We are midway through 2026 and it feels like we are on the same track as 2025 regarding the AI ecosystem — major AI tech players doing their thing: buyouts, new tools, new models, etc.

With this post I’m going to try to predict how the AI scene would look by the end of 2026 — if that is even possible given the speed things are moving.

My inspiration was 2 posts on HN I saw today (22/04/2026):

• The first was the news that Elon Musk is in talks to buy Cursor for some billions (link).
• The second was that for some users (2% to be precise) Anthropic wasn’t offering Claude Code for Pro users (link).

Let’s talk a bit more about those 2.

Buyouts are not slowing down

It seems that the buyouts of smaller companies (smaller than the AI giants) are continuing into 2026. In 2025 and early 2026, OpenAI for example made a bunch of acquisitions of either AI startups or AI tools like Openclaw. This trend is a continuation of the acquisitions that happened in previous years due to the explosion of AI usage and popularity. Almost all big players (or at least OpenAI and Anthropic) are spending billions on top talent and tools that can eventually:

• a) get more users using and buying their product
• b) drive progress in AI tech

And I ask: who’s going to pay for those spendings? The answer is obvious — you and me, the people using AI tools (i.e. vibe coding), either for fun (hobbyists) or for professional work.

Subscriptions are next

The latter ties in with the second post about Anthropic. In a later statement (through X), Anthropic admitted they were basically running an A/B test for new users (or prosumers as they call them). Yes, this is one side of the coin. If one reads between the lines — or beyond the statement and the user comments in the threads — you can sense that some fundamental changes are coming to AI tool subscriptions.

Take for example the memory situation: major memory chip providers turned their backs on consumers and made datacenters serving AI their number one priority. This could be a sign that smaller players (us developers) are second-class citizens.

The same can apply to subscriptions: the ~$20 that most paid plans cost is too little for AI companies. This could trigger a push toward pricier subscriptions (Max plans at $100+) or giving priority to organizations. This is backed up by the news of Microsoft stopping new sign-ups for educational Copilot subscriptions (link).

My prediction

My prediction for 2026 (or early 2027) is that subscriptions will get pricier and getting inference slots will get trickier.

Of course, there is always the option of using local LLMs (good luck with current GPU/RAM prices) or going with cheaper and more available Chinese LLMs — same story with the new emerging Chinese memory chip companies that stepped in to fill the void.

Time will tell, as always. But one thing is for sure: long gone are the days of 24/7 running AI tools and endless vibe coding. We must adapt — i.e. start writing more code by hand again like we used to!