Executives at artificial intelligence startup Anthropic have warned that a move by the U.S. government to place the company on a national security blacklist could slash billions of dollars from its projected 2026 revenue and damage its reputation with customers and investors.

The warnings emerged in federal court filings tied to a lawsuit the company filed Monday seeking to block the U.S. Department of Defense from placing it on the blacklist. The legal challenge marks a dramatic escalation in a dispute between the AI firm and the Pentagon over restrictions governing how Anthropic’s technology can be used in military-related work.

Senior executives described the potential consequences in sworn statements submitted to the court, arguing that the designation could destabilize the company’s commercial and government-facing businesses.

Krishna Rao, the company’s chief financial officer, said the impact could reach into the billions.

“Across Anthropic’s entire business, and adjusting for how likely any given customer is to take a maximal reading, the government’s actions could reduce Anthropic’s 2026 revenue by multiple billions of dollars,” Rao said in the filing.

Rao warned that if the government’s actions are allowed to stand, the consequences for the company would be “almost impossible to reverse.”

According to the filing, Anthropic estimates that hundreds of millions of dollars in revenue tied specifically to work with the Defense Department could be at risk in 2026 alone. The company also fears broader ripple effects across the defense technology ecosystem.

Rao said Anthropic could lose between 50% and 100% of revenue linked to defense contractors and other companies whose operations depend heavily on Pentagon relationships.

Beyond direct contract losses, the company argues that the designation could undermine investor confidence at a time when AI firms are spending heavily on computing infrastructure and research.

Rao said the move would likely increase the cost of raising capital needed to fund operations and growth.

Anthropic’s Head of Public Sector, Thiyagu Ramasamy, described what he called immediate damage to the company’s standing in the government technology market.

“The government’s actions immediately and irreparably harm Anthropic,” Ramasamy said in the court filing.

“The designation also impugns Anthropic’s integrity and reputation as a trusted partner, having a real but incalculable effect on sales to non-governmental customers.”

According to Ramasamy, the company expects an immediate loss of more than $150 million in annual recurring revenue tied to current and anticipated Defense Department contracts.

The company’s public sector business had been expanding rapidly before the dispute. Ramasamy said that between December 2025 and January 2026, Anthropic recorded a fourfold increase in its annual recurring revenue run rate from government clients.

Internal projections suggested that over the next five years, the public sector business could generate revenue in the multiple billions of dollars.

Those projections now appear uncertain.

If defense contractors distance themselves from the company to avoid potential compliance risks, Ramasamy warned that Anthropic’s expected public sector annual recurring revenue — projected to exceed $500 million in 2026 — could “shrink substantially or disappear altogether.”

The company’s commercial relationships are also showing signs of strain.

Paul Smith, Anthropic’s chief commercial officer, said the controversy has already disrupted negotiations and prompted some clients to reconsider their partnerships.

In one case, a partner with a multi-million-dollar annual contract switched from Anthropic’s Claude AI system to a competing generative AI model for a deployment at the U.S. Food and Drug Administration. Smith said the move eliminated an anticipated revenue pipeline worth more than $100 million.

Other deals have also been affected.

Smith said negotiations with financial institutions representing roughly $180 million in potential contracts have been disrupted, while a $15 million agreement was paused.

One fintech customer also cut the value of an existing contract in half — from $10 million to $5 million — citing concerns over the company’s dispute with the Pentagon.

According to Smith, enterprise customers are increasingly anxious about the implications of the government’s actions.

Anthropic has received inquiries from more than 100 corporate clients expressing what he described as “deep fear, confusion and doubt” about the risks of continuing to work with the company.

The case highlights the growing tension between rapidly expanding AI firms and government efforts to regulate how advanced technology can be used in national security contexts. Companies developing powerful AI models are increasingly pursuing contracts with government agencies and defense contractors, seeing the public sector as a major growth market.

But that relationship also exposes them to heightened scrutiny from regulators and national security officials concerned about how the technology is deployed.

Goldman Sachs (GS.N) is actively marketing a derivative product to hedge funds that enables them to take long or short positions on corporate loans, with a particular focus on debt issued by software companies, according to a source familiar with the matter who spoke to the Financial Times.

