Salesforce CEO Marc Benioff Says Company Could Spend $300 Million on Anthropic AI Tokens in 2026
- Utshab Biswas
- 3 minutes ago
- 6 min read
What You need to Know
Salesforce CEO Marc Benioff revealed the company could spend nearly $300 million on Anthropic AI tokens in 2026, mainly for AI-powered coding tools.
Salesforce says AI is already handling between 30% and 50% of the company’s overall workload, leading to a freeze in software engineering hiring since early 2025.
Benioff confirmed Salesforce engineers are now working alongside AI tools like Anthropic Claude, OpenAI Codex, and Cursor instead of manually writing every line of code.
The AI race within enterprise software has moved beyond experimentation, piloting, or impressive demos. The battle for massive investments is now underway, and Marc Benioff has just dropped one of the largest spending figures we've seen yet. On the All-In Podcast, the Salesforce CEO made the disclosure that Salesforce could be spending up to $300 million on Anthropic tokens in 2026, with the majority of this spending going towards AI-enabled coding.
Also Read: GTA 6 Pre-Orders Reportedly Go Live Early in Europe, Pricing Leak Sparks Massive Fan Frenzy
This is indicative of the pace at which AI coding assistants are revolutionizing software development within tech firms. Furthermore, it demonstrates the willingness of business leaders to invest vast sums of money in AI systems that can increase efficiency, save operational costs, and speed up product delivery.
The comments by Benioff come in a time when AI-enabled software development platforms are taking center stage in the development of new products. From coding to debugging applications and streamlining engineering workloads, AI coding assistants have come to play a crucial role in the development process.
Salesforce’s Massive AI Commitment Could Reshape Enterprise Software
According to Benioff, Salesforce is preparing for an unprecedented scale of AI usage internally. Anthropic’s AI models, especially its Claude family of systems, are expected to play a major role in helping Salesforce engineers write, optimize, and manage software.
Benioff said during the podcast:
“These coding agents are awesome. Anthropic is awesome. I am going to probably use $300 million of Anthropic tokens this year at Salesforce. Coding. Everything's going to be cheaper to make.”
The announcement is important since AI tokens are literally the fuel that powers any AI generators. From interactions, responses, lines of codes to AI workflows, anything that is done by AI consumes tokens, which are then charged for based on consumption.
Should the amount of expenditure materialize, Salesforce will most likely emerge as one of Anthropic’s largest corporate clients. While neither company has confirmed the exact amount via any regulatory announcements, the amount alone underscores the importance that Salesforce attaches to the coding AI infrastructure.
The announcement also highlights the rising clout of Anthropic in the enterprise AI sector. While their competitors have been working on developing AI capabilities in many consumer-oriented AI products such as video production, AI companions and even marketing AI, Benioff believes that Anthropic’s intense focus on coding is its main strength.
Marc Benioff Says AI Is Changing Software Development Forever
Benioff described the current AI shift as unlike anything he has witnessed in his long career in technology. According to him, AI coding agents are enabling developers and executives to move faster than ever before.
He explained:
“I can do things that I just could not do before. I can go faster than ever before. I can implement my software and sell it at the same time. I've never been able to do that before.”
These statements highlight a larger trend within the industry in which AI systems are not seen only as an assistant but rather as a collaborative partner who is able to perform even complex processes without much human assistance.
Furthermore, Benioff appreciated the strategy adopted by Anthropic, stating that “it was like a rocket ship that would not stop.” He added that he sees Anthropic’s focus on coding agents and reasoning systems bearing fruits, especially compared to other firms that are trying their hand at several products.
Salesforce Froze Engineering Hiring After AI Productivity Gains
One of the most interesting things about Salesforce's AI strategy is how much it has already impacted hiring strategies. Early in 2025, Salesforce decided to freeze software engineering recruitment because of the productivity gains brought about by AI tools such as the Agentforce platform.
