AI tools threaten the consulting firms that keep decades-old software running

Overview

An estimated 220 billion lines of COBOL code still run in production every day, processing 95% of ATM transactions and roughly $3 trillion in daily commerce. For decades, understanding and modernizing that code has required large teams of specialized consultants working for months or years. On February 23, Anthropic published a playbook showing how its Claude Code tool can automate the most labor-intensive phases of that work—mapping dependencies, documenting workflows, and identifying risks across thousands of files—and IBM shares immediately fell 13.2%, their worst single-day drop in more than 25 years.

The sell-off wiped over $31 billion from IBM's market value and dragged down Accenture and Cognizant, which fell 6% and 6.6% respectively. All three companies earn significant revenue from helping banks, airlines, and government agencies modernize legacy systems. The market reaction signals a broader reckoning: if AI can compress years of consulting work into quarters, what happens to the firms whose business model depends on that time scale? Some analysts say the sell-off vastly overstates the threat. Others see the beginning of a structural shift in how the world's oldest software gets maintained.

Play on this story Voices Debate Who Said What? Predict

Key Indicators

13.2%

IBM single-day drop

Largest percentage decline since October 2000, wiping over $31 billion in market value

$29.4B

Legacy modernization market

Estimated global market size in 2026, growing at 17.6% annually

220B+

Lines of COBOL in production

Powers 95% of ATM transactions and an estimated $3 trillion in daily commerce

~33%

IBM February decline

On track for IBM's worst single month since at least 1992

Interactive

Exploring all sides of a story is often best achieved with Play.

Ever wondered what historical figures would say about today's headlines?

People Involved

Dario Amodei

Leading Anthropic's expansion into enterprise tooling

Arvind Krishna

Defending IBM's mainframe and consulting franchise amid AI disruption concerns

Dan Ives

Publicly arguing the software sell-off is overdone

Organizations Involved

Anthropic

AI model developer

Released Code Modernization Playbook targeting legacy COBOL systems

Anthropic is the maker of Claude, a family of AI models, and Claude Code, an agentic coding tool whose new COBOL modernization capabilities triggered the largest single-day sell-off in IBM stock in over 25 years.

International Business Machines Corporation (IBM)

Technology and Consulting Corporation

Stock down ~33% in February 2026; defending mainframe and consulting business

IBM is the world's dominant mainframe vendor and a major provider of consulting services for modernizing the COBOL systems that run on its hardware—the exact work Anthropic's new tool targets.

Accenture

Global Consulting and Professional Services Firm

Stock declined 6% on February 23; significant legacy modernization practice at risk

Accenture is one of the world's largest consulting firms and derives meaningful revenue from legacy system modernization engagements that AI tools now threaten to compress or automate.

Cognizant Technology Solutions

IT Services and Consulting Corporation

Stock declined 6.6% on February 23; exposed to legacy modernization disruption

Cognizant is a major IT services provider with deep roots in maintaining and modernizing legacy COBOL systems for financial institutions, making it directly exposed to AI-driven disruption of that work.

Timeline

Anthropic publishes COBOL modernization playbook; IBM crashes 13.2%
Product Launch / Market Event

Anthropic released a Code Modernization Playbook demonstrating how Claude Code can automate the exploration and analysis of COBOL codebases. IBM shares plunged 13.2% to $223.35, their worst single-day percentage drop since October 2000, wiping over $31 billion in market value. Accenture fell 6% and Cognizant fell 6.6%.
February 23rd, 2026
Wedbush analyst calls software sell-off 'most disconnected trade'
Analysis

Dan Ives of Wedbush Securities publicly pushed back against the software sector sell-off, calling it the most disconnected trade he had seen in his career on Wall Street.
February 14th, 2026
IBM shares fall 6.5% as AI disruption fears mount
Market Event

IBM became the primary laggard on the Dow Jones Industrial Average as investors began repricing software and services companies that could be disrupted by AI automation.
February 12th, 2026
IBM reports record Q4 mainframe revenue
Earnings

IBM reported fourth-quarter 2025 revenue of $19.7 billion, beating estimates. Mainframe revenue grew 61% year-over-year, the highest Q4 figure in more than two decades, driven by the new z17 platform.
January 28th, 2026
IBM CEO touts mainframe AI adoption
Statement

Arvind Krishna said IBM's watsonx Code Assistant for Z had achieved wide adoption, with most customers using it to understand their COBOL code bases and plan modernization.
July 1st, 2025
DeepSeek shock rattles AI and tech stocks
Market Event

Chinese AI lab DeepSeek released a model rivaling Western frontier systems at a fraction of the cost, triggering a broad sell-off in AI-adjacent stocks and introducing the template of an 'AI scare trade' that would recur throughout 2025 and into 2026.
January 27th, 2025
IBM launches AI tools for mainframe modernization
Product Launch

IBM unveiled watsonx Code Assistant for Z, its own AI tool for understanding and converting COBOL code to Java on mainframes. The tool marked IBM's first major move to bring generative AI into its legacy modernization business.
August 22nd, 2023

Scenarios

Predict which scenario wins. Contrarian picks score more — points lock in when the scenario resolves.

