AI and Energy Infrastructure, Explained Plainly, The Physical Layer

The Physical Layer

AI and Energy Infrastructure, Explained Plainly

A journal on the physical constraints behind the AI buildout: power, grid connection, data centres, cooling and the supply chains that gate them. Written to be understood by anyone.

Equity Bottom-Up

TSMC (Taiwan Semiconductor Manufacturing) / Equity Bottom-Up

TSMC (Taiwan Semiconductor Manufacturing) (2330 TT)

TSMC (Taiwan Semiconductor Manufacturing)

<ul><li>Almost every advanced AI chip, whoever designs it, is made by one company, TSMC, mostly in Taiwan.</li><li>Its chief executive says those factories, not electricity, are what limits AI today, and the company's own numbers show why: a new chip factory takes years to build and more years to fill.</li><li>That is a fair point, and it changes how long I think the chip side of the shortage lasts.</li></ul>

Almost every advanced AI chip, whoever designs it, is made by one company, TSMC, mostly in Taiwan.Its chief executive says those factories, not electricity, are what limits AI today, and the company's own numbers show why: a new chip factory takes years to build and more years to fill.That is a fair point, and it changes how long I think the chip side of the shortage lasts.

TSMC's chief says the AI bottleneck is its factories, not the power grid

AI | Semiconductors | Commodities, Mint Finance

Raghav Vashisht

AI | Semiconductors | Commodities

Deep Fundamental Analysis

Mint Finance

Mint Finance's investment research blends deep fundamental analysis with technical & quant insights. Investment strategies straddle across multiple asset classes with risk management embedded into every trade to deliver an outsized reward-to-risk ratio.

Mint Finance's research is centered around three key themes. First, energy transition and climate change. Second, artificial intelligence. Third, shifting world order with its impact across politics and monetary economics.

Thematic (Sector/Industry)

<ul>
<li>With inference workloads scaling, NAND flash and HBM are now central to AI infrastructure, reshaping demand, pricing, and capex.</li>
<li>SSD demand is booked through 2026, HDD lead times stretch to 52 weeks, and persistent storage (commodity, you say?) has become mission-critical for inference data trails.</li>
<li>HBM to cross 43% of DRAM TAM, while packaging complexity and competition from Chinese players raise fresh and long-term questions on valuation and defensibility.</li>
</ul>


With inference workloads scaling, NAND flash and HBM are now central to AI infrastructure, reshaping demand, pricing, and capex.
SSD demand is booked through 2026, HDD lead times stretch to 52 weeks, and persistent storage (commodity, you say?) has become mission-critical for inference data trails.
HBM to cross 43% of DRAM TAM, while packaging complexity and competition from Chinese players raise fresh and long-term questions on valuation and defensibility.


From Commodity to Catalyst: Memory as the New Bottleneck

Weekly podcast series by SemiWiki.com, SemiWiki.com

SemiWiki.com

Weekly podcast series by SemiWiki.com

Semiconductor Insiders: A weekly podcast series by SemiWiki.com, the open forum for semiconductor professionals.

<ul><li>RISC-V allows for design and implementation locally, compliant to global standards, with specifications available under Creative Commons and no licensing fees</li><li>RISC-V's customizable architecture, with a small core base ISA and various extensions, enables workload optimized silicon for different applications such as AI, ML, and vector processing</li><li>Recent highlights include the availability of the RBA23 profile specification for high-performing application processors, Nvidia's adoption of RISC-V for CUDA porting, and Infineon's announcement of an all RISC-V automotive microcontroller roadmap, showcasing the growing momentum of RISC-V across various industries.</li></ul><p><em>This content is sourced through publicly available sources and has been machine generated. Information displayed is for general informational purposes only.</em></p>

RISC-V allows for design and implementation locally, compliant to global standards, with specifications available under Creative Commons and no licensing feesRISC-V's customizable architecture, with a small core base ISA and various extensions, enables workload optimized silicon for different applications such as AI, ML, and vector processingRecent highlights include the availability of the RBA23 profile specification for high-performing application processors, Nvidia's adoption of RISC-V for CUDA porting, and Infineon's announcement of an all RISC-V automotive microcontroller roadmap, showcasing the growing momentum of RISC-V across various industries.This content is sourced through publicly available sources and has been machine generated. Information displayed is for general informational purposes only.

Podcast EP309: The State of RISC-V and the Upcoming RISC-V Summit with Andrea Gallo

Derivatives and Quantitative Strategist

Gaudenz Schneider

Equity Derivatives

<ul>
<li>
<a href="https://www.smartkarma.com/entities/sk-hynix-inc" target="_blank" data-smartkarma-citation="entities:16955:1" rel="noopener">SK hynix (000660 KS)</a> gained 11.2% in September MTD and its 2x Leveraged Index rose 20.9%, but combined AUM increased only 10% as redemptions offset performance.</li>
<li>Units outstanding fell across all seven long leveraged ETFs. In contrast, units in the inverse product more than doubled and its market capitalisation rose 56%, despite negative performance.</li>
<li>Annualised 21-day volatility more than halved from its late-July peak of 179% to 55%. The decline coincided with the collapse in ETF trading activity and fund flows following regulatory changes.</li>
</ul>



SK hynix (000660 KS) gained 11.2% in September MTD and its 2x Leveraged Index rose 20.9%, but combined AUM increased only 10% as redemptions offset performance.
Units outstanding fell across all seven long leveraged ETFs. In contrast, units in the inverse product more than doubled and its market capitalisation rose 56%, despite negative performance.
Annualised 21-day volatility more than halved from its late-July peak of 179% to 55%. The decline coincided with the collapse in ETF trading activity and fund flows following regulatory changes.


