In 2026, the technology sector is living through a troubling contradiction. On one side, the world’s biggest companies — Google, Amazon, Microsoft, Meta — are cutting tens of thousands of jobs at a pace not seen in a decade. On the other, those same companies are collectively planning $725 billion in investments in artificial intelligence infrastructure, while facing a severe shortage of qualified talent: 275,000 AI-related positions remain unfilled worldwide. This paradox reveals a deep and often misunderstood transformation of the tech labor market.

150,000 job cuts: a pace unseen since 2020

Since January 2026, the global technology sector has recorded more than 150,000 job cuts, according to data compiled by specialized aggregators such as Layoffs.fyi. This pace far exceeds the whole of 2025. The “Big Four” — Google, Amazon, Microsoft and Meta — account for most of these cuts, affecting entire teams in product, sales, human resources and operations.

The wave is not limited to the United States. In Europe, subsidiaries of major tech groups have also reduced headcount, using in their official communications vague wording around operational streamlining and prioritizing AI. Analysts have not missed this semantic shift.

275,000 unfilled AI jobs: the other side of the paradox

While large groups part with employees in roles deemed “transformed” by automation, they are simultaneously struggling to recruit profiles trained for the new demands of AI. According to several converging industry sources, 275,000 artificial-intelligence-related positions remain unfilled worldwide — because candidates with the required skills are lacking.

Machine-learning engineers, AI ethics experts, architects of agentic systems, data scientists able to operate foundation models at scale: these profiles are rare, highly sought after and unevenly distributed between countries. Universities are struggling to train people fast enough, and companies that lay off generalist developers do not automatically recover the specialists they need.

AI washing: when AI becomes an excuse for restructuring

The term AI washing — by analogy with greenwashing — refers to the practice of invoking artificial intelligence to justify job cuts that in reality stem from other causes: falling demand, a post-Covid correction after excessive hiring, shareholder pressure or simple strategic reorganization.

Data from Challenger, Gray and Christmas, a U.S. firm specialized in tracking job cuts, illustrates the phenomenon well. Of the 108,000 job cuts recorded in the United States in January 2026, only 7% explicitly cited AI as a direct cause. This figure contrasts sharply with the public statements of many CEOs, who routinely mention AI to explain restructuring.

AI often serves as a narrative shield for decisions that would have been made anyway. It is not a lie — AI is indeed transforming work — but it is not the single cause that press releases suggest either.

The debate remains open between analysts and unions. On one side, empirical data struggles to establish a direct and massive short-term link between AI adoption and job cuts. On the other, signs of deep transformation are accumulating across entire sectors: customer support, translation, content creation and software development.

In France: Mistral AI and the state hand in hand

France is not a passive spectator of this revolution. The startup Mistral AI, valued at €11.7 billion in 2026, is one of the few European players capable of competing technically with American labs. Its CEO Arthur Mensch announced that he is targeting one billion euros in revenue by the end of the year.

On the public-sector side, the rollout program for an interministerial AI assistant based on a Mistral model is entering its final phase. Launched in October 2025 with 10,000 civil servants across several ministries, this pilot is one of Europe’s most advanced sovereign-AI programs. A decision to expand it to the entire civil service is expected in the coming weeks.

Mistral AI has also launched Forge, a platform enabling companies and governments to train frontier-level models from their own data, without dependence on U.S. players. Its first customers include ASML, Ericsson and the European Space Agency.

AI Act: the regulatory countdown accelerates

On the regulatory front, the situation is more complex than it appears. The provisional agreement of May 7, 2026 on the Digital Omnibus postponed the compliance obligation for high-risk AI systems to December 2, 2027, and even to August 2, 2028 for systems embedded in industrial products.

But this postponement does not mean obligations are lifted. The text still has to be formally adopted by the European Parliament and the Council before August 2, 2026 — the date on which the initial provisions of the AI Act would have taken effect. Companies that have slowed their compliance programs are navigating an uncomfortable gray area.

• Bans already in force (exploitation of vulnerabilities, social scoring) have applied since early 2025.
• Penalties for non-compliance can reach 7% of global revenue.
• In France, the CNIL, DGCCRF and Arcom have been designated as competent national authorities and are intensifying their communications.

For companies still waiting on the sidelines, regulators’ message is clear: postponement does not mean impunity.

What this means for workers

Faced with this transformation, what conclusions should people in work or retraining draw?

First, the jobs/AI paradox shows that technology does not eliminate “work” as a whole, but shifts the skills required. Repetitive tasks, simple information-processing roles and standardized support functions are on the front line. Roles requiring specialized AI expertise, human supervision of automated systems or skills in high-value relational fields are more resilient.

Second, the gap between the 150,000 jobs cut and the 275,000 unfilled positions shows that tech companies suffer from a skills-transformation problem, not merely an excess workforce. Redeployment plans and continuous training are becoming strategically urgent, for individuals as much as for public policy.

Third, regulatory pressure from the AI Act will create growing demand for hybrid profiles: lawyers specialized in AI law, compliance managers and algorithmic ethics experts. These are jobs that barely existed five years ago.

A rapidly changing market, not a collapse

The tech sector is not collapsing. It is mutating. The 2026 wave of job cuts looks less like job destruction than a massive reallocation of human resources toward new profiles, new skills and new ways of organizing work. But this transition is creating real turbulence for hundreds of thousands of people.

France, with Mistral AI, its commitments to digital sovereignty and its European regulatory ambitions, is trying to navigate this episode with its own strategy — neither a simple vassal of American platforms nor withdrawn into itself. The success of this path will largely depend on how fast skills can be developed, both in companies and in public institutions.