Walmart and Anthropic Signal a Major Shift Toward Autonomous Commerce as AI Infrastructure Spending Nears One Trillion Dollars

The global retail and technology landscapes reached a significant inflection point this week as Walmart and the artificial intelligence startup Anthropic unveiled separate but deeply interconnected advancements in autonomous software. These developments suggest a transition from human-directed digital tools toward independent AI agents capable of executing complex economic transactions. While Walmart is building the digital architecture to allow software agents to shop on behalf of human consumers, Anthropic’s recent financial disclosures reveal the massive capital requirements and improving unit economics necessary to sustain this next generation of computing.

The Rise of Agentic Commerce: Walmart’s Non-Human Shoppers

Last Wednesday, Walmart Inc. (WMT) officially began restructuring its digital infrastructure to accommodate a new class of consumer: the autonomous AI agent. The retail giant announced a suite of tools designed to allow customers to delegate shopping tasks to software assistants. Unlike traditional e-commerce interfaces designed for human eyes and manual scrolling, Walmart’s new system includes a dedicated gateway, currently in beta testing, that provides AI agents with a direct, streamlined connection to Walmart and Sam’s Club shopping systems.

This move represents a fundamental shift in the retail experience. By authorizing these agents to browse, select, and purchase goods, Walmart is positioning itself as the primary infrastructure for "agentic commerce." To maintain human oversight, the system allows users to set strict spending limits, require manual approval for specific categories of purchases, and track agent-led orders through a separate interface. This infrastructure addresses a primary hurdle in the adoption of AI: the transition from software that merely provides information to software that executes real-world actions.

Simultaneously, Anthropic introduced a research preview of Claude Duet, an AI assistant designed to operate across various connected applications. Claude Duet is engineered to continue tasks autonomously after a user has stepped away, checking in for final confirmation before completing significant actions. Together, these announcements underscore a growing industry-wide push to move AI beyond the "chatbot" phase and into a functional role where software handles recurring household and business responsibilities.

Anthropic’s Financial Disclosure: A Path to Profitability

The operational shift toward AI agents is mirrored by the financial maturation of the companies building the underlying technology. On Monday, Anthropic PBC filed documentation revealing significant growth in its business model. The company disclosed second-quarter revenue of $11.6 billion and its first GAAP operating profit of $974 million. This financial milestone is particularly noteworthy given the high costs associated with training and running large language models (LLMs).

A historical analysis of Anthropic’s financials shows a rapid improvement in efficiency. In early 2023, the company faced a significant deficit in compute-to-revenue ratios; for every dollar spent on infrastructure and computing power, the company generated only 40 cents in revenue. However, over the subsequent five quarters, the scaling of enterprise adoption began to outpace the growth of computing expenses. By the second quarter of 2024, Anthropic reported that every dollar of compute spending supported approximately $2.08 in revenue.

While the filing also showed a staggering $80.7 billion net loss for the first half of the year, market analysts note that the vast majority of this figure is attributed to non-cash accounting adjustments related to convertible financing. The emergence of a $1 billion operating profit suggests that the underlying business of selling AI intelligence is becoming fundamentally viable. Furthermore, Anthropic revealed that it is utilizing its own technology to drive internal productivity, with AI generating more than 90% of the company’s code contributions as of September.

Enterprise Adoption and Customer Retention Data

The demand for agentic AI is reflected in Anthropic’s customer data, which shows a significant expansion in high-value enterprise contracts. The number of customers spending at least $100,000 annually nearly doubled in the third quarter, reaching more than 11,900. Perhaps most telling is the company’s net revenue retention rate, which reached approximately 900%. This indicates that existing customers are not only staying with the platform but are exponentially increasing their spending as they integrate AI more deeply into their operations.

Currently, more than half of the Fortune 500 companies are spending at least $100,000 annually with Anthropic. However, industry analysts point out that this spending currently represents less than 0.1% of the total operating expenses for these corporations. This suggests a significant "runway" for growth; if AI agents become reliable enough to handle broader categories of corporate work—such as software engineering, data analysis, and customer support—the share of corporate budgets allocated to AI providers could increase by orders of magnitude.

