AI is a Commodity Product
At the end of the day, an LLM is the aluminum soda can or the HFCS inside it, not the fizzy branding that separates Dr Pepper from Mr. Pibb.
Every week brings highly-hyped launches and announcements from the AI giants. This week, it’s Gemini 3, from Google. Other weeks it’s been the GPT-OSS models from OpenAI, or the announcement of some big new operating agreement where chip companies, AI companies, and cloud service providers cycle massive amounts of money between each other. Signs of a bubble? Most would say yes, and yet the drive for capacity and scale continues.
Will investors start demanding returns? Will capacity keep up with demand? Will promised productivity boosts and cost savings ever materialize? All pressing issues that call into question what the future of AI will look like, and who the eventual winners and losers will be.
But even as we mint more and more celebrity AI tech entrepreneurs and founders, there’s something else to consider: the mostly unspoken third rail of the AI industry is that, in the long run, AI is a commodity product.
And that’s a very different business than the one most people in the AI industry think they’re in.
The unspoken third rail of the AI industry is that, in the long run, AI is a commodity product.
My mind goes here because of my early days as a trade magazine reporter covering the beverage industry, where scintillating topics included the shape of wheel wells on beverage delivery truck bodies and the ratio of surface area to volume in different sizes of PET bottles.
Working at a trade magazine, as many writers and editors of my (and earlier) generations did, taught us how to find stories in the driest of topics, how to focus on the mechanics of reporting, writing, and editing, and in many cases, taught us the basics of business journalism. And the key business lesson for many industries was: What is a commodity product and what isn’t?
A commodity product is one where the brand name doesn’t matter terribly — it just needs to work. And in those cases, the buying decision comes down to price.
RAM is largely a commodity product, so are hard drives. From back in my days as an Assistant Editor at Beverage World magazine (founded in 1882 as the National Bottlers’ Gazette!) aluminum is a commodity product, so is sugar or corn syrup. Arguably, premium colas are a differentiated (non-commodity) product; ginger ale is not.
Dear reader...at this point in my initial draft, I went on a long tangent about the difference between The Coca-Cola Company and its main bottler (up until 2010 a separate-but-related company called Cola-Cola Enterprises) -- one handled the secret formula and the marketing/advertising, the other did the actual dirty, low-margin, high-cost work of manufacturing, bottling, and distribution. In this analogy, AI and startup branding was the former, actual LLM tech the latter. I have, in the interests of your sanity, excised this section.
How does this play out for AI?
AI tech will continue to shift towards being integrated invisibly into the hardware and software you use regularly, rather than interacted with through standalone apps and brands. In that scenario, things like Time to First Token and Tokens per Second will matter, as will the quality of the AI output, but the brand names won’t.
There will be no loyalty for particular LLM models and versions, with Qwen up one week and Gemma up the next. It takes seconds to swap models in an on-device environment like LM Studio, and yesterday’s LLM is about as useful as yesterday’s newspaper (to use a deliberately dated metaphor).
Just as Apple is reportedly replacing the inner workings of Siri with Google’s Gemini, consumer-facing brands will continue to switch AI providers or models around under the hood without the end user caring much one way or another. No AI brand is “cool.” None have cachet.
Apple is a differentiated brand. So are Google, Microsoft, Meta, etc. These companies may continue to develop their own AI tools, or they may slot whoever has the best current model into their consumer and business-facing products. The actual LLMs evolve so quickly and haphazardly that sticking with the same one almost becomes a competitive and perceptual disadvantage.
How do we know this? Many small-to-mid-size AI startups are basically marketers making wrappers for existing AI models, because they know that the dirty, low-margin, high-cost work of building the actual AI part is...a commodity business.
And that’s a distinction worth keeping in mind as AI bubble talk grows louder, even as Google’s CEO says there are “elements of irrationality“ in AI investment. Or there may be an even starker road sign for those who know where to look -- clips from two bubble-busting films, Margin Call (2011) and The Big Short (2025) have suddenly taken over my TikTok feed.
Gift Guide Season is Here
I’ve been writing, editing, assigning, and/or publishing Holiday Gift Guide articles for 20-plus years. No doubt you’ve been inundated with them this month, especially if you’re looking for early Black Friday or holiday deals.
Some people like to structure them around hypothetical giftees (“Best gifts for the home cook”), others use broad categories "(“Best gaming gifts for 2025”). I’d be the first to admit it’s largely an exercise in SEO optimization driven by publications’ affiliate commerce needs.
So for the 2025 version at Micro Center News, I took some key categories — Local AI PC Building, Cards & Collectibles, etc. and assigned them out to some of my favorite freelancers and told them to just go with their guts. I saved the laptops category for myself, of course. You can see some of them here, with more going live over the next week or so:
Holiday Gift Guide 2025: The Best PC Parts for the Local AI PC Builder
Holiday Gift Guide 2025: The Best STEM Gifts for Kids and Makers
Holiday Gift Guide 2025: The Best Streaming and Podcasting Gifts
In the next newsletter, I’ll pinpoint a few of my favorite tech products of the year.
One More Gift Idea — Techlandia: The Game of Smartphones and Supernatural Horror
But wait, there’s more! If were doing a “one fun fact about me” exercise, I might mention that during the pandemic, I created a Kickstarter tabletop game, called Techlandia. It was funded, manufactured, and fulfilled, but even after all that, I’ve got a few cases of game left and they make great stocking stuffer gifts.
Techlandia is a solo-friendly co-op dungeon crawler for up to four players, where you infiltrate the high-tech headquarters of the all-powerful Techlandia corporation and stop the launch of a new smartphone that just might destroy the world...
You can order your own copy here for $19 — about half the original price. I even made a new promo video for this Holiday 2025 special offer:
The Cheat Sheet
> Read my NYT-reviewed gaming history book here: The Tetris Effect
> Threads: threads.net/@danack
> Bsky: danackerman.bsky.social
> LinkedIn: linkedin.com/in/danackerman
> TikTok: tiktok.com/@danacknyc
> YouTube: youtube.com/danackerman
> IG: instagram.com/danack
> MC News: microcenter.news





