Google AI Mode Tops 1B Users, Easing Search Disruption Fears

🧠 Original Insights Summary

$GOOGL Search risk looks less binary. AI Mode has surpassed 1B monthly users, and AI Mode + AI Overviews are creating incremental searches instead of cannibalizing existing activity. That clears the first bear-case debate: users are still staying inside Google in an AI-first world. 1
Google should be framed as a long-duration holding, not a next-day trade. If Buffett buys Google or any stock, the playbook is multi-year compounding: at least 3–5 years, not tomorrow or next month’s P&L. 2

Intel 2Q26 reads like a supply-constrained recovery. It posted a seventh consecutive beat, demand is still far ahead of supply, and capex is finally being raised from $18B flat to more than $20B in 2026, with 2027 significantly above. With ASML EUV output growing 30% for the next 2 years, the setup points to capacity finally catching up to demand. 3
The China memory “supply shock” bear case looks overblown. The key logic: China itself is still short of memory, and state-backed buyers are first in line for domestic supply. If CXMT is selling above market pricing, the fear that Chinese DRAM will flood global markets loses force. 4
$AMD and $DELL are showing relative strength despite a $QQQ mini correction. If the Qs bounce, the clean read is momentum rotation back into leaders, with AMD and Dell likely pushing back to ATHs. 5

$T looks ready to commercialize direct-to-device satellite service with $ASTS without waiting for another hardware cycle. The important signal: John Stankey said “We’re not betting on the next turn of a chip”. Translation: the product can compete today, not only after future tech breakthroughs. 6

Open-weight AI is being positioned as a national competitiveness issue, not just an open-source preference. $NVDA, $MSFT, $PLTR, and $NOW argue that relying only on closed models is not inherently safer, because concentrating advanced AI behind a few providers creates single-point risk. 7
Kimi K3 is now assessed as the strongest Chinese model for cyberattack capability in the CAISI evaluation, clearly ahead of GLM-5.2, earlier DeepSeek, Qwen, and Kimi models. The direction is clear: China’s frontier-model cyber capability is improving fast, but still materially trails the most advanced U.S. models from the same period. 8

Fang Xinghai’s downfall may mark a policy reckoning for the securities-lending regime and quant/high-frequency trading issues he previously supported. The read-through: regulatory clean-up may be moving from rhetoric into personnel-level accountability. 9

$MRVL was managed with a clear green-to-red stop. The trade was fully exited, booking just shy of 3% on common shares, with no position left. Clean risk control. No bag-holding. 10