[conspire] Discussion: Using LLMs the Right Way: 10/1/2025 7pm Eastern Daylight time

[Ivan Sergio Borgonovo]

On 10/5/25 3:59 AM, Deirdre Saoirse Moen wrote:
> 
> 
>> On Oct 4, 2025, at 12:32, Steve Litt <slitt at troubleshooters.com> wrote:
>>
>> I don't understand the point of the preceding paragraph.
> 
> Perhaps you could get one of them fancy LLMs to explain it to you.

DONE

This situation in Oregon is highly relevant to Large Language Models 
(LLMs) and data centers because it highlights the growing **conflict 
between the massive energy demands of the tech industry, particularly 
AI, and the capacity and cost of the public power grid.**

Here is a breakdown of the relevance of each point:

### 1. The Energy Demand of LLMs and Data Centers

* **LLMs are Power-Hungry:** Large Language Models (like the one you are 
interacting with) require immense computational power for both 
**training** and **inference** (running the model for user queries). 
This computation is housed in **data centers**.
* **Data Centers as "Voracious" Users:** Data centers, especially those 
hosting AI and cloud computing, are industrial-scale operations that 
consume as much electricity as small-to-medium-sized cities. The demand 
from a single, large-scale data center can be in the tens or even 
hundreds of megawatts (MW). For example, one large data center can use 
the energy equivalent of tens of thousands of homes.
* **The AI Boom Exacerbates Demand:** The recent explosion in generative 
AI has created a new, rapidly accelerating demand for electricity. Some 
projections suggest data center energy consumption could *double or 
triple* in the next few years.

### 2. Oregon's Attractiveness and Subsequent Strain

* **The Hydro Lure:** Oregon historically attracted data center 
companies (like Google and Amazon Web Services) because of its 
**relatively cheap and clean hydropower electricity** from the Columbia 
River dams and a cool climate ideal for cooling servers. This was a 
major business incentive.
* **Exceeding Capacity:** The original abundance of electricity and low 
rates began to be strained by the massive and continuous influx of data 
centers. Even though hydropower is clean, it is a **constrained 
resource** that can be "maxed out." This rapid and dramatic load growth 
has put enormous stress on the local power grid and required expensive 
upgrades to transmission and other infrastructure.

### 3. Relevance of the New Laws (The POWER Act)

The laws you mentioned are a direct regulatory response to the energy 
demands of large users like data centers:

* **Strict Limitations on Construction (The POWER Act):**
     * Oregon passed legislation (like the **POWER Act**, or HB 3546) to 
create a **separate customer classification and rate class** for "large 
energy use facilities" (data centers, cryptocurrency, etc., typically 
those using over $20$ MW).
     * The primary goal is to **stop shifting the high costs** of grid 
expansion and new infrastructure (needed to serve these massive new 
loads) onto residential and small business customers, whose rates have 
reportedly risen by as much as $50\%$ in recent years.
     * The law effectively **limits growth** or makes it significantly 
**more expensive** for new, large-scale data centers to be built, 
especially those powering LLMs and AI, which rely on the lowest possible 
electricity rates to be profitable.

* **Rate Caps (Winter/15-Month Limits):**
     * These rate-capping measures are aimed at **protecting residential 
ratepayers** from the volatility and significant increases in electric 
bills that have been driven, in part, by the surging demand from data 
centers.
     * By ensuring residential stability, the state is essentially 
drawing a line, asserting that the expansion of the tech industry will 
no longer be subsidized by other ratepayers.

In summary, the Oregon laws represent a crucial moment where the 
environmental and economic cost of the **AI and LLM boom** is becoming a 
significant political issue, forcing a state known for cheap power to 
actively **limit the expansion of the very industry it once heavily 
courted.**


-- 
Ivan Sergio Borgonovo
https://www.webthatworks.it https://www.borgonovo.net