into perspective the extraordinary scale of AI buildout ambitions, which are running into the limits of what capital markets can provide. As I warned in my September report on "The AI Trade" ( "In the first five months of this year, Alphabet, Amazon, Microsoft, Meta, and Oracle borrowed as much as they had in the five years prior. Morgan Stanely expects AI-related debt issuance to top $570 billion this year, more than double any previous record. Just for a bit more context about the scale of the borrowing binge, here’s an FT recap of Goldman analysis: 'The numbers are just bonkers. Goldman Sachs said in a note earlier this month that US dollar investment-grade issuance had already exceeded $1.5 trillion so far this year, which put 2026 on track to beat the previous record-breaking blowout from the pandemic era. The bank had previously penciled in a total of $2.1 trillion for the whole of this year but now it thinks that may be looking too cautious. The average deal size is $1.7 billion, the biggest in the post-crisis era. This includes at least 20 deals at $10 billion or more at announcement—a size it describes as “jumbo”. (I’m old enough to remember when “jumbo” transactions were $1 billion or more, so thank you Goldman for making me feel ancient.) In the whole of 2025, there were just 12 deals on this scale and from 2022 to 2024, there were 15 in total.' Market participants remain complacent about tech giants continuing to borrow at the rate they’re borrowing if borrowing costs keep rising. And if tech giants begin to pull back from the debt binge, it’ll ripple far beyond the bond market, through to everything from equities to consumer spending."

FT: "China is building data centres at rapid speed across its energy-rich hinterland, as Beijing seeks to turn abundant electricity and cheap land into an advantage in the global race to develop AI...While China is struggling to secure enough advanced chips to satisfy soaring AI demand, it has been able to mobilise the land, electricity and construction capacity needed to build the data centres that house them...China already has 24 gigawatts of operational data centre computing capacity, more than the rest of Asia combined but less than half the 56GW in the US, according to SemiAnalysis, the chip-focused research firm. A further 50GW is under construction or has been announced." As I argue in my upcoming report on "Global Electricity Undersupply" (learn about our reports here: China's electricity advantages remain an underestimated edge in the AI arms race, especially as models rapidly grow in size. Chinese models continue to widen the gap with US models in terms of token usage. Just between October 5th and 11th, AI models worldwide processed 172 trillion tokens, with Chinese models accounting for 86.74 trillion of those tokens versus 18.97 trillion for US AI models, according to OpenRouter data. It was the 24th consecutive week in which Chinese models outpaced US models in token usage. This has continued to put downward pressure on token prices, with the LLM Token Expenditure Index falling by more than half since this summer, begging even more questions about the unit economics of US hyperscalers versus their Chinese competitors. The US bet is that bigger and therefore, smarter models are their competitive advantage over China. It’s staggering how rapidly frontier models are growing. As The New York Times reported recently: "In 2026, GPT 6 Astra took nearly an octillion calculations to train—over 50 million times as many calculations as GPT 1 did. Even calling this figure astronomical understates its size: An octillion is a thousand times the number of stars in the known universe…Researchers predict that by 2030, training the largest A.I. models may use electricity at a rate similar to New York City today." The bigger models get, the more electricity they demand and therefore, the bigger China’s advantage becomes given their electricity supply capacity and pricing. According to Wall Street Journal reporting from late last year, Chinese data center power rates range between 3 and 5.6¢ per kWh while US data center power rates range between 7 and 9.1¢ per kWh. Far more to come in the report! FT link:

To my mind, OpenAI and Sam Altman committed the original AI sin, debuting ChatGPT to the public, following the platform-era assumption that a first mover could blitzscale consumer usage, establishing monopoly or duopoly pricing power. This neglected that consumer usage wouldn't prove adequately monetizable and model-capability leads weren't sustainable amidst an arms race. If instead, they'd started by targeting high-value enterprise use cases across sectors, they'd have had a CAPEX path more in line with an ROI path, a better chance of entrenching service loyalty, and less risk overhang, whether existential, political, or logistical. Healthcare exemplifies that squandered opportunity. Recent survey results from Bessemer: "In 2025, 84% of executives expected Gen AI to transform clinical treatment decisions within 3 to 5 years. Just a year later, 71% say it already has. Healthcare AI ROI arrived 2x as fast and at a higher level than expected. Buyers projected ~24 months, and realized it in ~12, averaging 3.5x, exceeding expectations in ~40% of use cases. 54% see material ROI inside the first year." I did a deep-dive on the healthcare AI opportunity in my July report on "The Medical Innovation Inflection" ( "We see compelling reason to doubt AI’s ability to deliver transformational change at the scale and on the timeline the market currently expects across sectors. We do not have the same skepticism about the impact it will have on healthcare. Yes, healthcare has long lagged other sectors in technological transformation, held back by a combination of bureaucracy, profit seeking, personal privacy concerns, and the life-and-death consequences of misguided implementation. Those issues are already dragging on and will continue to drag on AI integration in healthcare. However, we also see AI’s strengths as so obviously well-matched to so many existing healthcare shortcomings, its impact can still be game-changing even if constrained. Study after study has shown AI’s ability to assist physicians in coming to more accurate diagnoses. For context, each year, there are more than 12m “serious diagnostic errors” just in the US, costing the US health system an estimated $100b annually. Meanwhile, AI has already transformed medical devices. The FDA has cleared more than 1,300 medical AI devices since 2015, with 331 cleared in 2025 alone. It is also transforming drug discovery. In total, there are 173 AI-originated drug programs currently in clinical development. And finally, AI is altering how care is delivered. According to survey results released in March, roughly 80% of physicians now use AI professionally. AI is reducing the administrative burden on healthcare workers, allowing more time to care for patients." Chart link:

