# AI 推高 NSA、美国医院与保险公司的成本账单

- 来源：The Decoder：AI News（RSS）
- 作者：Maximilian Schreiner
- 发布时间：2026-09-25 20:03
- AIHOT 分数：55
- AIHOT 链接：https://aihot.news/items/cmugxuqhh1hbdrogvo9swp2g8
- 原文链接：https://the-decoder.com/intelligence-doesnt-come-cheap-as-ai-drives-up-costs-for-the-nsa-hospitals-and-insurers

## AI 摘要

NSA 告知国会今年将花费数十亿美元测试先进 AI 模型，算力是最大开支，立法者预计全面 AI 监管每年可能耗资数百亿美元，远高于此前 CBO 对类似法案约 2000 万美元一年的估算。

## 正文

Nano Banana Pro prompted by THE DECODER

The NSA is reportedly spending billions of dollars to test AI models, mostly on computing power. In US health care, AI drives up costs in a different way. It makes care itself more expensive because hospitals and insurers use it to fight over billing.

The National Security Agency told Congress it will spend billions of dollars this year testing advanced AI models, The Washington Sun reports, citing two sources familiar with the classified estimates. According to one source, computing power is the biggest expense. Staffing adds to the bill because the NSA has to compete with the pay packages AI labs offer.

Because of this, lawmakers expect that full-scale AI oversight could cost tens of billions of dollars a year, according to the sources. Earlier estimates were far lower. The Congressional Budget Office put the cost of a bipartisan bill to create a center for AI risks at about $20 million a year.

Paying by the token means paying full price for compute

AI providers' pricing models show why computing power costs large customers so much. Consumer flat-rate subscriptions are heavily subsidized. According to SemiAnalysis, a maxed-out $200 ChatGPT Pro subscription is worth up to $14,000 at API list prices, while Claude Max, which costs the same, comes to up to roughly $8,000.

Businesses, on the other hand, increasingly pay based on usage. OpenAI bills new enterprise contracts in tokens. At Anthropic, 75 to 85 percent of revenue comes from usage-based contracts, according to SemiAnalysis. Agent workflows can use up to a thousand times more tokens than a regular chat.

Providers are responding. 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. That doesn't apply to the most powerful systems. Frontier models remain expensive, and providers increasingly sell their capabilities through closed cybersecurity programs.

Business customers appear to be lucrative for providers. SemiAnalysis estimates the gross margin on Anthropic's API business at more than 80 percent. The companies need that money, though. 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.

In health care, AI billing fights push up the cost of care

US health care has a different problem. What drives costs there isn't the price of AI but how hospitals and insurers use it for billing. In a report, the Blue Cross Blue Shield Association accuses hospitals of using AI-assisted coding to run up nearly $1 billion in extra costs over two years, The New York Times reports. According to the report, 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, it pays off. The hospital system McLaren Health Care brings in an extra $1 million a 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.

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.