Skip to content
Marsanne Studio
← The AI watch

Analysis ·

Apple wants Siri AI to be everywhere. At the same time, it caps its employees' token spending at $200

Original title, in Spanish: “Apple quiere que Siri IA esté en todas partes. Al mismo tiempo, limita a 200 dólares el gasto de tokens para sus empleados”

Publication
Xataka (Spain)
Author
Rubén Andrés
Published on
5 November 2026 · read in the page structured data

Summary

Apple has quietly cut the daily AI token budget allocated to some of its engineers from $300 to $200, and is considering lowering it further, according to Bloomberg. Teams outside engineering would see their limit fall from $100 to $25 per day. This move, described as the end of "tokenmaxxing", also affects Meta, Amazon and Microsoft, where one employee spent $28,000 in a month according to 404 Media. Nvidia, by contrast, encourages token spending, through the voice of Jensen Huang. To reduce costs, Revolut has switched 2,000 engineers to GLM, an open-source Chinese model hosted in-house, which costs ten times less than frontier models, according to its head of infrastructure David Abitbol.

Summary based on the full text of the article.

Read the article on Xataka →

Figures cited

  • 200 USD/day

    Daily token limit for some Apple engineers, reduced from $300 to $200

    Some Apple engineers, since the month before the article

    « el mes pasado Apple redujo de forma discreta el límite diario de uso de tokens de algunos ingenieros de 300 a 200 dólares »

    Source: Bloomberg

  • 100 USD/day

    New limit under consideration for some Apple teams

    Some Apple teams

    « Ahora, la compañía de Ternus estudia rebajarlo todavía más y dejarlo en 100 dólares para determinados equipos. »

    Source: Bloomberg

  • 25 USD/day

    Reduction in the token limit for Apple employees outside engineering

    Groups of Apple employees outside engineering, previously at $100/day

    « Fuera de ingeniería, algunos grupos de empleados pasarían de 100 dólares en tokens a 25 dólares diarios. »

    Source: Bloomberg

  • 45,765 USD million

    Apple's R&D spending in the last financial year

    Apple, last fiscal year

    « Según sus cuentas, la compañía gastó 45.765 millones en I+D en su último ejercicio, un 32% más que en el anterior. »
  • 32 %

    Increase in Apple's R&D spending compared with the previous financial year

    Apple, last fiscal year vs previous year

    « Según sus cuentas, la compañía gastó 45.765 millones en I+D en su último ejercicio, un 32% más que en el anterior. »
  • 28,000 USD

    Amount spent on tokens by a Microsoft employee in one month

    One Microsoft employee, in one month

    « En Microsoft, un solo empleado llegó a quemar 28.000 dólares en tokens en solo un mes. »

    Source: 404 Media

  • 250,000 USD

    Token spending threshold deemed insufficient by Nvidia for a well-paid engineer

    Nvidia engineer on a $500,000 salary, according to Jensen Huang

    « Le preocuparía mucho que un ingeniero con 500.000 dólares de sueldo gastara menos de 250.000 dólares en tokens de IA. »

    Source: Podcast All-In

  • 2,000 USD million/year

    Nvidia's estimated annual token spending for its engineers

    Nvidia, engineers

    « Cuando le preguntaron si Nvidia gasta unos 2.000 millones al año en tokens para sus ingenieros, el CEO respondió: "Lo estamos intentando". »

    Source: Podcast All-In

  • 2,000 engineers

    Number of Revolut engineers using the GLM model by default

    Revolut

    « Revolut ha puesto a unos 2.000 ingenieros a programar por defecto con GLM, un modelo chino de pesos abiertos. »
  • 10 % (one tenth of the cost)

    Cost of the GLM model for Revolut compared with frontier models

    Revolut

    « El responsable de infraestructura de Revolut, David Abitbol, contó que pagan una décima parte de lo que cuestan los modelos de frontera. »
  • 300 USD/month

    Previous monthly budget per Revolut engineer for Claude Code

    Revolut, before the switch to GLM

    « Antes, los 300 dólares mensuales por ingeniero para Claude Code no duraban todo el mes, ahora el límite de consultas la definen sus propios servidores. »

Classification

Country
United States, United Kingdom
Company size
Large companies
Industries
Software vendors, IT services companies, telecoms · Banking, insurance, asset management, finance departments
Companies
Revolut (deploys AI), Amazon (cited), Meta (cited), Anthropic (Claude) (supplier), OpenAI (ChatGPT) (supplier), Google (Gemini) (supplier), Microsoft (cited), Apple (deploys AI), Nvidia (cited)
Keywords
tokenmaxxing, generative ai costs, token budget, open-source models, coding agents, enterprise ai

Entry written by Marsanne Studio from the cited article, with the help of an AI model (claude-sonnet-5). The rights to the article and its illustrations belong to Xataka and its author. Entry translated from French by Claude Sonnet 5.5.