Nvidia Corp. (NASDAQ: NVDA), Palantir Technologies Inc. (NASDAQ: PLTR), and Booz Allen Hamilton Holding Corp. (NYSE: BAH) are restricting their use of advanced AI models from Anthropic and OpenAI due to concerns about exposing sensitive corporate information.

The companies are demanding stronger guarantees or dropping specific models to prevent providers from retaining or learning from proprietary customer data, according to a report by The Information on Monday.

Nvidia shares fell about 3% on Monday amid a broader selloff in AI-linked stocks.

Alternative Models for Sensitive Data

Ohalo CEO David Friedberg described similar concerns during a podcast appearance on Friday. He stated that Ohalo is "moving off Claude for a lot of our sensitive data" and shifting work toward locally run models.

Friedberg cited anecdotal cases where novel scientific ideas discussed with one model appeared to resurface when querying another version from a different account. His concern extends beyond leaking confidential documents to potentially giving AI providers valuable internal know-how.

Nvidia and Palantir Build an Alternative

Nvidia is applying a similar approach to its supply chain operations. Last week, Nvidia and Palantir unveiled a system using Palantir software and Nvidia’s open Nemotron models to analyze Nvidia’s own supply-chain data.

Nvidia is the first customer for this technology. The company says clients can customize models using proprietary operational data while retaining control over the data, models, and infrastructure location.

This approach allows Nvidia to capture how its planners make decisions without handing supply-chain knowledge to an outside model provider.

Anthropic Responds to Customer Pressure

Anthropic introduced 30-day data retention with Fable 5 in June to detect misuse across multiple sessions and accounts. The company states this information is not used to train its models.

Anthropic later acknowledged that many enterprises, particularly in regulated industries, found the retention requirement difficult to accept. Its proposed solution, Enterprise Frontier Safeguards, will keep activity data used for misuse monitoring in customer-controlled cloud infrastructure. It begins rolling out in phases later this fall.

Despite these data concerns, expectations for Anthropic’s IPO remain high. Polymarket puts the odds of a listing this year at 86%.

Nvidia is also reportedly considering a roughly $10 billion anchor investment in an offering that could value Anthropic near $2 trillion.

The debate echoes Palantir CEO Alex Karp’s recent warning that companies risk handing AI providers their proprietary "alpha" along with their data.

Nvidia Corp (NASDAQ: NVDA) generated $177.8 billion in revenue during the first half of fiscal 2027. The chipmaker’s Data Center segment drove this growth, with quarterly revenue jumping 117% year-over-year to $89 billion.

Customer concentration intensified in the period. Three direct customers accounted for 44% of total revenue, or $78.2 billion, up from two customers contributing 35% a year earlier. This shift highlights the company’s reliance on a small group of tech giants for its top-line expansion.

Nvidia Wants to Get ‘Paid Twice’

To mitigate concentration risk and broaden its revenue base, Nvidia is advancing its neocloud strategy. Under this model, the company sells data center infrastructure to AI cloud providers while securing cloud-service agreements to finance capacity. In return, Nvidia captures a share of rental revenue when third parties utilize that compute power.

CFO Colette Kress described the approach on the earnings call: "In this model, we get paid twice—once on the hardware sale and again through the share of rental revenue." This structure allows Nvidia to move beyond one-time hardware sales to participate in the recurring economics of AI demand.

The Customer Base Could Get Much Bigger

Nvidia projects that non-hyperscaler businesses—including sovereign AI initiatives, regional neoclouds, enterprises, edge computing, and air-gapped data centers—will eventually comprise roughly half of its Data Center business. The neocloud agreements are designed to unlock this segment.

As of July 26, Nvidia’s commitments under the new model totaled $36 billion. These agreements typically span six years and decline as third-party customers or Nvidia itself consume capacity. This creates a feedback loop where GPU sales expand the customer base, driving utilization and generating secondary revenue streams.

What the Numbers Show

The divergence between overall revenue growth and customer concentration warrants attention. While Data Center revenue surged 117%, the share of revenue from just three customers rose from 35% to 44%. This indicates that the bulk of recent growth is concentrated among a shrinking number of buyers, increasing dependency risk despite the absolute scale of sales.

For investors, the critical metric is whether Nvidia can convert its hardware dominance into a recurring claim on the broader AI economy. The success of neocloud utilization and non-hyperscaler demand expansion will determine if customer concentration remains a vulnerability or becomes a lever for sustained leverage.