Cohere’s enterprise AI strategy now includes a quieter channel into Palantir’s customer base. According to information discussed in a Palantir video, Cohere models are already being used by various unnamed Palantir customers.
The partnership matters because it connects one of the best-known AI startups outside of OpenAI and Anthropic with a software company known for serving major enterprises as well as U.S. defense and intelligence agencies. The details available so far are technical, but they point to a broader shift: large language models are increasingly being delivered inside platforms that already sit close to sensitive business and operational data.
What Cohere Is Bringing To Palantir
Cohere is based in Toronto and San Francisco and sells AI to enterprise customers. Unlike companies that became widely recognized through consumer chatbots, Cohere has not built its public identity around a viral consumer product.
The company is also closely tied to the modern large language model wave. It was co-founded by an author of the Attention Is All You Need paper, which helped launch the LLM revolution. As of July, Cohere had reached a $5.5 billion valuation.
That enterprise focus makes the Palantir connection especially important. Palantir customers can access Cohere’s latest AI models through compute modules inside Foundry, according to remarks made by Cohere engineer and former Palantir employee Billy Trend.
In a November 2024 presentation at DevCon1, Palantir’s first developer conference, Trend said Cohere was already deploying to Palantir customers. He also said the presentation would explain how Cohere could serve those customers through Palantir.
How The Deployment Works Inside Foundry
The most concrete information in the Palantir video concerns how customers reach Cohere’s models. Trend said Palantir customers can use Cohere’s latest models through compute modules within Foundry.
That point is significant because Foundry is one of Palantir’s flagship platforms. Palantir has described Foundry as geared more toward commercial customers, while Gotham, its older main platform, was designed for defense and intelligence agencies.
The distinction does not identify the organizations using Cohere through Palantir. But it does suggest that some deployments could involve corporations rather than government or intelligence users. Palantir works with major enterprises, including Airbus, while also speaking openly about its close work with U.S. defense and intelligence agencies.
For customers, the appeal is clear from the information available: Cohere’s models can be made available where Palantir customers already work with data and operational systems. That reduces the partnership to a practical enterprise AI question: how can advanced models be deployed inside existing software environments while respecting the rules that customers place around their data?
Why Data Constraints Matter
Trend did not name specific Palantir customers in the video. He did, however, describe one deployment involving a Palantir customer with really strict constraints on where it can store its data.
That customer also wanted to do inference in Arabic. Trend described that as a strong fit for Cohere because, in his words, it is an area where Cohere excels.
Those details are limited, but they explain why enterprise AI deployments often depend on more than model quality alone. A customer may need language capability, but it may also need strict control over where data lives. In that setting, the delivery mechanism can be as important as the model itself.
The source does not identify the customer, the industry, or the exact use case. It also does not say whether the deployment is commercial, defense-related, intelligence-related, or something else. The useful takeaway is narrower: Cohere’s models are already being delivered through Palantir to customers with specific operational and data-location requirements.
The Defense Question Remains Open
The Palantir connection naturally raises questions because Palantir serves both large enterprises and U.S. defense and intelligence agencies. Palantir has also published a manifesto about how to rebuild the defense-tech sector.
TechCrunch asked Cohere whether its AI is being used for military or intelligence-related use cases, and what Cohere’s general policy is toward those deployments. Cohere declined to comment. Palantir did not immediately comment.
That leaves a clear information gap. We know Cohere is working with Palantir. We know Cohere models are already in use by unnamed Palantir customers. We know those models can be accessed through Foundry compute modules. We do not know which customers are using them, or whether any of those deployments involve defense or intelligence work.
The broader market context is also moving in the same direction. Anthropic made headlines for a deal with Palantir and AWS to sell AI to defense customers. OpenAI is also being used by defense tech, with news that it had penned a deal with Anduril.
Why The Quiet Partnership Stands Out
Cohere has publicized partnerships with major technology companies such as Fujitsu. But according to a review of its website and announcements cited in the source article, the company has stayed quiet about any deals with Palantir.
That contrast is notable. A partnership with Palantir may be technically useful and commercially important, but it also sits in a more sensitive public context because of Palantir’s government, defense, and intelligence work.
For now, the available facts point to a partnership that is already operational but only partially visible. Cohere’s models are being used by Palantir customers, Foundry is the access point, and at least one deployment involves strict data-storage limits and Arabic inference.
The unanswered questions are just as important as the confirmed details. Until Cohere or Palantir provides more information, the public picture remains incomplete: a prominent enterprise AI company is already embedded in Palantir customer deployments, but the customer list and use cases remain undisclosed.