Deepseek Is Cheap. People Like Cheap
October 1, 2025
![green-dino_thumb_thumb[1]](https://www.arnoldit.com/wordpress/wp-content/uploads/2025/09/green-dino_thumb_thumb1.gif "green-dino_thumb_thumb[1]")
This essay is the work of a dumb dinobaby. No smart software required.
I read “Deepseek Has ‘Cracked’ Cheap Long Context for LLMs With Its New Model.” (I wanted to insert “allegedly” into the headline, but I refrained. Just stick it in via your imagination.) The operative word is “cheap.” Why do companies use engineers in countries like India? The employees cost less. Cheap wins out over someone who lives in the US. The same logic applies to smart software; specifically, large language models.
Cheap wins if the product is good enough. Thanks, ChatGPT. Good enough.
According to the cited article:
The Deepseek team cracked cheap long context for LLMs: a ~3.5x cheaper prefill and ~10x cheaper decode at 128k context at inference with the same quality …. API pricing has been cut by 50%. Deepseek has reduced input costs from $0.07 to $0.028 per 1M tokens for cache hits and from $0.56 to $0.28 for cache misses, while output costs have dropped from $1.68 to $0.42.
Let’s assume that the data presented are spot on. The Deepseek approach suggests:
- Less load on backend systems
- Lower operating costs allow the outfit to cut costs to licensee or user
- A focused thrust at US-based large language model outfits.
The US AI giants focus on building and spending. Deepseek (probably influenced to some degree by guidance from Chinese government officials) is pushing the cheap angle. Cheap has worked for China’s manufacturing sector, and it may be a viable tool to use against the incredibly expensive money burning U S large language model outfits.
Can the US AI outfits emulate the Chinese cheap tactic. Sure, but the US firms have to overcome several hurdles:
- Current money burning approach to LLMs and smart software
- The apparent diminishing returns with each new “innovation”. Buying a product from within ChatGPT sounds great but is it?
- The lack of home grown AI talent exists and some visa uncertainty is a bit like a stuck emergency brake.
Net net: Cheap works. For the US to deliver cheap, the business models which involved tossing bundles of cash into the data centers’ furnaces may have to be fine tuned. The growth at all costs approach popular among some US AI outfits has to deliver revenue, not taking money from one pocket and putting it in another.
Stephen E Arnold, October 1, 2025
AI Will NOT Suck Power Like a Kiddie Toy
October 1, 2025
This essay is the work of a dumb dinobaby. No smart software required.
The AI “next big thing” has fired up utilities to think about building new plants, some of which may be nuclear. Youthful wizards are getting money to build thorium units. Researchers are dusting off plans for affordable tokamak plasma jobs. Wireless and smart meters are popping up in rural Kentucky. Just in case a big data center needs some extra juice, those wireless gizmos can manage gentle brownouts better than an old-school manual switches.
I read “AI Won’t Use As Much Electricity As We Are Told.” The blog is about utility demand forecasting. Instead of the fancy analytic models used for these forward-looking projections, the author approaches the subject in a somewhat more informal way.
The write up says:
The rise of large data centers and cloud computing produced another round of alarm. A US EPA report in 2007 predicted a doubling of demand every five years. Again, this number fed into a range of debates about renewable energy and climate change. Yet throughout this period, the actual share of electricity use accounted for by the IT sector has hovered between 1 and 2 per cent, accounting for less than 1 per cent of global greenhouse gas emissions. By contrast, the unglamorous and largely disregarded business of making cement accounts for around 7 per cent of global emissions.
Okay, some baseline data from the Environmental Protection Agency in 2007. Not bad: 18 years ago.
The write up notes:
Looking the other side of the market, OpenAI, the maker of ChatGPT, is bringing in around $3 billion a year in sales revenue, and has spent around $7 billion developing its model. Even if every penny of that was spent on electricity, the effect would be little more than a blip. Of course, AI is growing rapidly. A tenfold increase in expenditure by 2030 isn’t out of the question. But that would only double total the total use of electricity in IT. And, as in the past, this growth will be offset by continued increases in efficiency. Most of the increase could be fully offset if the world put an end to the incredible waste of electricity on cryptocurrency mining (currently 0.5 to 1 per cent of total world electricity consumption, and not normally counted in estimates of IT use).
Okay, the idea is that power generation professionals are implementing “logical” and “innovative” tweaks. These squeeze more juice from the lemon so to speak.
