OpenAI signs agreement with AWS to sell artificial intelligence to U.S. government personnel

OpenAI is accelerating its layout in the U.S. government market, with its strategic path becoming increasingly clear: using government contracts as a springboard to replicate Palantir's successful model, ultimately penetrating the high-value enterprise market.

On March 17, according to media reports citing two informed sources, OpenAI signed a new contract with Amazon Web Services (AWS) last Friday to provide artificial intelligence services to U.S. government personnel through AWS, covering both classified and unclassified operations. AWS is already a major cloud service provider for several U.S. government agencies and has agreed to sell OpenAI products to other government clients.

This collaboration with AWS is a key outcome of OpenAI's efforts to secure government contracts over the past few months, contingent upon adjustments to the agreement terms with Microsoft.

Previously, OpenAI's AI had to be exclusively hosted on Microsoft's platform. After completing its for-profit restructuring last fall, OpenAI signed a new agreement with Microsoft, allowing it to collaborate with competitive cloud service providers like AWS to sell AI services to national security clients such as the Pentagon.

However, there are still differences in applicability: for non-classified operations involving civil agencies like the Department of Health and Human Services, if OpenAI wishes to host AI models on AWS, it still needs to apply for a waiver from Microsoft; whereas for agencies involved in classified operations like the State Department, it can directly collaborate with AWS without Microsoft's approval.

Government contracts leverage corporate endorsement, direct revenue is limited

For OpenAI, the strategic value of government contracts goes beyond direct revenue; it lies in using them as a springboard to enter the enterprise market. High-profile government projects are often seen by large enterprises as a reliable endorsement of technology suppliers.

The software company Palantir has validated this path: as a long-term military contractor, its revenue from private sector clients last year was approximately $2 billion. OpenAI is clearly attempting to emulate this model.

However, the direct value of government contracts themselves is quite limited. Reports citing an informed source indicate that the agreement signed by OpenAI at the end of last month to provide ChatGPT and customized AI products to 3 million Department of Defense employees is expected to generate only a few million dollars in revenue over 15 months, which is negligible compared to the company's projected total revenue of $30 billion this year.

Securing defense contracts but losing user trust, OpenAI sets urgent limits

OpenAI's government market layout has not been smooth sailing. The rushed signing of the contract with the Pentagon at the end of last month triggered a serious user trust crisis.

The contract was signed on February 27, authorizing the Pentagon to use OpenAI models for classified military operations, coinciding with hours after competitor Anthropic's negotiations with Defense Secretary Pete Hegseth broke down. According to Sensor Tower data, on February 28, the uninstall rate of ChatGPT on U.S. mobile devices surged by 295% compared to the previous day, and one-star ratings skyrocketed by 775%. Internal employees also expressed dissatisfaction, with graffiti protesting "against mass surveillance" appearing outside the San Francisco office.

In the face of internal and external pressure, CEO Sam Altman publicly acknowledged on March 2 that the signing "seemed opportunistic and hasty." He announced additional contract terms, explicitly prohibiting the use of AI systems for surveilling U.S. citizens and temporarily excluding intelligence agencies such as the National Security Agency from the contract scope. **