The $850 Billion Infrastructure Build-Out

The shift toward autonomous AI agents requires a massive expansion of physical and digital infrastructure. Anthropic remains "compute-constrained," meaning its growth is limited by the availability of high-end chips and data center capacity. The company’s filing suggests that its fixed and minimum supplier commitments are approaching $850 billion, with some estimates for the broader ecosystem total nearing $1 trillion.

These commitments involve a complex web of partnerships across the semiconductor and hardware industries:

• Compute Ecosystems: Anthropic relies heavily on the Tensor Processing Unit (TPU) ecosystem developed by Google (GOOG, GOOGL) and Broadcom (AVGO).
• Semiconductor Demand: The expansion continues to fuel demand for Advanced Micro Devices (AMD) and Nvidia (NVDA) GPUs, as well as high-bandwidth memory and fabrication services from Taiwan Semiconductor Manufacturing Company (TSM).
• Infrastructure and Power: To secure its future, Anthropic is moving toward directly leasing chips and developing dedicated data centers with on-site power generation. This has created a secondary wave of demand for power management, cooling systems, and networking hardware.

Connectivity specialists like Astera Labs (ALAB) have become integral to this build-out. Their technology allows massive arrays of AI chips to work in concert, a requirement for training the next generation of models that will power Walmart’s shopping agents and Anthropic’s Claude Duet.

Supply Chain Confirmation and Market Reactions

Earnings reports from across the AI supply chain confirm that the investment in infrastructure is not merely speculative but is backed by concrete orders. Several key players have recently reported record-breaking figures:

• Cadence Design Systems (CDNS): Reported a record $8.2 billion backlog in chip design software and services.
• Corning (GLW): Saw its optical communications business grow by 44%, driven by the need for fiber-optic connectivity in AI data centers.
• Celestica (CLS): Provided aggressive forward guidance, forecasting $40 billion in revenue by 2027—a 90% increase over its 2026 outlook—citing sustained demand for finished AI systems.
• Amkor Technology (AMKR): Disclosed roughly $3 billion in customer prepayments, indicating that tech giants are paying in advance to secure advanced packaging capacity for AI chips.

These figures suggest that the "build phase" of the AI era is intensifying. The breadth of the companies reporting growth—spanning from chip design and memory interfaces to packaging and finished server racks—indicates a synchronized effort to prepare for the high-compute demands of agentic software.

Implications and Future Outlook

The convergence of Walmart’s retail innovation and Anthropic’s financial milestones marks a new chapter in the AI trade. However, the path forward is not without risks. Analysts have raised concerns regarding the disparity between the massive supplier commitments (nearing $1 trillion) and the current contracted future customer revenue. There is also the risk that a massive Anthropic initial public offering (IPO) could drain liquidity from other sectors of the technology market.

Furthermore, the relationship between AI developers and their infrastructure providers is becoming increasingly complex. As companies like Anthropic move toward dedicated power and proprietary data centers, they are effectively becoming infrastructure operators themselves. This shift is driven by the necessity of securing the technology and energy required for "sovereign AI"—the ability of a company or nation to maintain its own computing capabilities independent of third-party bottlenecks.

The move toward AI agents represents the ultimate test of the generative AI thesis. If software can successfully take over routine tasks for millions of people and businesses, the current levels of infrastructure spending may eventually be viewed as a necessary foundation for a new era of productivity. If, however, issues of trust, reliability, or cost-efficiency stall the adoption of these agents, the industry may face a period of significant consolidation.

For now, the momentum remains with the builders. Walmart’s preparation for non-human shoppers and Anthropic’s transition into operating profitability provide a clear signal: the industry is moving beyond the experimental phase and into a period where AI is expected to perform, transact, and generate economic value autonomously. Raising the capital for this transition was the first hurdle; securing the physical components—the chips, the fiber, and the power—remains the defining challenge of the coming decade.