Lowering "AI costs" is an inevitable trend! Meta is working on "model routing," replicating OpenRouter

```

Under the pressure of continuously expanding AI infrastructure spending, Meta is trying to find cost reduction solutions internally.

On July 21, according to tech media The Information, Meta’s internal AI incubator AAI Labs is developing an AI model routing tool called "Switchboard." Its core logic is highly similar to OpenRouter's Auto Router product—by assessing task difficulty, it diverts simple requests to cheaper smaller models, thereby reducing overall inference costs.

Switchboard is still in its early stages, and it is not certain whether it will ultimately be implemented. But according to internal documents obtained by The Information, the Meta team has outlined two potential paths: one is internal deployment to reduce costs; the other is public release to external organizations running large-scale AI programming agents.

Analysis points out that this means the tool is not only a cost-cutting measure but may also become Meta’s attempt to open a new source of revenue in the AI tools market.

It is also reported that this project directly addresses a current pain point for the company: paying top-tier model prices for every programming request, even for simple tasks that could be easily handled by small models.

Targeting the Pain Point: Inference Cost is the Biggest Obstacle for Scaled Deployment

Reportedly, the logic for initiating the Switchboard project is articulated quite plainly in internal files: "We pay top-tier model prices for every programming request, including the simple ones."

The documents further state that most programming agent tasks can be handled entirely by smaller models, and only a few tasks truly require state-of-the-art large models. Yet currently, "all requests are sent to the same model, which leads to excessive expenditure on simple tasks or insufficient performance on complex tasks."

The documents explicitly describe inference cost as the "primary obstacle" to broader deployment of agents within the company, stating bluntly: "Cost is the key factor limiting our ability to run agents at scale."

This statement aligns with a series of Meta’s recent actions to rein in AI spending. According to a previous report by The Information, Meta notified employees this June that, only weeks after encouraging broader adoption of AI tools company-wide, it would begin to set limits on AI token usage, while building internal platforms to track AI spending and enforce token budgets.

Notably, the Switchboard project belongs to Meta’s AAI Labs, an internal incubator under the Meta Applied AI Engineering team, officially established in March this year. According to internal documents reviewed by The Information, the mechanism allows employees to submit AI product and service proposals, and once approved, small teams can build them with the possibility of public release.

As of this July, AAI Labs has approved about 200 projects, covering three main areas: consumer products, developer tools, and internal infrastructure. Switchboard is one of them.

This mechanism reflects Meta CEO Mark Zuckerberg's broader strategic vision—using AI to enable small teams to quickly build products. In April, Zuckerberg told analysts that AI agents mean "small teams can make very rapid progress" and predicted this technology would drive "a lot of innovation." He also said Meta might build as many as 50 new applications.

Model Routing Track: Tech Giants Are Entering the Arena Beyond OpenRouter

The model routing track targeted by Switchboard is attracting more and more attention.

OpenRouter has gained considerable popularity among developers by helping them access a variety of AI models at lower costs. According to a report by The Information last week, OpenRouter has entered talks with a much larger tech company about a potential acquisition, which could boost its valuation to several billion dollars—up from $1.3 billion in April this year.

The broader interest in the model routing concept stems from the routing feature built into OpenAI’s GPT-5—this feature can automatically switch to a cheaper model when user prompts are relatively simple. Later, companies like Databricks and Palantir also developed their own routing tools to manage costs and enhance efficiency.

Meta’s self-developed Switchboard is both a proactive response to its own cost pressures and a strategic move to build its own capabilities in this rapidly heating sector.

Greater Ambition: AI Investment Seeks Multiple Monetization Paths

Behind the Switchboard project is Meta’s overall pursuit to turn massive AI investments into new tools, new businesses, and new sources of income.

Meta previously projected that spending on AI infrastructure and other hardware and facilities could reach as much as $145 billion this year, more than double its 2025 level. At the same time, Meta is restructuring its engineering teams to strengthen AI development capabilities.

The projects incubated by AAI Labs go beyond Switchboard. According to another internal document obtained by The Information, AAI Labs is also developing an AI navigation app for drivers, which can run via Apple CarPlay and Android Auto, using AI to explain nearby landmarks and allow drivers to ask questions.

The document positions this product as an extension of Instagram’s map experience, which may in the future integrate location-based Reels content, travel recommendations, and even Meta’s Ray-Ban smart glasses.

Reportedly, these projects together sketch out Meta’s path: starting from employee ideas, rapidly prototyping AI products, and seizing opportunities to launch them to the external market—controlling costs while exploring incremental revenue beyond advertising.

Risk Disclaimer and Legal StatementThe market has risks; investment must be cautious. This article does not constitute personal investment advice and does not take into account individual users’ specific investment objectives, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article fit their particular circumstances. All investments made based on this are at one’s own risk. ```