The write up ends with a note that power generation and investors are not into “degrowth”; that is, the idea that investments in new power generation facilities may not be as substantial as noted. The thirst for new types of power generation warrants some investment, but a Sputnik response is unwarranted.
Several observations:
- Those in the power generation game like the idea of looser regulations, more funding, and a sense of urgency. Ignoring these boosters is going to be difficult to explain to stakeholders.
- The investors pumping money into mini-reactors and more interesting methods want a payoff. The idea that no crisis looms is going to make some nervous, very nervous.
- Just don’t worry.
I would suggest, however, that the demand forecasting be carried out in a rigorous way. A big data center in some areas may cause some issues. The costs of procuring additional energy to meet the demands of some relaxed, flexible, and understanding outfits like Google-type firms may play a role in the “more power generation” push.
Stephen E Arnold, October 1, 2025
Google and Its End Game
October 1, 2025
No smart software involved. Just a dinobaby’s work.
I read “In Court Filing, Google Concedes the Open Web Is in Rapid Decline.” The write up reveals that change is causing the information highway to morph into a stop light choked Dixie Highway. The article states:
Google says that forcing it to divest its AdX marketplace would hasten the demise of wide swaths of the web that are dependent on advertising revenue. This is one of several reasons Google asks the court to deny the government’s request.
Yes, so much depends on the Google just like William Carlos Williams observed in his poem “The Red Wheelbarrow.” I have modified the original to reflect the Googley era which is now emerging for everyone, including Ars Technica, to see:
so much depends upon the Google, glazed with data beside the chicken regulators.
The cited article notes:
As users become increasingly frustrated with AI search products, Google often claims people actually love AI search and are sending as many clicks to the web as ever. Now that its golden goose is on the line, the open web is suddenly “in rapid decline.” It’s right there on page five of the company’s September 5 filing…
Not only does Google say this, the company has been actively building the infrastructure for Google to become the “Internet.” No way, you say.
Sorry, way. Here’s what’s been going on since the initial public offering:
-
- Attract traffic and monetize via ads access to the traffic
- Increased data collection for marketing and “mining” for nuggets; that is, user behavior and related information
- Little by little, force “creators,” Web site developers, partners, and users to just let Google provide access to the “information” Google provides.
Smart software, like recreating certain Web site content, is just one more technology to allow Google to extend its control over its users, its advertisers, and partners.
Courts in the US have essentially hit pause on traffic lights controlling the flows of Google activity. Okay, Google has to share some information. How long will it take for “information” to be defined, adjudicated, and resolved.
The European Union is printing out invoices for Google to pay for assorted violations. Guess what? That’s the cost of doing business.
Net net: The Google will find a way to monetize its properties, slap taxes at key junctions, and shape information is ways that its competitors wish they could.
Yes, there is a new Web or Internet. It’s Googley. Adapt and accept. Feel free to get Google out of your digital life. Have fun.
Stephen E Arnold, October 3, 2025
Will AI Topple Microsoft?
October 1, 2025
At least one Big Tech leader is less than enthused about AI rat race. In fact, reports Futurism, “Microsoft CEO Concerned AI Will Destroy the Entire Company.” As the competition puts pressure on the firm to up its AI game, internal stress is building. Senior editor Victor Tangermann writes:
“Morale among employees at Microsoft is circling the drain, as the company has been roiled by constant rounds of layoffs affecting thousands of workers. Some say they’ve noticed a major culture shift this year, with many suffering from a constant fear of being sacked — or replaced by AI as the company embraces the tech. Meanwhile, CEO Satya Nadella is facing immense pressure to stay relevant during the ongoing AI race, which could help explain the turbulence. While making major reductions in headcount, the company has committed to multibillion-dollar investments in AI, a major shift in priorities that could make it vulnerable. As The Verge reports, the possibility of Microsoft being made obsolete as it races to keep up is something that keeps Nadella up at night.”
The CEO recalled his experience with the Digital Equipment Corporation in the 1970s. That once-promising firm lost out to IBM after a series of bad decisions, eventually shuttering completely in the 90s. Nadella would like to avoid a similar story for Microsoft. One key element is, of course, hiring the right talent—a task that is getting increasingly difficult. And expensive.
A particularly galling provocation comes from Elon Musk. Hard to imagine, we know. The frenetic entrepreneur has announced an AI project designed to “simulate” Microsoft’s Office software. Then there is the firm’s contentious relationship with OpenAI to further complicate matters. Will Microsoft manage to stay atop the Big Tech heap?
Cynthia Murrell, October 1, 2025
-
- Subscribe to Beyond Search
Feature archive
News archive
-
