Anthropic’s Big Software Reset: Winners & Losers

In early February 2026, the tech sector saw a sharp and sudden valuation reset, with nearly $300 billion in market capitalization wiped out in roughly a single trading day. The declines were concentrated in software, data services, and IT outsourcing stocks. The trigger was Anthropic’s release of Claude Cowork, a set of open-source plugins that enable AI agents to execute tasks autonomously, end-to-end, using raw inputs rather than operating inside existing software workflows. Demonstrations showed the system independently conducting legal research and drafting filings.

This likely changes the narrative of AI from a mere productivity enhancer into a direct substitute for entire layers of software and services.  The scale of the sell-off raises a critical question for investors: if so much value was destroyed so quickly, where exactly could it be reallocated? In the medium term, markets rarely erase hundreds of billions of dollars without simultaneously pricing in new winners. So how should investors play this shift?

The sell-off is a reminder that individual stocks can be volatile and shake you out, but strategic allocation and diversification help you stay invested. Our Boston-based wealth management partner’s asset allocation approach is designed exactly for that.

Which sectors of the market were badly hit?

• Seat-Based SaaS: Giants like Salesforce (NYSE:CRM), ServiceNow, and Adobe (NASDAQ:ADBE) saw 6% to 8% declines. These companies typically charge “per user.” If AI agents make one human as productive as ten, the total number of “seats” required by an enterprise drops precipitously.
• IT Services (Outsourcing): Indian firms like Infosys and TCS were hit hard. Their revenue model relies on billing hours for manual data tasks and junior-level coding, which are the exact tasks Claude’s new plugins could potentially automate.
• High-Margin Data Providers: Stocks like Thomson Reuters and LegalZoom fell 15% to 20% as investors priced in the risk that autonomous AI can handle legal triage, research, and drafting, reducing the willingness to pay premium prices for search-oriented legal databases. However, this does not mean the underlying data is becoming obsolete. Much of the value in these businesses sits in proprietary, licensed, and continuously updated legal, tax, and financial datasets that are legally and practically “un-scrapable” by large language models.

So is the software industry dying?

Probably not, but the business model could very well be changing. We could be moving from “Software-as-a-Service” or buying the tools to effectively buying the outcomes. In other words, value could be moving from the interface, or the buttons one clicks, to the intelligence and the task that is getting done. The “subscription per head” model could be under siege. New leaders like Palantir and Harvey are increasingly moving toward outcome-based billing.  Instead of charging $100 per month for a seat, they charge based on the successful completion of a task (for example, say, a contract reviewed or a bug fixed).  For the layman, this might mean that software is becoming “invisible.” One may no longer need to log into multiple platforms to get work done. Instead, they will state the objective, and an AI agent will navigate systems, tools, and data to deliver the outcome.

Who are the primary winners?

Semiconductor vendors like Nvidia (NASDAQ:NVDA) are clear beneficiaries because autonomous agents do not merely generate text; they reason. Reasoning requires far more compute cycles per task than simple inference. Nvidia’s Rubin platform was explicitly designed for agentic workloads, optimizing both compute density and data movement. As agents shift from episodic use to continuous execution, demand for high-performance accelerators rises structurally, not cyclically.

Model developers such as Anthropic and OpenAI sit at the intelligence layer that replaces multiple categories of software simultaneously. When a single agent can perform legal research, customer support, data analysis, and code generation directly, value consolidates at the model level rather than being distributed across dozens of vertical SaaS tools. As pricing shifts from access to software toward completed outcomes, model builders gain leverage. They are not selling interfaces or workflows. They are selling task completion. That allows them to capture economics that previously accrued to software vendors and services firms.

Cloud infrastructure providers such as Amazon and Google benefit because autonomous agents run continuously rather than intermittently like human-driven SaaS usage. Traditional SaaS is constrained by login frequency and human attention. Agents operate persistently and in parallel. This makes AI workloads significantly more compute-, storage-, and power-intensive, directly increasing cloud consumption. Companies with control over data centers, power access, and land increasingly resemble infrastructure owners. In that sense, they become the landlords of the agent economy.

Physical-world AI: The reset in software valuations could also rotate capital toward businesses that AI cannot easily replace. As pure software becomes increasingly commoditized through APIs, markets may begin to reward physical AI. Tesla has increasingly decoupled from the broader software cohort, trading as a play on robotics and autonomous mobility through Optimus and Cybercab. The logic is straightforward. An AI agent can draft a legal brief, but without a physical body, it cannot run a warehouse floor.

What should investors do?

Approach this shift with caution, recognizing that value may gradually move away from seat-based SaaS and labor-driven services toward compute, cloud infrastructure, and outcome-oriented AI platforms. Positioning should emphasize balance: selective exposure to semiconductors and hyperscalers, careful scrutiny of software pricing models, and measured interest in physical-world AI, while avoiding assumptions of rapid or uniform disruption across the sector.

The Trefis High Quality (HQ) Portfolio, with a collection of 30 stocks, has a track record of comfortably outperforming its benchmark that includes all three – the S&P 500, S&P mid-cap, and Russell 2000 indices. Why is that? As a group, HQ Portfolio stocks provided better returns with less risk versus the benchmark index; less of a roller-coaster ride, as evident in HQ Portfolio performance metrics.