Benioff says that AI was already responsible for between 30% and 50% of all work done by the company as early as mid-2025. This number is very impressive considering the size of Salesforce's operations.
Salesforce is convinced that AI coding agents help its engineers do their work more efficiently, thus making it unnecessary to resort to aggressive hiring even though there is constant product expansion.
Nevertheless, Benioff assures us that engineers will remain an essential part of Salesforce's operations. It is just that the approximately 15,000 engineers at the company are now collaborating with AI tools instead of writing lines of code themselves.
They supervise AI workflows that are built using Anthropic models, OpenAI Codex, and Cursor. This represents a bigger shift that is taking place throughout the tech industry. Engineers are increasingly becoming AI workflow managers, reviewers, and system architects instead of simple coders.
Slack Could Soon Get Powerful AI Coding Features
Benioff also revealed that Salesforce is developing new AI coding capabilities directly inside Slack, the workplace collaboration platform Salesforce acquired in 2021.
According to him, Salesforce wants to make coding accessible to a wider range of users through AI-powered tools integrated within Slack’s ecosystem.
He stated:
“We're even working on technology inside Slack to make it easier for everybody to code.”
He also hinted at the possibility of more announcements in the coming period; however, he did not go into specifics regarding any new product launches.
This is a very significant strategic step for Salesforce as the use of artificial intelligence code writing tools within corporate communication channels can help employees create automated processes, workflows, integrations, and applications without any programming skills.
In case of success, Slack would turn from a mere team communication tool into a full-fledged AI-enabled development environment.
Salesforce Wants Smarter AI Cost Management
In spite of Salesforce’s large budget for AI investment, Benioff admitted that continuous execution of frontier AI models can be extremely costly. He stressed the need for intelligent layers in infrastructure that could allocate tasks to the most cost-efficient models. He explained that not all the tasks require the most sophisticated reasoning systems.
As Benioff put it, organizations should deploy inexpensive AI models for basic tasks, leaving the more complex tasks of coding and reasoning to the Anthropic Claude models. The so-called “intermediary layer” strategy could become especially important when AI adoption is scaled to the point of being used by thousands of workers in millions of interactions each day. The strategy follows the industry trend aimed at optimizing costs for AI inference.
Agentforce Growth Shows Salesforce’s AI Push Is Paying Off
It looks like Salesforce’s aggressive approach towards AI is already bearing fruit for the company. For example, according to reports, its AI agent platform called Agentforce has already generated annual recurring revenue of $800 million.
This means that there has been an enormous 169% year-over-year growth with Salesforce closing some 29,000 deals associated with the platform. In addition, earlier this year, the company invested more than $300 million into Anthropic and now reportedly holds a 1% stake in the AI startup.
At the same time, even though Salesforce halted engineering hiring, it continues to grow elsewhere. For instance, the company plans to hire up to 2,000 salespeople who will be responsible for helping enterprises embrace AI technologies and solutions. All of this shows that, in many ways, AI is not cutting back hiring but rather reshaping it by focusing it on specific needs.
In this respect, Benioff’s statements reveal a fundamental shift in the software industry. AI coding assistants are no longer experimental productivity instruments confined to small developer communities; instead, they have started to become part of the infrastructure of some of the world's largest enterprise technology companies. The potential $300 million annual expenditure on Anthropic tokens reveals just how valuable AI-assisted coding has become for large enterprises striving to boost efficiency and speed up software creation.
The general meaning behind this development is that software engineering may continue to develop rapidly within the next few years as developers concentrate more on overseeing AI systems, improving their output, creating architecture and solving high-level problems, and letting the AI agents do the routine programming.
On the other hand, enterprises like Salesforce are sure that AI integration will help reduce development expenses, make production cycles shorter, and scale operations across the entire enterprise.
It still remains unclear whether Salesforce’s massive AI investments will bring the company any long-term competitive advantage, yet one thing becomes increasingly obvious: enterprise AI expenditures are about to enter a new era.