Bullish on AI despite disappointing August performance
We lost approximately 2.2% in August. This was slightly more than the SP’s decline of 1.77% and was largely due to our aggressive positioning going into September. Thankfully, we did not suffer the full impact of the S&Ps decline of nearly 5% in mid-August while our accounts were generally flat. At the end of the quarter we were penalized by two of our small healthcare investments.
At the end of Q1 when most Investor earnings calls provided evidence that the economy was likely to slow. Nvidia’s amazing earnings beat, and guidance raise, brought new life to Information Technology. Today we believe the IT Economy will grow as a result of the AI paradigm shift.
We see the primary beneficiaries as Semiconductor companies that manufacture chips that move massive amounts of data and the processor companies like Nvidia that make chips architected for AI applications. Currently no competitors are shipping alternatives to Nvidia. We expect AMD with its graphics expertise will likely be the next entry. I am not sure where ARM is but I would bet they get there before Intel.
WHY AI, and what is the big deal?
AI applications bring very high ROE to the customer. This rapid payback of AI applications creates a reason for IT budgets to hold or increase. Much of our IT infrastructure needs retooling.
AI has been a holy grail of the information industry almost since inception. It was present during the Carter administration when I entered the business. At that time, AI meant “Fuzzy Logic,” a very labor-intensive programming concept intended to provide intelligent answers without unduly taxing the computer. It was way too early.
Then, Japan’s MITI central planning group took lead in the AI charge with fourth generation languages and a vision for massive storage farms. Fourth Generation Languages faded to black by the late 1980s as natural language computer programming became pervasive. The storage farms happened, from there AI’s launchpad formed.
Generative AI expands its data sets with findings from prior models. Traditional AI (including machine learning) does not generate additional data. To us this means unit demand for storage will accelerate.