AI Infrastructure —
Share
What changed in the AI trade, and why it mattered.
Memory chip makers surge as Micron warns of shortages running through 2027
A multi-year supply squeeze in memory reinforces pricing power across the memory chain, which is why Sandisk and SK Hynix jumped sharply and why Intel's reported memory manufacturing talks with SK Hynix carry weight for the AI hardware buildout.
Compute · Memory· MU, INTC, SNDK, 000660.KS· Source 1· Source 2
Semicap equipment stocks jump on India expansion and broad AI capex forecasts
Applied Materials' $5 billion India investment and UBS's near $1 trillion AI capex call this year, rising to $1.4 trillion by 2027, point to sustained equipment demand, and the move showed up across the group with Lam Research, Applied Materials, Onto Innovation, Teradyne, Amkor and KLA all closing sharply higher.
Semiconductor equipment· AMAT, LRCX, ONTO, TER, AMKR, KLAC· Source 1· Source 2
Arista raises full year outlook on sustained AI data center demand, networking names rally
A guidance raise from a bellwether networking supplier signals demand is not slowing, and it coincided with strong moves in Coherent and Credo, though the sector overall remains a mixed picture rather than a uniform advance.
Corning signs multibillion dollar fiber deal with Verizon for AI and broadband
Optical and fiber capacity is a bottleneck for AI data center buildouts, and this deal alongside Ciena's target of $14 billion revenue by 2029 on optical demand outrunning supply points to durable order books for the physical layer of AI networks.
GE Vernova sees $200 billion backlog by early 2027 as power demand surges, Oracle locks in 1.7GW of Texas wind
Power availability is a key constraint on AI data center growth, so a large equipment backlog and a hyperscaler securing long-term wind supply both point to infrastructure providers moving to meet that demand.
U.S. House passes bill requiring AI data centers to pay for grid upgrades
If adopted by states within the two year window, this would shift grid upgrade costs onto data center operators, raising the cost of new AI infrastructure buildouts and potentially slowing the pace of new site development.
Data-centre infrastructure · Power· Source
Qualcomm fell sharply while the rest of compute held firm
The decline was isolated to Qualcomm rather than spreading through the compute group, which stayed strong and unchanged on the week, suggesting a name specific move rather than a shift in compute demand.
Compute· QCOM
Semicap strengthened sharply this week while broader AI infrastructure names were roughly steady
Semicap firmed up notably from a week ago with Applied Materials, KLA and Onto Innovation all turning positive, while memory and servers stayed strong and dc-infrastructure and neocloud remained split, showing the rally is concentrated in equipment and memory rather than broad based.
Semiconductor equipment · Memory · Servers · Data-centre infrastructure · Neoclouds· AMAT, KLAC, ONTO
Published 19/09/2026, 08:15 UTC · claude-sonnet-5