Introduction
In a week where the VIX went parabolic up:
In a week where the VIX went parabolic up:
VIX volatility index went sharply up this week
And Crypto went in the other direction:
Bitcoin / Crypto went sharply down this week
We can start to see the setup for 2026 - S&*^'s getting real. The markets have gone from risk on to risk off, and from everyone's a winner, to we are going to start picking and choosing.
In 2024, inference was too slow, too expensive, and too random. Significant progress has been made on all fronts, with 2025 being a magical year in the progress of the current approach to GenAI, even if it is nto the path to AGI. Now, demand is greater than supply. There is no token bubble, but there may have been a market bubble, which are two different things.
Not just because I have been going on about efficiency and complexity for some time and therefore have cognitive bias, I feel in my bones we are going to see much more about efficiency in 2026. There is already signs of increased chatter about the efficiency of new models. OpenAI was accused - rightly or wrongly - of simply brute force. GPT 5, and Google pushed out a new foundation model on highly optimized TPUs.
Short of a black swan, supply constraints are not likely to disappear over night, so we efficiency will be the order of the day. Scaling laws are not dead. However, efficient scaling is not the worse idea in the world. We can see some signs of the efficiency to come:
These are all signs the supply side is saying: we are ok holding the line on pricing, because we have more demand than we can serve, so if we dampen demand a little, that is OK.
In my gut, I feel like 2026 is going to be a big year on the energy front in terms of transistions and mix, but more on that later.
Below is an AI-generated summary of the issues that got my attention this week, the full dashboard you can see here: https://www.bohcay.com/library/consider-this
In late 2025, rapid AI advancements—from groundbreaking systems like SciAgent and AgentEvolver that enable self-improving, multi-agent reasoning to efficient TPUs and RAG innovations—have fueled unprecedented productivity and infrastructure demands, yet they coincide with mounting economic strains, including surging credit card delinquencies, record-high CDS spreads for AI companies, and a volatile stock market where Nvidia oscillates wildly while high-quality stocks underperform amid fears of overvaluation and defaults [Note: there has been a pivot to quality this week].
The AI boom has driven massive data center expansions and energy consumption, straining grids and pushing electricity rates upward, while paradoxically accelerating global transitions to solar and wind dominance, with China leading in renewables adoption and fossil fuel declines, though crypto volatility—exemplified by Bitcoin's sharp drawdowns and outflows—reflects speculative bubbles tied to AI hype. Amid layoffs at tech giants like Amazon, rising bankruptcies, and housing depreciation, consumer spending falters, yet innovations like AI-powered travel bookings and stablecoin integrations by Visa and Mastercard signal a shift toward digital economies, underscoring how AI's transformative potential is reshaping industries while exposing vulnerabilities in finance, energy, and labor markets.
See a dashboard of interesting news at: https://www.bohcay.com/library/consider-this