Predictions leaderboard

AI compresses consulting timelines but firms adapt, preserving most revenue

AI tools like Claude Code and IBM's watsonx automate the initial analysis phases of COBOL modernization but cannot handle the full migration—rewriting business logic, testing against production systems, managing regulatory compliance, and coordinating organizational change. Consulting firms absorb the tools into their own workflows, using them to accelerate delivery rather than losing the work entirely. Margins compress modestly, but total addressable market grows as modernization becomes affordable for mid-size organizations that previously couldn't justify the cost. IBM's stock recovers as investors recognize the threat was overstated.

Discussed by: Wedbush Securities analysts, Jefferies, IBM management

Consensus —

AI triggers structural decline in legacy consulting, reshaping the services industry

AI tools prove capable of not just analysis but increasingly competent code migration, eating into the most profitable phases of modernization engagements. Large consulting firms see utilization rates drop as projects that once required 50-person teams for 18 months shrink to 10-person teams for six months. Indian IT services firms—Tata Consultancy Services, Infosys, Cognizant—face disproportionate impact because of their heavy reliance on labor-intensive modernization work. The $29 billion legacy modernization market doesn't disappear, but the labor component shrinks permanently, forcing services firms to reinvent their business models around AI-augmented delivery.

Discussed by: Bloomberg, technology industry analysts, McKinsey internal assessments

Consensus —

COBOL modernization proves too complex for AI, and the sell-off fully reverses

Production COBOL systems turn out to be far more tangled than AI can reliably untangle. Decades of patches, undocumented business rules, interdependencies with other legacy systems, and regulatory constraints make AI-generated analysis unreliable for mission-critical environments. Banks and government agencies, unwilling to risk systems that process trillions of dollars daily, continue relying on human consultants who understand institutional context. Claude Code becomes a useful assistant for consultants but not a replacement, and the market recalibrates IBM and services stocks back toward pre-panic valuations.

Discussed by: Mainframe industry veterans, IBM consulting leadership, enterprise CIOs

Consensus —

AI modernization accelerates mainframe migration, undermining IBM's hardware franchise

The deeper threat isn't that AI replaces consulting work—it's that AI makes it faster and cheaper to move off mainframes entirely. If organizations can now understand their COBOL codebases quickly enough to plan full migrations to cloud platforms, IBM loses not just consulting revenue but the recurring infrastructure revenue from mainframe licensing and maintenance—a segment that just recorded its best quarter in 20 years. This scenario would represent a genuine existential challenge to IBM's business model, not just a margin compression for consultants.

Discussed by: Financial analysts focused on IBM's infrastructure segment, cloud computing advocates

Consensus —

Historical Context

IBM's near-death experience and the mainframe's first obituary (1993)

1991-1993

What Happened

In the early 1990s, the rise of client-server computing led analysts to declare the mainframe dead. IBM posted an $8 billion loss in 1993—then the largest annual loss in American corporate history—and brought in Lou Gerstner as CEO. The company's stock fell roughly 75% from its 1987 peak, and layoffs cut headcount from 405,000 to 220,000.

Outcome

Short Term

Gerstner scrapped a plan to break IBM into pieces and instead bet on services and software, pivoting the company away from hardware dependence.

Long Term

Mainframes didn't die. They evolved into high-throughput transaction processors that became even more deeply embedded in banking and government. IBM rebuilt itself around them, and three decades later mainframe revenue hit a 20-year high.

Why It's Relevant Today

The current panic follows the same pattern: a new technology is declared capable of replacing mainframes and their surrounding ecosystem. History suggests the mainframe may adapt rather than disappear—but also that IBM's stock can fall dramatically before the market recognizes that.

Y2K and the rise of Indian IT outsourcing (1998-2000)

1998-2000

What Happened

The Year 2000 bug required companies worldwide to audit and fix billions of lines of COBOL code before date fields rolled over. American firms couldn't find enough COBOL programmers domestically—the language had already fallen out of university curricula—so they turned to Indian IT companies like Infosys and Tata Consultancy Services, where COBOL was still taught. Indian IT services revenue surged.

Outcome

Short Term

The Y2K remediation wave generated enormous revenue for consulting firms and established Indian IT companies as credible enterprise partners.

Long Term

After Y2K passed, the Indian IT firms pivoted to ongoing legacy maintenance and modernization, building the multi-billion-dollar outsourcing industry that persists today. The same COBOL expertise that powered Y2K fixes became the foundation for decades of modernization consulting.

Why It's Relevant Today

The legacy modernization consulting industry was born from a COBOL skills shortage. AI tools now threaten to fill that same skills gap mechanically, potentially disrupting the business model that Indian IT firms and IBM Consulting built on top of Y2K-era relationships.

Kodak and the digital photography disruption (1975-2012)

1975-2012

What Happened

Eastman Kodak invented the digital camera in 1975 but chose to suppress it to protect its film and processing business, which generated enormous margins. By the time Kodak attempted a digital pivot in the early 2000s, Sony and Canon had seized the market. Kodak's market capitalization fell from $31 billion at its 1997 peak to bankruptcy in 2012.

Outcome

Short Term

Kodak tried several strategies to compete in digital, including online photo sharing and digital camera lines, but couldn't replace the profit margins that film had provided.

Long Term

Kodak became the canonical example of an incumbent destroyed by a technology it invented but couldn't adopt without cannibalizing its core business.

Why It's Relevant Today

IBM finds itself in a structurally similar position: it launched its own AI modernization tool (watsonx Code Assistant for Z) in 2023, but fully automating COBOL modernization could undermine the consulting revenue that depends on the process being slow and labor-intensive. The question is whether IBM can deploy AI to modernize its own business model faster than outside competitors.