AI costs surge for US national security and healthcare as NSA spends billions on testing

The National Security Agency is spending billions of dollars this year to test advanced AI models, with computing power accounting for the bulk of the expense, according to a report from The Washington Sun. The revelation comes as US lawmakers anticipate that full-scale AI oversight could cost tens of billions of dollars annually, far exceeding earlier estimates.

The NSA’s spending on AI testing is driven primarily by the cost of computing infrastructure, with staffing expenses also rising as the agency competes with AI labs for talent. Lawmakers now expect that comprehensive AI oversight could cost tens of billions of dollars a year, according to sources familiar with classified estimates. This contrasts sharply with earlier projections from the Congressional Budget Office, which estimated the cost of a bipartisan bill to create a center for AI risks at about $20 million per year.

The NSA’s AI spending highlights a broader trend: the cost of deploying and overseeing advanced AI systems is far higher than early estimates suggested. For Indian readers tracking global AI policy and technology costs, the NSA’s experience offers a cautionary tale about the real expenses involved in AI adoption at scale.

Why AI computing costs so much

The high cost of AI computing for the NSA is not unique to government agencies. AI providers’ pricing models reveal why computing power is so expensive for large customers. Consumer flat-rate subscriptions are heavily subsidized, while businesses increasingly pay based on usage.

According to analysis from SemiAnalysis, a maxed-out $200 ChatGPT Pro subscription is worth up to $14,000 at API list prices. Similarly, Claude Max, which costs the same, represents up to roughly $8,000 in API value. For enterprise customers, OpenAI now bills new contracts in tokens, while Anthropic generates 75 to 85 percent of its revenue from usage-based contracts.

Agent workflows, which are increasingly used by businesses, can consume up to a thousand times more tokens than a regular chat session. Providers are responding to this demand. According to The Information, OpenAI now lets some large customers pay only for completed tasks, and the company introduced GPT-6 Sol and Luna, models priced at half the usual API rate. However, frontier models remain expensive, and providers increasingly sell their capabilities through closed cybersecurity programs.

The business of AI appears lucrative for providers. SemiAnalysis estimates the gross margin on Anthropic’s API business at more than 80 percent. But the companies need this revenue. According to the Financial Times, OpenAI plans to spend about $856 billion on computing power through 2030, and Anthropic is planning its IPO for November at the latest.

AI in US healthcare: billing disputes add nearly $1 billion

In US healthcare, AI is driving up costs in a different way. According to a report from the Blue Cross Blue Shield Association, hospitals are using AI-assisted coding to run up nearly $1 billion in extra costs over two years, as reported by The New York Times. Hospitals coded cases as more severe and added secondary diagnoses, collecting nearly $12,000 more per case on average without any apparent change in treatment.

For hospitals, the investment pays off. The hospital system McLaren Health Care brings in an extra $1 million per month using software from SmarterDx, says CFO Dave Mazurkiewicz. The vendor gets a cut of the added revenue. Insurers, meanwhile, use AI to comb through records for claims they can deny, Mazurkiewicz says.

Caroline Pearson of the Peterson Health Technology Institute says both sides now go several rounds per dispute “because it’s cheap.” Writing in Health Affairs, former US health official David Brailer says he expects AI to raise overall health care spending, just as electronic health records did before it.

Datos clave

Item	Details
NSA AI testing costs	Billions per year, mostly on computing power
US healthcare AI billing costs	Nearly $1 billion extra over two years
OpenAI planned computing spend	$856 billion through 2030
Anthropic gross margin on API	More than 80 percent

Who pays for AI oversight?

For the NSA, the debate over who should foot the bill has already started. Nat Purser of the AI Verification and Evaluation Research Institute proposes a levy that would have developers help pay for independent testing. According to The Washington Sun, Anthropic and OpenAI have signaled they might be willing to pay for more government oversight.

The rising costs of AI adoption raise important questions for Indian businesses, policymakers, and technology professionals. As AI deployment scales globally, the hidden costs of computing infrastructure, regulatory oversight, and unintended consequences such as billing inflation will need to be factored into planning. For Indian startups and enterprises considering AI adoption, the US experience suggests that the total cost of AI goes far beyond subscription fees and API charges.

Source: The Decoder – https://the-decoder.com/intelligence-doesnt-come-cheap-as-ai-drives-up-costs-for-the-nsa-hospitals-and-insurers/