The instrument — a total return swap (TRS) — allows investors to gain synthetic exposure to the performance of underlying corporate loans without owning them directly. Under a TRS, one party (typically the hedge fund) receives the total economic return of the reference asset (interest payments plus any price appreciation or depreciation), while paying a financing cost (usually based on SOFR or LIBOR plus a spread) to Goldman. The structure can be used to express bullish views (long the loans) or bearish views (short the loans) on credit quality and pricing.

No trades have been executed using this specific strategy to date, the source said.

Goldman Sachs declined to comment on the specifics of the offering but provided a general statement.

“As a market-maker, we obviously engage constantly with clients on facilitating the trading strategies they want to execute. This happens every day, across many asset classes, in every market environment,” the Wall Street giant said.

The pitch comes at a moment of acute stress in the software sector. Software-as-a-service (SaaS) and legacy enterprise-software companies have seen their stock prices and credit spreads widen significantly in 2026 as investors grow increasingly concerned that rapid advances in generative AI and agentic systems could automate or commoditize many core software functions. AI agents capable of performing complex multi-step tasks across applications — from email management and data analysis to code generation and customer support — are viewed as existential threats to traditional SaaS growth models.

The primary leveraged-loan and high-yield bond markets for software issuers have been effectively frozen. No major debt deals backed by software companies have priced in the primary market since Oracle’s $25 billion debt package closed on February 2, 2026. Secondary-market liquidity has thinned, with loan prices for many software borrowers trading at deep discounts to par amid fears of slower growth, margin compression, and potential rating downgrades.

Goldman’s TRS offering would allow sophisticated investors to express directional views on software-loan performance without needing to source physical paper in a constrained market. A short position via TRS would profit if loan prices fall further (due to credit deterioration or forced selling), while a long position would benefit from any stabilization or recovery in valuations. The product could also serve as a hedging tool for banks and other loan holders looking to mitigate downside risk.

The software sector’s credit and equity weakness is seen as part of a broader re-rating of growth stocks in the AI era. Multiples across enterprise software have compressed sharply since late 2025, with many names trading at single-digit forward EV/Revenue or EV/EBITDA levels — a stark contrast to the 15–30x multiples seen during the 2020–2022 boom.

Investors worry that AI agents could erode subscription-based recurring revenue by offering low-cost or free alternatives to legacy SaaS tools. At the same time, leveraged-loan and high-yield spreads for software issuers have widened considerably. Secondary loan prices for many B-rated software borrowers are now trading in the mid-80s to low-90s, underlining both macro caution (higher rates, slower growth) and sector-specific AI disruption risk.

Goldman’s move to facilitate TRS trading in this space underscores the Street’s view that the software loan market — once seen as a stable, high-conviction asset class — is now a fertile ground for relative-value and directional trades. The bank’s role as a leading arranger and market-maker in leveraged loans gives it unique visibility into pricing dislocations and hedging demand.

While no trades have yet been executed, Goldman’s active pitching suggests growing hedge-fund interest in expressing bearish or hedging views on software credit. If AI disruption fears intensify, or if primary issuance remains sidelined, secondary-market volatility could increase, creating more opportunities for TRS-based strategies.

For software borrowers, the lack of primary-market access raises refinancing risks, particularly for companies with near-term maturities or covenant pressure. But for investors, Goldman’s product offers a way to monetize or hedge those risks without needing to source illiquid physical loans.

The development highlights how quickly market dynamics are shifting in the AI era. What was once considered a defensive, cash-flow-rich sector (enterprise software) is now viewed by some as structurally vulnerable. Goldman’s facilitation of TRS trading on software loans is an early sign that Wall Street is adapting to this new reality — turning credit dislocation into a tradable opportunity while the primary market remains closed.

Apple is now manufacturing roughly a quarter of its iPhones in India, reaching a milestone that analysts at JPMorgan projected in 2022 as the company gradually reduces its reliance on China as its primary manufacturing hub.

According to a report by Bloomberg, about 55 million iPhones produced last year came from India, out of an estimated 220 million to 230 million units manufactured globally. The figure indicates that Apple has quickly scaled production in the country, turning India into one of the most important pillars of its global supply chain.

The expansion is part of a broader strategy by the company to spread manufacturing across multiple countries, limiting exposure to geopolitical tensions, supply disruptions, and shifting trade policies that have complicated operations in China in recent years.