SK Hynix (000660 KS): Volatility Falls to 55% as Leveraged ETF Activity Fades

Global Materials | Industrials | Commodities

Rahul Jain

<ul>
<li>Asia mirrors the U.S. “boring tech” rally: storage, memory, and components once seen as cyclical are being re-rated as AI infrastructure enablers.</li>
<li>Valuation gaps present opportunities: Kioxia and SK Hynix trade at meaningful discounts vs Seagate/WDC/Micron, while Murata and TSMC command premiums; Chinese names appear expensive.</li>
<li>Key risks remain: cyclicality in DRAM/NAND, potential subsidy overbuild, and currency volatility could pressure valuations if AI demand slows.</li>
</ul>


Asia mirrors the U.S. “boring tech” rally: storage, memory, and components once seen as cyclical are being re-rated as AI infrastructure enablers.
Valuation gaps present opportunities: Kioxia and SK Hynix trade at meaningful discounts vs Seagate/WDC/Micron, while Murata and TSMC command premiums; Chinese names appear expensive.
Key risks remain: cyclicality in DRAM/NAND, potential subsidy overbuild, and currency volatility could pressure valuations if AI demand slows.


Asia’s “Boring Tech” Rerating: Memory, Components & AI Infrastructure

United States

Micron Technology, Inc., through its subsidiaries, manufactures and markets dynamic random access memory chips (DRAMs), very fast static random access memory chips (SRAMs), Flash Memory, other semiconductor components, and memory modules.

Price Gauge
52-week Range	0.00 - 0.00
Market Cap (USDm)	0.00
P/E	0.00
Dividend (Yield)	0.00

Micron Technology (MU US)

TSMC's chief says the AI bottleneck is its factories, not the power grid

Almost every advanced AI chip, whoever designs it, is made by one company, TSMC, mostly in Taiwan. Its chief executive says those factories, not...

The Physical Layer

From Commodity to Catalyst: Memory as the New Bottleneck

AI infrastructure is increasingly reliant on NAND flash and HBM, reshaping demand and pricing, with SSD demand booked through 2026 and HDD lead...

Mint Finance

Podcast EP309: The State of RISC-V and the Upcoming RISC-V Summit with Andrea Gallo

RISC V is a customizable and globally recognized architecture used by companies like Nvidia, with ongoing development and expansion in various...

SemiWiki.com

SK Hynix (000660 KS): Volatility Falls to 55% as Leveraged ETF Activity Fades

Despite positive performance, some long-leveraged ETFs lost AUM, while the inverse product attracted significant assets. Collapsing investor...

Gaudenz Schneider

Asia’s “Boring Tech” Rerating: Memory, Components & AI Infrastructure

Asia's tech rally focuses on storage, memory, and components as AI infrastructure enablers, with valuation gaps offering opportunities but risks...

Rahul Jain

Price Gauge

52-week Range

Market Cap (USDm)

P/E

Dividend (Yield)

TSMC's chief says the AI bottleneck is its factories, not the power grid

Almost every advanced AI chip, whoever designs it, is made by one company, TSMC, mostly in Taiwan. Its chief executive says those factories, not...

The Physical Layer

From Commodity to Catalyst: Memory as the New Bottleneck

​AI infrastructure is increasingly reliant on NAND flash and HBM, reshaping demand and pricing, with SSD demand booked through 2026 and HDD lead...

Mint Finance

Podcast EP309: The State of RISC-V and the Upcoming RISC-V Summit with Andrea Gallo

RISC V is a customizable and globally recognized architecture used by companies like Nvidia, with ongoing development and expansion in various...

SemiWiki.com

SK Hynix (000660 KS): Volatility Falls to 55% as Leveraged ETF Activity Fades

Despite positive performance, some long-leveraged ETFs lost AUM, while the inverse product attracted significant assets. Collapsing investor...

Gaudenz Schneider

Asia’s “Boring Tech” Rerating: Memory, Components & AI Infrastructure

​Asia's tech rally focuses on storage, memory, and components as AI infrastructure enablers, with valuation gaps offering opportunities but risks...

Rahul Jain

AI infrastructure is increasingly reliant on NAND flash and HBM, reshaping demand and pricing, with SSD demand booked through 2026 and HDD lead...

Asia's tech rally focuses on storage, memory, and components as AI infrastructure enablers, with valuation gaps offering opportunities but risks...