Hello everyone, I'm a grassroots developer who codes without much regard for specific rules, relying entirely on intuition and muscle memory, pounding the keyboard like crazy.

Last night, I stayed up late watching the Google I/O 2026 keynote. At $100 per month, this subscription fee is outrageously expensive for ordinary people. But for developers, it's significantly cheaper than the previous price of $250—still a painful cut, but now a barely affordable productivity tool.

However, the most chilling part of this update isn't the price at all. It's that one casually mentioned line at the end of the article: usage limits will shift from counting by requests to counting by compute (Compute).

The real cold logic is hidden in that single sentence.

1. The era of empty promises on slides is over—the real money harvesting has begun. Previously, the generative AI scene was all about storytelling and painting rosy pictures. Grand narratives about AI changing the world and boosting human productivity tenfold emerged endlessly. You pay $20 a month, and everyone gets to share a sweet dream together.

But deep down, we all know that with AI running 24/7 in the background, writing long code, and performing large-scale refactoring, how could Google possibly recover the server electricity and computing costs at just $20 a month?

The answer is obvious: it was all at a loss before. So Google decisively pivoted this time. In the era where AI agents automatically run automation tasks in the background for you, the pastoral pay-per-query model is bound to collapse. The rules have changed: how much work you have the AI do, and how much computing power it consumes, you have to pay real money for. That’s the underlying logic behind this adjustment.

1. The $100 Plan: A Textbook Bundle Trap At $100 a month, even considering US inflation, ordinary consumers would absolutely not buy it. But Google never intended to sell it to ordinary people; its target is extremely precise — developers and the B2B sector.

The pricing strategy for this plan is extremely cunning. The core lies in that 20TB of cloud storage. You see, previously buying 20TB of Google One separately would cost about $100 as well. So Google’s actual subtext is:

Professional users who originally needed large storage capacity, you continue to pay the $100 storage fee as usual. In return, the latest AI tools (such as Antigravity) and YouTube Premium are on me for free.

From the user's perspective, this is practically a feast of freebies. But from Google's perspective? Carving out an extra 20TB of storage space in their own data center comes at almost zero marginal cost. Using an asset that costs almost nothing, packaged as a $100 giveaway, they precisely shift the huge GPU costs of developing AI onto professional users. While we cheer about getting a great deal, we voluntarily hand over our money to Google to fund their AI war chest. The ones being played are, in fact, ourselves.

1. Ordinary people don't even have the chance to be "leeks" anymore. So, what happens to the freeloaders and regular users who can't afford $100? Are they left behind by the times? Quite the opposite—they will be firmly trapped in an inescapable net.

Google has given up on forcing ordinary people to subscribe to a fourth or fifth paid software. Instead, they have directly embedded AI into underlying systems like Android and Chrome, which are as inescapable as air.

Users may not even realize they are using an AI subscription service. They will simply get used to letting Chrome quietly help them compare prices, book flights and hotels in the background. Until one day, a line suddenly pops up on the screen: "Your computing power quota for today has been exhausted. To continue completing tasks, please pay 5 yuan or upgrade to a premium plan."

Just as mobile games once used stamina to harvest from the entire population, in the future, ordinary people will unknowingly pay bit by bit for computing time in order to have AI handle trivial tasks. Even more insidiously, the underlying systems of e-commerce sites or enterprise internal systems that ordinary people frequently use will all be connected to Google's AI infrastructure. Ultimately, this computing power tax will turn into product premiums and service fees, making us pay indirectly.

1. The upgrade to operators' 5G premium packages is subtly tied to the hidden computing power tax. Following this logic, what will the three major domestic carriers do next?

Ordinary people will never voluntarily buy a $100 AI subscription, and everyone is already numb to all kinds of membership subscriptions. So operators will definitely resort to their most skilled and most hated old trick: bundling high-end 5G plans.

In the past few years, we paid for mobile data by the GB. But as 5G becomes a mere pipe and video data exemption becomes widespread, operators can no longer squeeze profits out of simply selling data. Look now, in order to get you to upgrade to expensive high-end plans, they have to desperately stuff in iQiyi memberships or Bilibili premium members.

Then the next step, their new selling point will inevitably become: upgrade to the 199 yuan large data plan each month, and directly receive a Gemini exclusive self-disciplined AI agent computing power package, enjoy exclusive computing power channel, and skip the queue.

The operator's calculation: data is no longer valuable, so turn AI computing power into a new premium point for plans, forcing you to maintain a high-end plan in order to keep the AI assistant in your phone from lagging.

Google's calculation: Those ordinary people who stubbornly refuse to buy the $100 original price plan are packaged in batches through carrier channels, becoming the base that steadily supplies infrastructure costs to Google.

For users, you don't even need to sign up for any third-party payment deduction. You just feel that your monthly phone bill plan, which never changes, has become more expensive, but since the AI in the background runs more smoothly, you have no choice but to hold your nose and accept it.

This kind of trick leaves you no time to think about subscription fatigue; it directly smashes the electricity and computing costs generated by AI into the fixed communication expenses you have to pay every month. In essence, it's exactly the same recipe and the same flavor as the notorious SP business practices of China Mobile and China Unicom back in the feature phone era, which included arbitrary charges and forced bundling of value-added services.

After watching this press conference, my intuition tells me: The era of purely treating AI as a toy and slacking off is completely over.

Tech giants simply don't care whether ordinary users hate subscription models. On one hand, they use storage space as bait to directly extract $100 to $200 in cash flow from paying professionals and enterprises; on the other hand, through high-end carrier packages from telecom operators and various big tech system services, they disguise computing power taxes as communication plan fees and platform handling fees to indirectly skim profits.

They don't need everyone to willingly fork over $100. As long as they use AI infrastructure to firmly grip the underlying funnel of global digital transactions and communications, every level of human economic activity will automatically pay tribute to them.

In this ruthless computing war, how are we grassroots programmers, who are stuck staring at the backend Token interfaces madly churning data, supposed to survive?

We used to complain all day about "not having enough data," but in a few years, we'll likely be staring at our phone's prompt that says "your basic computing power for this month's carrier plan has been used up," and obediently pay up to upgrade to a more expensive plan.

Alright, I'll keep staring at that $100 subscription page and go back to figuring out how to use my intuition and code to create something new.