Apple has accelerated its manufacturing shift over the past year. The company produced the entire iPhone 17 lineup in India ahead of its global launch in September, marking a significant development in Apple’s production strategy. Previously, the earliest batches of new iPhone models were almost exclusively assembled in China, with other countries joining the production cycle later.

Speaking about the company’s evolving supply chain strategy, Apple chief executive Tim Cook said the majority of iPhones sold in the United States are now supplied from India-based manufacturing facilities. That shift indicates that India has rapidly moved from a secondary assembly location to a central manufacturing base.

Industry analysts say the transformation of Apple’s supply chain has been years in the making. The COVID-19 pandemic first exposed the risks of heavy dependence on Chinese manufacturing when lockdowns disrupted operations at major iPhone assembly plants. Since then, Apple has steadily expanded production partnerships in India with key contract manufacturers.

The pace of diversification increased in 2025 as uncertainty grew around U.S. tariff policies affecting Chinese imports. The shifting trade environment has prompted companies with global supply chains to reassess production locations and build redundancy into manufacturing networks.

The issue has also drawn political attention. At a business summit in Doha in May, U.S. President Donald Trump reportedly warned Cook against accelerating Apple’s manufacturing expansion in India, highlighting how the company’s supply chain strategy intersects with wider economic and trade priorities.

On the other hand, Apple’s expansion represents a major win in India’s effort to position itself as a global electronics manufacturing hub. The country has introduced a range of incentives under its production-linked incentive (PLI) scheme to attract large-scale technology manufacturing investments from companies seeking alternatives to China.

Apple’s contract manufacturing partners — including large electronics assemblers — have significantly expanded factory capacity in India in recent years, allowing production of newer iPhone models to begin earlier in the product cycle. The shift has also coincided with rapid growth in Apple’s presence in the Indian consumer market, which the company increasingly views as one of its most promising long-term growth opportunities.

According to data from analyst firm Counterpoint Research, Apple shipped about 14 million iPhones in India last year, representing a 9% increase from the previous year. That growth reflects the gradual expansion of Apple’s customer base in a country long dominated by lower-cost Android devices.

In revenue terms, the market has become increasingly significant. Bloomberg reported that iPhone sales in India exceeded $9 billion last year, highlighting how rising incomes and expanding premium smartphone demand are reshaping the market.

Apple has also been steadily increasing its retail footprint in the country. The company opened its sixth Apple Store in India last month, adding to the flagship outlets launched earlier in major cities as part of its push to strengthen direct engagement with customers.

The company is also exploring ways to deepen its ecosystem presence in the country. Apple is reportedly in discussions to launch its digital payments service, Apple Pay, in India later this year, a move that would expand its services portfolio in one of the world’s fastest-growing digital payments markets.

Analysts say the parallel expansion of manufacturing and consumer sales in India gives Apple a strategic advantage. Producing devices locally not only helps the company diversify supply chains but also reduces import costs and improves its ability to compete in a price-sensitive market.

At the same time, the growing scale of production in India illustrates how the global electronics supply chain is slowly evolving away from a single-country manufacturing model toward a more distributed network spanning multiple regions.

That shift could prove essential for Apple as geopolitical tensions, trade disputes, and regulatory pressures continue to reshape the global technology industry.

Nvidia is developing an open-source platform for artificial intelligence agents called “NemoClaw,” positioning itself to capitalize on the explosive growth of agentic AI tools, according to a Wired report, citing anonymous sources familiar with the matter.

The company has begun pitching the platform to major enterprise software providers, including Salesforce, Cisco, Google, Adobe, and CrowdStrike, in hopes of securing partnerships that would integrate NemoClaw into their ecosystems. Neither Nvidia nor the named companies immediately responded to requests for comment on the report.

Sources indicated that no formal partnerships have been finalized, though early access to the platform is expected to be offered in exchange for contributions to the open-source project. Because NemoClaw will be fully open source, partners would gain free usage rights, with Nvidia aiming to build a broad, collaborative ecosystem around agent development and deployment.

The platform is designed to enable companies to deploy AI agents that perform complex, multi-step tasks for employees — such as automating workflows across CRM, security, collaboration, and productivity tools — while incorporating built-in security and privacy controls. A key feature is hardware-agnostic access: companies can use NemoClaw regardless of whether their infrastructure runs on Nvidia GPUs, lowering barriers to adoption and positioning the platform as a neutral, foundational layer for enterprise agentic AI.

