Fifteen years ago, 'phone' meant a device for making calls. Today, a phone is the camera, the bank, the map, the boarding pass, the home remote, and lastly, rarely, a device for making calls. The same shift is starting with the company computer. Nvidia and Microsoft jointly announced the first reinvention of the PC in 40 years, and the change goes well beyond 'it got faster'. This post explains what is changing, why it matters for a company of any size, and what makes sense to think about right now for the 2027 budget.
The new name is RTX Spark. It is a computer that still runs everything it ran before (Windows, Office, the company's system, all existing programs), but it has a new capability inside it: running an AI agent around the clock, without needing the internet, without a cloud subscription, without sending data outside. That sentence changes a lot. It is worth understanding why.
1. What this computer does that the current one does not
The current computer turns on, you open a program, use it, close it, shut down. It is a tool that waits for orders. It does only what you tell it, when you tell it.
The new computer is different. It has inside it an agent of its own, running 24 hours, listening, observing, executing tasks in the background. You ask 'organize my files from the week by client', you go grab a coffee, you come back and it is done. You ask 'prepare me a summary of yesterday's three meetings with the three decisions still pending', you leave for a meeting, you come back and it is done. This agent is a digital employee that lives inside your machine.
Think of an assistant who never sleeps, never asks for a raise, never takes vacation, and is on every computer in the company. This is not an exaggeration. It is literally the product being sold. The real question is: what will you ask it to do?
2. Why running locally, and not in the cloud, changes everything for a company
Almost all AI today runs 'in the cloud'. You send the question, it goes to the AI company's server, it processes, it comes back. It works, but it has three problems that any serious company needs to weigh:
- It costs per use: the more you use, the more you pay. Some months come out expensive.
- It sends data outside: every question you ask involving client data, contract data, payroll data, leaves your company. In some businesses that is prohibited by law. In others, it is just uncomfortable.
- It depends on the internet: the link goes down, everything stops. And in some regions of Brazil, the link goes down every day.
The new computer solves all three. The agent runs inside the machine, without a subscription, without sending data, without needing the internet. For a law firm, an accounting firm, a healthcare office, for any business that handles sensitive client information, this difference is not a detail. It is what separates 'usable' from 'not usable at all'.
3. The example that proves it: a complete house project in one session
Nvidia showed an architect working on the new PC during the presentation. He described in text what he wanted (a house, on a given plot, with a given style, a given budget). The agent, running inside the machine, opened the design program, modeled the terrain, proposed three ways of building, chose one with him, generated the interior layout (walls, doors, windows, structure), exported it to another rendering program, chose the angles, generated the realistic photos of the finished house.
All of that in one session. All of that inside the machine, without sending the project to any cloud. The architect stepped in at the points that mattered (choosing the concept, adjusting the material, validating the layout). The rest, the agent did. What used to be a week of work came out in one afternoon.
The case is architecture, but the pattern applies in any office where there is a chained creative or analytical process: a design agency, an engineering firm, a marketing department, a sales proposal team. The same computer, with the same kind of agent, executing the grunt work and leaving the human on the decisions.
4. It is not just a laptop: there is a whole new line
Nvidia announced three types of machine, all with the same agent:
- Laptop (RTX Spark): for the team, mobility, day to day.
- Desktop: stays on the desk, on all the time, ideal for the agent that needs to keep working while the person is away. It can be the office's 'AI hub', controlling cameras, systems, company integrations.
- Workstation (DGX Station): for a company that wants an agent far smarter than the standard one, trained on the business's own data. It costs more. It delivers more.
All three run 100% of Windows, 100% of the programs the company already uses. It is not a change of system. It is a change of machine, with everything continuing to work. Simple migration. The cost of changing is low. The cost of not changing is leaving the team working slowly while the competitor speeds up.
5. What to think about now for next year's budget
It is not a matter of replacing every machine tomorrow. It is a matter of thinking about the natural renewal cycle. Every company replaces its computers every three, four, five years. The next round of replacement is an important decision: do you keep buying a machine designed for a user who types, or do you start buying a machine designed for a user who delegates to an agent?
For those replacing in 2027 or 2028, the decision is made today. You cannot wait for the catalog to fill with options and the price to drop, because meanwhile the competitor is already using it. For those who replaced recently, it is worth starting with the machines in the roles where the gain is greatest: creation, design, analysis, client service that involves reasoning over information.
The important reframe: the PC of the next five years will not look like the current one. It will look more like a 'team member' than a 'tool on the desk'. This is the kind of change that looks like an exaggeration when it starts and looks obvious five years later. The smartphone also looked like an exaggeration in 2008. Today nobody wants to be without one. The difference is that, this time, the company that buys first gains a real commercial advantage, not just comfort.