Salesforce rebounds on AI: Agentforce starts booking revenue and Anthropic's IPO boosts the value of its bet

🕒 Published on Zendoric: September 7, 2026 · 02:54
✨ AI-generated · how it's made
Salesforce stock is up more than 80% from its year lows, to $260, after results that belie — for now — the feared "SaaSpocalypse". The other engine: its stake in Anthropic, which is reportedly preparing an IPO valued at up to $2 trillion.
By Zendoric · September 7, 2026.
Salesforce (NYSE: CRM) stock has risen from the $144 it touched as its low for the year to the current $260, a gain of more than 80% in just a few months. The trigger was second-quarter results: cRPO (the value of contracts already signed that the company has yet to bill, the indicator that best anticipates future revenue in a subscription business) grew 14%, to $33.5 billion, and quarterly revenue rose 11%, to $11.3 billion. Informatica, the data management company Salesforce bought last year, contributed $440 million of that figure, according to data published by Salesforce itself. Management also raised its annual guidance to a range of $46.1-46.4 billion, 12% more than the previous year.
What matters from the AI standpoint is not just that Salesforce is billing more, but where that growth comes from. Agentforce (its AI agent platform for automating customer service, sales and support tasks) and Data 360 (the layer that unifies a company's internal data so those agents can operate on it) already generated close to $3.9 billion in annual recurring revenue, and the company says it has delivered 7 billion «agentic work units» -its own metric for counting tasks executed autonomously- through products such as Agentforce and Slack. It is the figure the company sets against the «SaaSpocalypse», the thesis circulating in the industry that software companies will end up being displaced by artificial intelligence tools.
The second driver of the rise has nothing to do with Salesforce's own business: its bet on Anthropic. The company took a stake in Anthropic in the 2023 Series C round and has taken part in every subsequent round; its holding is now estimated at $5 billion. As Benzinga's own article reports, Anthropic is said to be preparing an IPO at a valuation of up to $2 trillion. It bears emphasizing: this is a reported plan, not a confirmed prospectus price, and figures of this kind tend to move a great deal before debut day. If it were confirmed even approximately, Salesforce's stake would be worth, in the source's own words, «substantially more» -without specifying how much-. Salesforce is not the only party involved: Amazon, Microsoft and Nvidia are also among the large corporate shareholders that would benefit from an eventual Anthropic IPO at that valuation. Several analyst houses have already raised their price target on CRM: Morgan Stanley lifts it to $315, and Needham, BTIG, Cantor Fitzgerald and Deutsche Bank have moved theirs in the same direction.
As sector context, this pattern -big tech companies financing AI labs in exchange for equity, compute credits or preferential access to the models, and then seeing that stake surface as an argument for their own share price- has become the central mechanism of the AI race. We have already seen it with Nvidia taking a stake in labs such as Safe Superintelligence in exchange for multiplying its access to compute; here the same mechanism appears on the corporate customer side, with Salesforce, Amazon, Microsoft and Nvidia all exposed to the same asset.
Our reading is that the Salesforce case is a data point in favor -though not definitive proof- that the fear of generative AI wiping out established software vendors is materializing less than feared, at least for the incumbent that manages to get agents into its own customer and data base: the Agentforce figure is the most concrete evidence so far that the disruption is being absorbed rather than replacing the long-standing vendor. That does not mean employment comes out unscathed. The very «agentic work units» Salesforce charges its customers for effectively replace administrative and support tasks that people used to do -the same erosion we described when analyzing the business administration sector, where the profiles that survive are those that orchestrate AI, not those that did the repetitive work-. Nor should the article's own technical warnings be lost sight of: the stock has run into resistance at $267 and is tracing a candlestick pattern («shooting star») that usually precedes a pullback, a sign that much of the optimism about Anthropic is already priced in before an IPO prospectus even exists. In the long run, if Agentforce and equivalent platforms really do free up staff from repetitive tasks so they can focus on customer relationships and expert judgment, the episode fits the underlying thesis: AI does not simply destroy work, it changes which work matters. But that is proven with employment data sustained over time, not with a few months' stock rally.
🔗 Related on Zendoric


