IBM shares plunge 10% after Anthropic's COBOL AI announcement
IBM stock dropped over 10%, its worst day since 2000, after Anthropic announced its Claude AI can modernize legacy COBOL systems.
English · Original discussion in Spanish · Published
IBM shares plunge 10% after Anthropic's COBOL AI announcement
On February 23, 2026, IBM shares recorded their worst session since October 2000, falling more than 10% after Anthropic announced that its Claude AI can accelerate the modernization of systems written in COBOL. Although this news initially concerns a programming language from the 1960s, it has shaken the foundations of a multi-billion dollar business: legacy system consulting. 95% of ATM transactions in the US are processed with COBOL, and that is where IBM derives part of its revenue.
Why does a technical announcement trigger a stock market crash?
The explanation is simple: IBM generates significant revenue by helping companies modernize their COBOL systems, primarily through consulting and mainframe services. If an AI can do this work faster and cheaper, the business shrinks. The market quickly discounted lower future revenues and sold shares en masse. This reaction is not isolated; it fits into a broader trend where AI news hits traditional software and service companies.
What stands out is that Anthropic's tool, Claude Code, not only automates complex tasks related to COBOL—such as analyzing, mapping dependencies, and documenting systems—but also accelerates processes that previously required many human experts and months of work. Now, these take just quarters. For a consultancy, this is an existential threat.
Traditional consulting business under scrutiny
Beyond the IBM case, the announcement has highlighted an uncomfortable reality: some argue that traditional consulting is a house of cards built on artificial knowledge scarcity. The potential displacement of COBOL analysts is just the tip of the iceberg. The real problem is that the entire ecosystem of digital transformation, high-end consulting, and outsourcing of critical services relies, according to this view, on complexity and antiestéticar of the unknown.
Some claim that COBOL is nothing more than a high-level assembler for a prehistoric compiler, and that any programmer can learn it in a month. The issue is not the language itself, but the business logic embedded in millions of lines of code that only a few understand. That logic is what an AI now promises to unravel. And that changes the rules of the game.
Is it realistic for AI to replace COBOL experts?
The answer is not unanimous. A COBOL migration tool is the holy grail of corporate IT, comparable to discovering a cure for cancer. If Anthropic truly possesses such a tool, it would already be deployed across all banks and insurance companies worldwide, and Anthropic would be the most valuable company on the planet. The reality is more nuanced: the challenge with COBOL is not writing the code, but managing the associated business logic.
Furthermore, there is a significant technical obstacle: banking and insurance systems require determinism, while AI models are inherently non-deterministic. Some argue that no rational person believes one can create an alternative to COBOL using machines that sometimes give different answers to the same question. In an environment where a calculation error can cost millions, this is an unacceptable risk.
The hidden cost of AI: who pays the bill?
Another open front is economic. Some suggest that AI companies themselves are on the verge of bankruptcy because they earn two euros for every ten they spend. The cost per use is unsustainable and, according to this view, they are surviving with accounting engineering while praying to find the key to monetizing the invention. IBM's drop is partly a symptom that the market is beginning to scrutinize the viability of the entire sector.
Added to this is an issue of involuntary industrial espionage: some claim that companies are discovering that if their developers ask AI to fix code, it learns from private company code, and when a competitor's developer asks something similar, the AI offers a solution based on the first company's data. A risk many companies had not considered.
What happens to COBOL analysts and IBM?
The immediate future is uncertain. On one hand, some argue that COBOL experts are highly valued not for knowing COBOL—which they consider trivial—but because they are also experts in mainframes and business logic. If you also know serious accounting, you are in high demand. On the other hand, migration to new systems raises doubts: Who will understand the newly generated code? Another AI? Anyone, or only the one that has you at its mercy?
Meanwhile, IBM saw its market value collapse in a single session. The company has not issued an official statement, but investors have already taken positions. The lingering question is whether this announcement marks the start of a deep sector transformation or is merely a temporary scare. What is clear is that AI has ceased to be a promise and has become a tangible threat to some pillars of the digital economy.
Key Data
Frequently Asked Questions
What is COBOL and why is it still used?
COBOL is a programming language created in the 1960s. Despite its age, it continues to run critical systems such as banking, airlines, and public administrations. It is estimated that around 95% of ATM transactions in the United States are processed with COBOL. Its robustness and the enormous amount of legacy code explain its persistence.
What exactly did Anthropic announce?
Anthropic presented a new tool within its Claude AI, called Claude Code, which automates complex tasks related to COBOL, such as analyzing, mapping dependencies, and documenting systems. According to the company, it can accelerate the modernization of legacy systems based on COBOL, processes that normally required many human experts and months of work, reducing them to quarters.
Why does this affect IBM?
IBM generates significant revenue by helping companies modernize COBOL systems, primarily through consulting and mainframe services. If an AI can do this work faster and cheaper, there is less business for traditional consultancies like IBM, implying lower expected future revenues. This triggered massive share sales and a drop of more than 10% in a single day.
On February 23, 2026, IBM shares recorded their worst session since October 2000, falling more than 10% after Anthropic announced that its Claude AI can accelerate the modernization of systems written in COBOL. Although this news initially concerns a programming language from the 1960s, it has shaken the foundations of a multi-billion dollar business: legacy system consulting. 95% of ATM transactions in the US are processed with COBOL, and that is where IBM derives part of its revenue.
Why does a technical announcement trigger a stock market crash?
The explanation is simple: IBM generates significant revenue by helping companies modernize their COBOL systems, primarily through consulting and mainframe services. If an AI can do this work faster and cheaper, the business shrinks. The market quickly discounted lower future revenues and sold shares en masse. This reaction is not isolated; it fits into a broader trend where AI news hits traditional software and service companies.
