deep? If you miss the last AI launch from Deepseek, you are not alone.

Deepseek updated the R1 AI a few days ago. It is better and still cheaper than most other models.

Did you miss it? I missed it. Or I saw the news shortly and then forgotten. Most of the technology industry and investors have received launch by ignoring a giant.

This is a somewhat blatant contradiction in early 2025 when I feared the Deepseek’s R1 model. Technology shares have decreased and the male penitent spending boom was seriously interrogated.

This time, Ross Sandler, chief technology analyst in Barclays, wrote in a note for investors.

“The stock market does not care less,” he added. “This tells us that the level of understanding the investment community in the trading of artificial intelligence may significantly improve in just five months.”

Unlawful poll

My teammates on Business Insider on Friday, just to see if I spend a lot of time watching Elus Musk and Donald Trump on social media (instead of doing my real job).

Here are some of their responses:

• One of the editors said that they did not notice the Dibsic update, but now they feel guilty for not discovering it. (Solid thinking. Just the madness of bone staying in the press).
• Another colleague said that they know this from their quick examinations, but they did not read much about it.
• A Reddit Technical Technical Reporting saw him, wiped it, and did not think about it again.
• Another correspondent said they completely missed them.
• Another editor: “TBH did not notice!”

Therefore, barely registered. These people are glued to technology news every second of the day.

Why no one really care now?

The latest R1 of Deepseek is the third best in the world at the present time, so why are not waves as before?

Sandler, Barclays analyst, indicated that the latest Depsik’s offers are not just as cheap as she was relatively talking. It costs just less than $ 1 per million symbols, which was approximately 27 times the OPENAI of OPENAI earlier this year.

Now, R1 Deepseek is “only” about 17 times cheaper than the highest model, according to Barclays Research and Data than the artificial intelligence index.

This shows a wider and more important point. Something I have told you since last year: Most of the artificial intelligence models are almost similar in performance because it was mostly training in the same data from the Internet.

This makes it difficult to stand out from the crowd, based on performance. When you jump forward, your inventions and gains are quickly integrated into anyone else’s offers.

The price is important, yes. But the distribution has become a key. If the employer has a Chatgpt account for the institution, for example, Openai models are likely to use at work. It is easier. If you have an Android smartphone, you will likely talk to Google Chatbot Gemini and have giant AI models to search.

Deepseek does not have this type of wide distribution yet, at least in the Western world.

Was the infrastructure of artificial intelligence in its place?

After that, there is awareness that “thinking” models, such as Deepseek’s R1 and Openai’s O3, require a huge amount of computing power. This is due to their ability to divide requests into multiple “thinking” steps. Each step is a new type of claim that is converted into a large number of new symbols that must be addressed.

Deepseek Freakout event mostly because the technology industry is concerned that the Chinese laboratory has developed more efficient models that do not need a lot of computing infrastructure.

In fact, this Chinese laboratory may instead helped to generalize these new types of thinking models, which may require more GPU equipment and other computing tools to operate them.

Subscribe to The weekly newsletter of technology BI BI is here.