Signal over Noise with Ulrike Hoffmann-Burchardi

Hello and welcome to Signal over Noise. I'm Ulrike Hoffmann-Borchati, CIO for the Americas and Head of Global Equities for UBS Wealth Management. The biggest market signal last week was the market's mounting concern about how AI CapEx will be financed and ultimately paid back.

The NASDAQ lost 3.1% last week, its worst week since April this year. The losses accelerated after OpenAI's CFO talked about government guarantees for debt to fund AI CapEx. Despite the retraction, the market quickly interpreted these remarks to mean that the government needs to be the lender of last resort to make these investments viable.

Credit default swap spreads for companies with AI data center debt jumped. Corbeef's CDS spread now stands at 5.1%, implying an 8 to 10% annual probability of default, assuming a 40% recovery rate. Amid this unease, it's worth noting that even with Meta raising 30 billion and Oracle 18 billion in debt, debt-funded AI CapEx still sits at less than 15% of overall U.S.

AI data center CapEx. That compares to about 50 to 70% of debt-financed telco CapEx in the early 2000s. From my years on the buy side, I have seen the fallout from cycles driven by excessive leverage, from subprime mortgages to the property boom in China.

Debt can amplify booms by multiplying the CapEx impulse. On the other side, it can lead to an investment bust with defaults that force painful balance sheet repair. It's therefore worth watching to what extent AI is moving from the cash flow-funded phase to the debt-financed one.

With CapEx projection above 500 billion for 2026, the risks of a potential air pocket between the demand and supply for compute are increasing. History suggests that a crunch for capacity has frequently turned into a glut. The reason is that supply cannot scale in perfect lockstep with demand.

Infrastructure build-out takes time for data centers two to three years, and demand projections can be off. Examples that are still top of my mind are the telco overbuilt in the early 2000s, the oil and gas shell boom in 2010 to 14, and the solar capacity build-out in the 2020s. Demand signals are therefore key to watch from here.

Aggregate token usage is the demand metric that matters. Tokens are the fundamental unit of compute. Every time a model processes a word, instruction, or line of code, it consumes tokens.

This year, Google has released three data points on token usage across its products. In May, it was 480 trillion tokens per month. In July, 980 trillion tokens per month.

And in October this year, 1,300 trillion tokens per month. The bottom line, growth is still strong, even if the token consumption growth decelerated from 104% to 33% in the latter part of this year. From here on, this will be the key metric to watch to judge the sustainability of AI CapEx investments.

Possibly related to AI are also the weak data points on the labor market last week. A number of private data points continue to paint the picture of a low-hire, low-fire labor market. ADP reported that private employers only added 42,000 jobs in October.

And Schellinger announced on Thursday that October was the worst October for job cut announcements since 2003. These softening labor market trends suggest to us that another federal reserve cut is in store for December, and likely again for Q1 next year. Turning to this week, the most heavily debt-financed public AI data center company, Corweave, reports earnings this Monday.

With 20 billion of CapEx, Corweave represents about 5% of US AI CapEx. While small relative to the overall footprint, the market attention will clearly center on commentary about AI demand. The more important earnings sprint will be NVIDIA Wednesday of next week.

Over 60% of the AI bill of materials are GPUs. On the clearly positive side, there are signals that an end to the government shutdown is in sight. Both economic and public opinion pressure is mounting.

The shutdown is affecting more and more Americans with cuts to food aid payments and flight disruptions. We expect strong bipartisan support for a relief before Thanksgiving. In conclusion, we remain optimistic about an end to the government shutdown and see more upside in AI infrastructure spend as AI usage remains strong.

At the same time, we remain vigilant on any changes in AI demand trends. With this, stay well and stay ahead. UBS Chief Investment Office's investment views are prepared and published by the Global Wealth Management Business of UBS AG or its affiliate, UBS.

This material has no regard to the specific investment objectives, financial situation or particular needs of any specific recipient and is published for informational purposes only. As a firm providing wealth management services to clients globally, UBS AG and its subsidiaries offer both investment advisory services and brokerage services. Investment advisory services and brokerage services are separate and distinct, differ in material ways and are governed by different laws and separate arrangements.

In the USA, UBS Financial Services, Inc. is a subsidiary of UBS AG and a member of FINRA SIPC. For information, please visit our website at ubs.com forward slash working with us. For a full legal disclaimer applicable to the independent investment views produced by UBS, please visit our website at ubs.com forward slash CIO dash disclaimer.