Nvidia’s Push into Agentic AI

Nvidia’s move denotes a deliberate shift from its core strength in AI training and inference hardware toward software and ecosystem layers that drive end-user adoption. The company has released several foundational models optimized for agentic use cases in recent months, including:

• Nemotron — a family of open models focused on reasoning, planning, and tool use.
• Cosmos — a multimodal agent framework designed for real-world task execution.

These models complement Nvidia’s existing NeMo platform, which provides end-to-end tools for building, customizing, deploying, monitoring, and optimizing AI agents — from data curation and fine-tuning to safety alignment and performance evaluation.

The NemoClaw name appears to be a deliberate nod to the viral success of OpenClaw (originally Clawdbot, then Moltbot), the open-source AI agent project that burst into mainstream awareness in late January 2026. Created by an Australian developer who was quickly acquired by OpenAI, OpenClaw demonstrated the power of lightweight, locally runnable agents that perform sequential tasks via natural-language interfaces on messaging apps (WhatsApp, Telegram, Discord).

Nvidia CEO Jensen Huang called OpenClaw “the most important software release probably ever,” underscoring the company’s view that agentic AI represents the next major wave beyond large language models.

Security and Enterprise Readiness

While OpenClaw’s consumer success has been dramatic, experts have repeatedly flagged security risks — especially for enterprise use cases. Agents with broad access to email, calendars, documents, codebases, and internal tools can introduce vulnerabilities ranging from data leakage and prompt injection to unintended actions with real-world consequences.

Nvidia’s NemoClaw is expected to address these concerns head-on with enterprise-grade features:

• Fine-grained access controls and permission boundaries.
• Audit logs and behavioral monitoring.
• Built-in safeguards against prompt injection and jailbreaking.
• Compliance tooling aligned with SOC 2, ISO 27001, and GDPR requirements.

By offering these controls natively, Nvidia aims to make agentic AI palatable to risk-averse Fortune 500 companies and regulated industries — sectors that have so far been cautious about adopting consumer-grade agent tools.

Developer Conference & The Market

Wired report arrives just one week before Nvidia’s annual GTC developer conference (March 17–20, 2026) in San Jose, where CEO Jensen Huang is expected to deliver a major keynote. GTC has historically been the venue for Nvidia’s most significant hardware and software announcements, including new GPU architectures (Blackwell in 2024, Rubin expected in 2026) and ecosystem expansions.

Industry watchers anticipate that NemoClaw — or at least a preview of its capabilities — could feature prominently, alongside updates to NeMo, Nemotron, Cosmos, and Nvidia’s broader AI software stack (NIM microservices, Blueprints, etc.). The conference is also likely to include announcements on partnerships with enterprise software vendors, potentially confirming some of the companies named in the Wired report.

Nvidia’s agentic push comes amid fierce competition:

• OpenAI has aggressively expanded ChatGPT’s agent capabilities (e.g., GPT-4o with tool use, custom GPTs, and enterprise connectors).
• Anthropic’s Claude has surged in enterprise adoption after refusing unrestricted Pentagon use, emphasizing safety and constitutional AI principles.
• Google has integrated Gemini deeply into Workspace (with Gemini for Gmail, Docs, Sheets, Meet) and is rolling out agentic features across Google Cloud.
• Microsoft has embedded Copilot agents throughout its 365 suite and Azure ecosystem.

Nvidia’s hardware-agnostic approach with NemoClaw — combined with its dominant position in AI training/inference chips — gives it unique leverage: it can court developers and enterprises regardless of which model they prefer, while ensuring that Nvidia GPUs remain the preferred compute backend.

The announcement (if confirmed at GTC) could reinforce Nvidia’s narrative as the “picks and shovels” provider of the AI era — not just selling GPUs but enabling the entire agentic software stack. Shares have been volatile in recent months amid concerns over AI spending sustainability, U.S.-China tensions, and emerging competition from Chinese AI chipmakers.

A successful agent platform launch could bolster investor confidence in Nvidia’s software moat and long-term ecosystem dominance. For OpenAI, Anthropic, Google, and others, Nvidia’s move is both an opportunity (easier integration with enterprise data) and a competitive threat (a neutral platform could reduce dependence on any single model provider).