Alphaville: "[There's a] mismatch between consensus forecasts for chip sales and US data centre completions...Morgan Stanley measured the gap earlier this week by estimating the shortfall in available power, concluding that >1/2 of the GPU servers sold between 2026 and 2028 might not have anywhere to be plugged in. Research from Jefferies reaches a similar conclusion using a very top-down method: space. Tracking US data centre construction sites by satellite imagery shows between 16 and 18 GW of gross capacity that can be energised this year...The rush to complete what’s already started makes it likely that for next year, data centre deployment by GW probably can’t go much above the low twenties. This represents a doubling versus the 11 GW deployed in 2025, but is a huge shortfall when compared with what’s implied by chip sales forecasts." This gets to two points I made in my recent report on "The AI Trade" ( First, hyperscalers appear likely to fail to build out the data center capacity required to improve their models enough to inspire adequate enterprise spending. To quote the report: "Delays have always been part of the data center construction reality—historically, only 72% of data center capacity has come online on time. However, things are clearly getting worse. Goldman expects only about 1/2 of the AI computing capacity scheduled to activate between now and 2028 via data center construction will actually come online by its target date. As asset valuation firm Barkr calculated in an August report, delays at that rate would translate to a compute supply/demand gap of 13.4 to 19.2 GW in 2027 and 27.3 to 36.8 GW in 2028." Second, market participants are underestimating how much chip stockpiling has been happening and what that could mean for chip demand expectations moving forward. To again quote the report: "There’s always been an element of suspension of disbelief in the triple-digit percentage stock gains made by everyone from Nvidia to Broadcom to TSMC, Micron, AMD, and GE Vernova—as if the AI revolution had rendered cyclicality a challenge of the past. Now, AI spending’s rate of change is poised to slow. Politics is likely to increase fear that the pace of the AI data center buildout falls short of expectations. And the threat grows that one or more hyperscalers decides to pull back from the AI arms race. If AI CAPEX slows, it’s likely to hit picks-and-shovels earnings greater than most will anticipate. From hardware shortages has come stockpiling. As one VC noted recently: “In some data centers, I’m hearing the usage of graphics processing units is only around 35% or 40% because some companies are hoarding colossal amounts in case they come to a point at which they don’t have enough chips to provide the computing capacity.” FT link:

AI's concentration risk: "Top 10% of customers account for 99.5% of model-serving spend and 99% of neocloud spend, leaving the bottom 90% of firms with 0.5% and 1%...The bottom line is that adoption is broadening while the spending base is not, and AI infrastructure will keep depending on a small set of heavy spenders until the tail scales up." This is certainly evidence of the technology's immaturity—over time the spending base will expand as more companies figure out how to effectively integrate AI to unlock operational value. However, it's also evidence that adoption challenges are far more persistent than the model builders anticipated. I quoted Sam Altman on this in my recent report on "The AI Trade" ( "The economy just has so much inertia. People just keep doing the same things. They keep buying from the same company. They keep using their tools in the same way. I think that’s actually a positive in many ways. It’s going to make this big transition in front of us go smoother and slower. But I think it means we’ve all been too ambitious on timelines." It's not just about inertia. AI is still plagued by its weaknesses, from hallucination to agentic workflows breaking down midstream. But to the inertia point, AI puts unprecedented transformational demands on enterprises. As I warned in my December report on "GenAI & Productivity" ( "As much attention was paid to the headline 95% failure estimate by MIT researchers, their explanation for that failure rate was likely a more important long-term consideration in understanding when and how companies will realize productivity gains from genAI. To quote the researchers: “The dominant barrier to crossing the GenAI Divide is not integration or budget, it is organizational design.” McKinsey is delivering a similar message: “Building a business for the agentic age will require a fundamental rewiring of how the business operates, innovates, and protects sources of value creation.” Deloitte is saying much the same: “This is not about adding another tool; it’s about fundamentally rethinking how work gets done from the top down.” It's difficult to look at modern history and identify an enabling technology that demanded the depth and speed of organizational transformation being suggested for genAI today." For all of AI's capabilities, there is no path to ~$2.5t in annual AI revenue (the estimatdd amount required to offset CAPEX) unless the vast majority of enterprises become relative "heavy spenders" on a manageable timeline. Instead, the tail is elongating slowly while evidence mounts that today's "heavy spenders" are pulling back their spending. According to Ramp data, the top 1% of spenders, the cohort that drives ~80% of OpenAI and Anthropic’s enterprise revenue, cut per-employee spend by nearly 10% in August. Chart link:

Stunning stat: "Anthropic’s investors are expecting the company to reach a valuation of $2tn when it goes public in the coming weeks. Add in SpaceX, which began trading at $2tn after its IPO in June, and OpenAI, which is considering raising money privately at $1.2tn ahead of a public listing next year, and these companies alone could be worth well north of $5tn. Now compare that with the entire history of IPOs from 1980 to 2025. The 3,365 tech companies that went public in that period were worth a combined $4.1tn when they started trading." While I've written much about AI's transformative potential across sectors, I've always been dubious about how much value hyperscalers can seize enabling that transformation. As I dissect in my recent report on "The AI Trade" ( it's not about user acquisition. OpenAI claims to have over one billion active users across its services and two million businesses using its AI models. Anthropic has claimed to have more than 300,000 business customers. The question is not whether they can bring users to their services, but rather how much average revenue they can generate per customer relative to the price of building and maintaining their models. The cost of compute is inflating at the same time competition is depressing token pricing power. That's a precarious dynamic when so much hinges on the success of two companies. To again quote the report: "It’s hard to overstate how much hinges on these IPOs. As mentioned in the Executive Summary, OpenAI and Anthropic will account for 13% of AWS revenue and 27% of Google Cloud revenue this year. As for Microsoft, OpenAI alone accounts for roughly 70% of its AI-specific revenue ($24.1 billion out of an estimated $34 billion total for the fiscal year ending in June 2026). OpenAI has committed to tens of billions of spending on chips from the likes of Nvidia and Broadcom. Deepening circularity concerns, tech giant earnings growth has been increasingly driven by paper gains in the private-market valuations of Anthropic and OpenAI. In Q2, Amazon, Alphabet, Nvidia, Meta, and Microsoft reported $160 billion in cumulative “other income”, trouncing the $69 billion in “other income” reported in Q1. To quote the FT: “These one-off valuation boosts, derived in large part from enthusiasm around AI, risk distorting the financial picture at a time when investors are closely scrutinizing tech earnings.” If either Anthropic or OpenAI stumble in their IPOs, it’ll hit tech giants on multiple balance-sheet fronts and likely ripple through the US and global economy. A pin-prick popping of the AI bubble may not be our base case, but if a pin is out there, it’s likely the revenue versus spending trajectories of Anthropic and OpenAI." FT link:

FT: "Leading US AI labs such as OpenAI and Anthropic are releasing cheaper models as they fight to retain cost-conscious customers who are switching to cut-price alternatives from Chinese rivals. The price war comes as rising AI bills push companies to curb usage and seek cheaper models, helping Chinese developers including Moonshot and DeepSeek make inroads with users from Silicon Valley to Europe. OpenAI recently said that it was slashing prices for GPT-5.6 Luna, its “fastest and most affordable model”, by 80 per cent. Anthropic has launched Claude Opus 5, touting the system’s “frontier intelligence . . . at half the price” of Fable 5, the company’s most capable model. The moves have helped decrease prices that customers are paying for models from leading US labs by almost a quarter since mid-July." In my December report on "GenAI & Productivity" ( I warned about the pricing power challenges faced by US hyperscalers: "While there’s a lot of speculative fear about how a single LLM could rise to dominance and what that could mean for economic, societal, and political stability, we believe the bigger concern for investors today is how relative model parity could compromise pricing power. Tech giants have thrived on monopolies and duopolies for a decade or more. Now, they’re in an LLM arms race where it’s unclear when or even if ever leadership will be sustainable." Since, my concern about the commoditization of AI has only intensified as Chinese models have risen to power. According to OpenRouter data, Chinese models accounted 4.4% of token usage by US companies in January. Today, that share is over 60%. Meanwhile, enterprise model router adoption has skyrocketed and frontier labs have been increasingly shifting from subscriptions to usage-based, metered billing, business models more akin to utilities than the per-seat models SaaS companies thrived on over the past decade. As RBC warned in July: “Oil, natural gas, and electricity are all important. Entire economies depend on them. But importance alone does not guarantee strong economics or durable profitability. When supply expands aggressively and is increasingly interchangeable, competitive forces tend to compress returns over time.” Far more to come in my next report! Learn about Sage Road Research here: Interested in subscribing? Message me. FT link:

FT: "Prices for credit default swaps, popular tools to bet against corporate debt, tied to Oracle, SpaceX, Alphabet, Amazon, Meta, Broadcom and Nvidia have risen to record highs in recent days...'Credit markets don’t deal well with uncertainty, and the sheer unpredictability of the pace and cost of AI financing is triggering a serious crisis of confidence right now.'"