Rates, Credit, Robotics, BloombergGPT

Rates vs Inflation

Yesterday the March CPI print showed a 5% increase YoY, down from 6% in February. The Fed’s interest rate is also 5%, marking this moment as the first time that interest rates have at least matched inflation. It seems likely the Fed will raise rates to 5.25% in three weeks unless something new gives them a reason to stop here.

At the same time, the Fed managed to shrink its balance sheet by $100B since it bailed out SVB last month. That bailout wiped out six months ($400B) of balance sheet contraction. At this turning point for the Fed, there is an exceptionally long list of positive and negative data to support all types of views.

On the one hand, earnings are expected to contract; on the other, companies are actively reducing headcount and cutting costs. Meta laid off 30% (26,000) of its workforce, and Twitter reduced its headcount from 6,000 to 1,500 employees. The dollar is down almost 15% from the peak in September, significantly reducing the pressure from international earnings.


Private Credit – Direct Lending

Direct lending has emerged as a significant investment asset class, with the market expanding from $200 billion in 2013 to over $800 billion in recent years. This investment strategy involves the creation of privately negotiated loans within a fund structure, where investors supply capital for the loans, while managers allocate capital and manage risk. Direct loans typically charge higher rates, between 10% and 12%, and can include bespoke terms that banks can’t or won’t offer. The growth of this asset gives us an insight into the shifting landscape of the financial sector, especially as banks tighten their own lending standards.

The appeal of direct lending as an asset can be partially attributed to the shortcomings of publicly traded fixed-income from investment grade to high-yield bonds. Excessive demand for liquid, publicly traded fixed-income securities pushes rates lower to a range that can be hard to justify. The illiquidity involved in direct lending can be considered a worthwhile trade-off for a longer-term bond allocation. The risks associated with direct lending should not be overlooked, and investors must remain vigilant in assessing borrowers, managers and market conditions.



We’ve written about warehouse automation in the past. It increases time efficiency, accuracy and safety while reducing costs, human error and response time. It’s a necessary evolution as the demand for supply chain efficiency exceeds human performance. We’re getting closer to fully automated “lights-out” warehouses thanks to a handful of companies from Honeywell to Amazon’s Kiva and a local MA company Symbotic.

Symbotic was founded by the C&S Wholesale Grocery billionaire Rick Cohen fifteen years ago to address his company’s logistics challenges. C&S is the 8th largest privately held company in the US. Their fully automated system uses computer vision algorithms to optimize pallet density and retrieves items at 25mph with near-perfect accuracy. Compared to Amazon’s Kiva system, this is an openly available ecosystem that can benefit more than just one company.

While bank failures and the Fed take the headlines, innovation tailwinds continue to march forward. Symbotic is a newly public company and this is not an endorsement or a recommendation.



My former employer Bloomberg recently announced BloombergGPT, an AI model trained on their extensive collection of financial documents. Everyone would like an AI investment analysis solution, but we’re not quite there yet.

Investment decisions require basic factual accuracy and precision, and so far, most LLMs routinely fail when it comes to basic algebra. The OpenAI plug-in with Wolfram Alpha helps address the mathematical limitations of LLMs, but that’s more of a proof of concept.

In all likelihood, BloombergGPT will likely require a Bloomberg subscription, which costs around $25k/yr. Given that all of the SEC filings are public, competitors can train models on more or less the same data, if they have the budget to do so. My guess is the BloombergGPT model will ultimately be more useful for drafting annual 10Ks and other regulatory filings rather than making investment decisions.

Weekly Articles by Osbon Capital Management:

"*" indicates required fields