REAL-TIME GLOBAL RESEARCH
US Hardware / Semiconductors Sector Specialist Commentary
Research evidence excerpt
US Hardware / Semiconductors Sector Specialist Commentary
$168B in Q2, up 74% YoY), and while Meta may rent out excess compute via a new cloud business—
which some read as a sign it overbuilt—most analysts think it’s a strategic option rather than a retreat from AI spending.
Markets are jittery about any slowdown signal anyway: chip and memory stocks sold off sharply on a single Bloomberg
report, showing how exposed the broader market has become to continued hyperscaler spending - The Wall Street Journal
• China's DeepSeek developing its own AI chip, sources say.DeepSeek is developing its own AI inference chip to cut reliance
on Nvidia and Huawei amid U.S. export controls, joining global peers in pursuing custom silicon, though the effort is early-
stage and faces steep manufacturing and memory-access hurdles. The move coincides with DeepSeek’s first-ever outside
funding round — reportedly $7B at a $52-59B valuation — signaling a strategic pivot from pure model research toward
vertical hardware integration and commercialization - Reuters
• Chinese AI models are gaining ground with U.S. companies as costs surge. Chinese open-source/open-weight models
(DeepSeek, Z.ai’s GLM 5.2, Alibaba’s Qwen) are rapidly gaining share among U.S. companies via platforms like OpenRouter
and Vercel—usage jumped from ~11% average to over 30-46% weekly since February—as they close the performance gap
(Brookings estimates 6-9 months behind frontier U.S. labs) while running 60-90% cheaper, prompting real production
switches. This is occurring against a backdrop of rising U.S. frontier model costs and tightening government oversight of
American AI exports, creating a bifurcated market where cost-conscious enterprises risk being pushed toward Chinese models
The English excerpt is extracted automatically from the cited source page and may contain layout or recognition errors. It is never batch translated.
Open report viewer