Ken Griffin on US Debt, Closed End VC, AI Update

Griffin on US Debt

Last week, Ken Griffin, arguably the most powerful finance figure in the world, released his annual letter. The only market commentary he offered in the brief letter was his growing concern about US Debt and excessive spending. We’ve written about this: the cost of funding US interest rates has gone parabolic and is now at a $1T annual run rate. For context, the US government currently receives about $4.8T in tax revenue and spends roughly $6.5T, which translates to just under a $2T annual deficit. Running a deficit this high is both stimulative and unsustainable. Deficit spending does not directly increase GDP and the US stock market, but it certainly doesn’t hurt!

The awkward part of the budget conversation is that two-thirds of our current annual budget is driven by “mandatory” expenses: Medicare, Medicaid, and Social Security. I don’t like the official use of this term because the same organization that makes these payments also sets and enforces the rules. Discretionary spending (like defense) and non-discretionary (rules-based) is the more appropriate label. Ken Griffin has said he wants better budget management but doesn’t want to see rate cuts due to inflation issues. I haven’t seen any comments on how he wants to achieve better budget behavior.

The status quo is a powerful force. Rate cut expectations continue to temper. Today’s consensus is that we might see two cuts by the end of the year, down from six. Shelter remains the most stubborn inflation metric. The main commodity index, the GSCI, is up 12% this year and 6% in the last 30 days. Rising commodity prices directly impact inflation.

Ken Griffin did not share his views on what the government should do to fix the deficit, so I asked ChatGPT to pretend to be Ken and to generate some ideas. KenGPT’s top two recommendations are to focus on auditing and waste reduction, and entitlement reform. I think those are fair assumptions for what Ken would suggest, although when pressed, it did not provide confident solutions to implement those changes successfully. Unfortunately, These ideas are both obvious and extremely difficult to implement due to factors like short-term re-election incentives or the overall complexity of the operation.

Closed-End Fund Venture Capital

Closed-end mutual funds combine a fixed number of illiquid assets into a single vehicle so that people can trade the basket on public markets. Closed-end funds trade at premiums and discounts to their fair value (NAV, net asset value). Most of them hold boring assets like bonds. Blackstone created a multi-billion dollar closed-end innovation fund called BIGZ in 2021 that holds public assets and roughly 25% of high-profile private market venture capital assets like Axon Space (the private space station). BIGZ trades at a -14% discount (not advice) and has an activist hedge fund, SABA Capital, working on closing the fund to distribute the shares at par value.

DXYZ is a new closed-end fund with just $50 million in “equity” of hot start-ups like Stripe and SpaceX. I put “equity” in quotes because it’s unclear whether the fund has secured its shares. Within ten days of listing, the $50m DXYZ portfolio of startup shares jumped to $1.1B value, a 20x premium. It’s fallen since then to a 12x premium.

Price is what you pay, and value is what you get. This is one of my favorite phrases in investing. There is an insane amount of pent-up demand for high-profile startups, so much so that a loosely organized $50m basket of startup equity gets bid up to insane valuations.

This story indicates that the IPO market is probably ready for more listings. Aside from hype, the market is currently rewarding cash flows, both positive and breakeven. Deeply unprofitable IPOs are unlikely to do well. If deeply unprofitable companies try to IPO, they will likely try what Reddit did and only sell a small amount of their shares publicly so that the demand outpaces the supply, just like DXYZ. In the meantime, private equity has recently taken many public software companies private at an average valuation of 6.8x revenue, indicating that they see attractive value at that price, on average.

AI Update

IO.NET is a blockchain-related AI protocol. The purpose is to allow participants to connect their GPUs (Nvidia or others) to the blockchain so anyone can rent it for model training or inference. Inference is what happens when ChatGPT calculates an answer. The live quotes for participation are here. I have a new Mac Air with an M3 chip that would allow me to earn roughly $0.13/hour. Of course, the highest-paying chip is NVidia’s H100 at $4/hr, although that rate appears to be above market, and no one is currently using them.

The only GPUs currently actively used on the network are the NVidia A100s at $.89/hr. The A100 is the backbone of the ChatGPT era, so its popularity is not surprising. There are currently 520,387 available GPUs for rent on the network, but only 320 or so A100s are in use (at the time of this writing).

While top-level AI requires the absolute best chips to provide dazzling ChatGPT-like results, not all machine learning or AI will need the most expensive units going forward. That said, I don’t foresee anyone renting my laptop’s compute for $.13/hr, or really any of the roughly 10,000 people with Apple laptops who also signed up. Either way, early users typically get rewarded with airdrops to incentivize participation. IO.Net plans to release their token $IO on April 28th. (Not advice)