Average age of first time home buyer rises to 40

As a father with adult-age children, the average age of the first-time home buyer rising to 40, really hit home. I purchased my first unit/condo in my early 20's and my first home before I was 30. This was one story among many that painted the picture of two Americas - one where people like me who have significant money in equities have been doing well, and another America where jobs are hard to get, real estate is unattractive to buy, inflation hits the pocketbook, and more. Some of that may be shaking out as the markets focus more on consumer strength and the echos of the telecom bust seem to be everywhere in AI (though it may be that institutional investors are running for the exits faster than retail investors). Nonetheless, an economy that lifts all boats is for cultural and practical reasons desirable.

Google Ironwood AI chip available to public

One conventional wisdom has been that the AI chips made by the hyperscalers were optimized for training their specific workloads with their massive amounts of data, for example recommendation systems etc. ==> simple is subversive. While that remains somewhat true, Google's Ironwood chip, announced earlier in the year but now available to customers, continues to lean more into inference with its dual-die architecture of SparseCore and Tensors combined with memory parallelism and system/pod-level optimizations. While the issue of chip/architecture is intriguing, another issue stuck out as well.

Some of the performance metrics quoted by Google are for FP8 (8-bit floating point). While this represents anticipated movement of the American market from FP16 and progress to greater efficiency, this week was also a week where the new Kimi 2 model was talking about INT4 performance. So while the gap may close between the bleeding edge Chinese models and the American models, the conversation around China pushing the envelope will likely continue in the near term.

Google Gemini with built-in managed RAG

Google announced this week its new filetool, built-in managed RAG (retrieval augmented generation) for Gemini. RAG has promised much, but not everyone is convinced it has delivered. In addition, there are nuances to how it is set up. If Google can make it easy to do, while producing good results, this moves the needle on grounding LLMs with data. I will be interested to see how this plays out for Google which is putting increasing focus on its Enterprise solutions.

Overall summary of this week's news

The below is an AI generated overview of my curated collection of news stories which can be found at "Consider This". Enjoy and have a great weekend ...

This week's news paints a picture of a rapidly evolving technological landscape dominated by the AI arms race, juxtaposed against a backdrop of growing economic concerns. The major tech giants are pouring arguably unprecedented resources into AI infrastructure, signaling a significant shift in their capital expenditure and a potential reshuffling of market dominance. Simultaneously, the broader economy is showing signs of strain.

AI Infrastructure Boom & Bottlenecks

The relentless demand for AI compute is driving massive investment in data centers and specialized chips. Hyperscalers like Amazon, Microsoft, and Google are set to spend nearly $500 billion annually by 2026 on infrastructure. However, this boom is not without its challenges. Power availability is emerging as a critical bottleneck, and companies like Tesla are pursuing vertical integration [Note: floating prospect of building its own fab / having discussions with Intel] to control their chip destiny / expected volume. The race for more powerful and efficient AI chips is intense, with Tesla's AI5 chip challenging industry leaders and TSMC implementing significant price hikes for advanced nodes.

Economic Headwinds and Market Concentration

The housing market remains frozen due to the "rate lock-in" effect, and consumer debt is at record highs, driven by mortgages, credit cards, and student loans. This financial strain is impacting retail sales, as evidenced by declining job postings and a slowdown in China-US trade [Note: The "consumer" is increasingly nuanced with different categories/layers]. The stock market, meanwhile, is showing extreme concentration, with the top 10 S&P 500 holdings accounting for a record 42% of its value. This suggests a widening gap between the performance of a few tech giants and the rest of the market. Some are giving the equal-weighted SP&500 more attention.

Flight to Safety and Shifting Investment

A trend over recent years has been strong inflows in to cash funds, bonds, and gold, Despite this, big tech equities have seen huge gains. In the context of visibility strugglels for Data stemming from the US government shutdown, spiking unemployment claims from Federal workers, and more chatter about institutional investors "trimming" their big tech holdings, this was a week where there was increased chatter about an AI bubble, despite great earnings reports and raised capex forecasts from most of the main players. There is also more chatter about a no-hire, no-fire environment potentially pivoting to no-hire, some-fire. The controversy over whether OpenAI did or did not request government backing is also being cited as a confidence hit.

Overall Takeaway

The dominant narrative is one of intense technological acceleration in AI, coupled with significant economic caution and market polarization. The "AI infrastructure arms race" is the primary driver of innovation and investment, particularly for the tech giants. However, this is occurring against a backdrop of increased chatter about weakening consumer finances (though, what is the "consumer"). The coming months will likely see continued focus on how these competing forces – technological advancement and economic headwinds – play out., mixed in with all the usual money/equity manager dynamics of closing out the year and pursuing strategies for overall year-end performance.

For more information see: "Consider This"