What stands out is that Anthropic's tool, Claude Code, not only automates complex tasks related to COBOL—such as analyzing, mapping dependencies, and documenting systems—but also accelerates processes that previously required many human experts and months of work. Now, these take just quarters. For a consultancy, this is an existential threat.
Traditional consulting business under scrutiny
Beyond the IBM case, the announcement has highlighted an uncomfortable reality: some argue that traditional consulting is a house of cards built on artificial knowledge scarcity. The potential displacement of COBOL analysts is just the tip of the iceberg. The real problem is that the entire ecosystem of digital transformation, high-end consulting, and outsourcing of critical services relies, according to this view, on complexity and antiestéticar of the unknown.
Some claim that COBOL is nothing more than a high-level assembler for a prehistoric compiler, and that any programmer can learn it in a month. The issue is not the language itself, but the business logic embedded in millions of lines of code that only a few understand. That logic is what an AI now promises to unravel. And that changes the rules of the game.
Is it realistic for AI to replace COBOL experts?
The answer is not unanimous. A COBOL migration tool is the holy grail of corporate IT, comparable to discovering a cure for cancer. If Anthropic truly possesses such a tool, it would already be deployed across all banks and insurance companies worldwide, and Anthropic would be the most valuable company on the planet. The reality is more nuanced: the challenge with COBOL is not writing the code, but managing the associated business logic.
Furthermore, there is a significant technical obstacle: banking and insurance systems require determinism, while AI models are inherently non-deterministic. Some argue that no rational person believes one can create an alternative to COBOL using machines that sometimes give different answers to the same question. In an environment where a calculation error can cost millions, this is an unacceptable risk.
The hidden cost of AI: who pays the bill?
Another open front is economic. Some suggest that AI companies themselves are on the verge of bankruptcy because they earn two euros for every ten they spend. The cost per use is unsustainable and, according to this view, they are surviving with accounting engineering while praying to find the key to monetizing the invention. IBM's drop is partly a symptom that the market is beginning to scrutinize the viability of the entire sector.
Added to this is an issue of involuntary industrial espionage: some claim that companies are discovering that if their developers ask AI to fix code, it learns from private company code, and when a competitor's developer asks something similar, the AI offers a solution based on the first company's data. A risk many companies had not considered.
What happens to COBOL analysts and IBM?
The immediate future is uncertain. On one hand, some argue that COBOL experts are highly valued not for knowing COBOL—which they consider trivial—but because they are also experts in mainframes and business logic. If you also know serious accounting, you are in high demand. On the other hand, migration to new systems raises doubts: Who will understand the newly generated code? Another AI? Anyone, or only the one that has you at its mercy?
Meanwhile, IBM saw its market value collapse in a single session. The company has not issued an official statement, but investors have already taken positions. The lingering question is whether this announcement marks the start of a deep sector transformation or is merely a temporary scare. What is clear is that AI has ceased to be a promise and has become a tangible threat to some pillars of the digital economy.
Key Data
- IBM Drop: More than 10% in a single day, the worst since October 2000.
- 95% of ATM transactions in the US are processed with COBOL.
- Anthropic states that Claude Code can accelerate the modernization of COBOL systems from months to quarters.
- AI companies earn 2 euros for every 10 spent, according to circulating calculations.
Frequently Asked Questions
What is COBOL and why is it still used?
COBOL is a programming language created in the 1960s. Despite its age, it continues to run critical systems such as banking, airlines, and public administrations. It is estimated that around 95% of ATM transactions in the United States are processed with COBOL. Its robustness and the enormous amount of legacy code explain its persistence.
What exactly did Anthropic announce?
Anthropic presented a new tool within its Claude AI, called Claude Code, which automates complex tasks related to COBOL, such as analyzing, mapping dependencies, and documenting systems. According to the company, it can accelerate the modernization of legacy systems based on COBOL, processes that normally required many human experts and months of work, reducing them to quarters.
Why does this affect IBM?
IBM generates significant revenue by helping companies modernize COBOL systems, primarily through consulting and mainframe services. If an AI can do this work faster and cheaper, there is less business for traditional consultancies like IBM, implying lower expected future revenues. This triggered massive share sales and a drop of more than 10% in a single day.
Summary of a discussion on Burbuja.info - Foro de economía, actualidad y política., translated from Spanish and reviewed before publication. Read the full discussion (194 replies).
More summaries
-
AI as an economic engine: bubble or structural shift?
Economic analysis of AI: Is it the growth engine or a speculative bubble? Energy costs and business adoption are examined.
-
Police Officer at 48 Receives €31,800 Incapacity Pension
A 48-year-old officer with 20 years of service receives €31,800 annually due to incapacity. The debate on privileges and sustainability resurfaces.
-
Zapatero's Daughter Sparks Aesthetic Debate in Spanish Forum
A forum thread analyzes the physical changes of former PM Zapatero's daughter, focusing on elite lifestyle contradictions.
-
Sánchez misstates King’s reign length amid power debate
Pedro Sánchez incorrectly claims the King has reigned for 20 years, sparking debate on his own tenure and the monarchy's role.
-
Argentine conspiracy theories surge after World Cup final loss
Following the World Cup final, Argentine fans unleashed conspiracy theories. We analyze refereeing data, historical controversies, and this cultural phenomenon.
-
70% of young adults live with parents, masking deeper housing crisis
While 70% of young adults reside with family, fewer than 10% achieve true independence. This highlights Spain's severe housing affordability and demographic challenges.