The next two years could become extremely turbulent as artificial intelligence displaces jobs and humanoid robots enter the market. Some experts predict that up to 30% of jobs across different sectors could disappear worldwide, creating serious challenges for economies and forcing governments to prepare for a difficult transition.

In the labor market, AI is expected to take over many managerial and cognitive functions. Professions built around direct work with people, however, have a stronger chance of lasting longer, and perhaps surviving.

Another critical issue is security. AI significantly reduces the cost of warfare through drone swarms and cheap lethal weapons.

All of this raises an uncomfortable question: can any of it be controlled? We are already seeing AI being captured by those who know how to turn technology into power, profit, control, and weapons.

Work That Can Be Turned Into Data

For most users, AI still appears to be just a chatbot, a window on a screen where they can upload a question, a document, an email, an image, or even their own anxiety. But interaction with artificial intelligence has long moved beyond a single interface. It is now embedded in search results, recommendations, translation, subtitles, document processing, banking services, workplace tools, advertising, media, and surveillance systems. The more familiar this infrastructure becomes, the less often we ask the central question: who actually owns the system that is beginning to think alongside us?

This question sits at the center of Mo Gawdat's warning. A former Google executive and an insider of tech culture, Gawdat is not talking about a future robot uprising. His point is colder and more practical: the real story of AI is not unfolding in our chats, but inside laboratories, corporations, military programs, government institutions, and data centers, where AI is no longer just software, but is turning into a new infrastructure of power.

For years we were told that AI would eliminate white-collar and knowledge work. In 2026, we can see what is actually happening. The U.S. tech industry has already lost more than 123,000 jobs, with AI being the most frequently cited reason. The total number of publicly reported AI-driven layoffs has approached 150,000.

Amazon has cut more than 30,000 corporate positions between October 2025 and January 2026. Officially, the company cites the need to reduce bureaucracy and invest in AI, while at the same time accelerating the automation of its warehouses.

Meta has reduced its headcount by 8-10%, more than 8,000 people, and frozen thousands of open roles, redirecting resources toward AI infrastructure. Some employees have been reassigned to AI projects, while others have been let go.

In May 2026, Cisco laid off 4,000 employees, openly stating that it was refocusing on AI-driven networking and security.

Salesforce has cut thousands of positions in support and related functions. CEO Marc Benioff had previously spoken about needing "fewer heads," noting that AI agents are already handling up to 50% of customer interactions.

Together, the major tech companies, Amazon, Microsoft, Google, and Meta, are spending nearly $700 billion on AI infrastructure in 2026. To justify these massive investments, they are optimizing "human capital."

There is a somewhat more hopeful side: some of those laid off are retraining and moving into AI-related roles. There is also demand for specialists who understand data, ethics, integration, and oversight. However, this transition is rarely smooth.

According to Forrester, 55% of companies that reduced headcount in the name of AI are already regretting the decision, often because service quality declined. In many cases, they quietly rehire, but at lower cost, frequently through offshoring to India and similar markets, or on contract terms. Klarna remains the classic example: after claiming that its AI assistant could do the work of 700 support agents, the company ran into quality problems and had to bring some humans back.

Sovereignty Is Moving Into Data Centers

Both governments and corporations now treat AI as a strategic asset that everything else depends on.

Models are both the new oil and the new weapon. Those who control the most powerful ones set the rules for everyone else.

In 2026, compute power and data centers are real power, not flags or parliaments. Amazon, Microsoft, Google, and Meta are collectively investing $630-700 billion in AI infrastructure this year. Amazon is spending around $200 billion, Google $175-185 billion, Meta up to $145 billion, and Microsoft more than $110-150 billion. In the first quarter of 2026 alone, these four companies spent over $130 billion. The money is going into hardware, energy, and chips, because without them there is no speed, no scale, and no advantage. Data centers already devour gigawatts, and companies are securing power capacity around the world.

Military applications are turning AI into a factor that makes wars cheaper and faster. In Ukraine, both sides are already deploying drones with autonomous targeting capabilities. Ukrainian systems such as Hornet and SmartPilot enable drones to independently identify and strike targets in the final stage of flight, even under heavy electronic jamming. AI-powered "mother" drones can now autonomously locate and attack targets from distances of up to 300 km. AI processes satellite imagery, drone footage, and ground data, generating targeting information in seconds rather than hours. Russia has also begun serial production of drones equipped with machine vision and swarm capabilities. All of this reduces the cost of warfare: instead of expensive missiles or aircraft, relatively cheap FPV drone swarms with AI can be used.

Control over information, through recommendations, content generation, moderation, and search, determines what society sees and does not see. In May 2026, Google released a new Core Update and, at Google I/O, announced the most significant redesign of search in more than 25 years. This included a new AI Mode based on Gemini 3.5 Flash, a completely reimagined intelligent search interface, and early versions of informational agents that monitor the web for users. These changes strengthen the zero-click effect, as people increasingly receive complete answers directly in search results without visiting original sources. At the same time, Google has intensified its efforts against scaled AI-generated content and spam. As a result, the company is tightening its control over the information landscape society sees, and over which content gets visibility at all.

Finally, there is the growing dependence of states on foreign infrastructure. Europe, along with many countries in Asia and Latin America, even those loudly promoting "sovereign AI," continues to rely on American models, chips, and cloud services. U.S. hyperscalers, AWS, Azure, and Google Cloud, control 70-72% of the European cloud market. Europe accounts for less than 5% of global frontier AI compute. Dependence on the United States and Taiwan for advanced chips is close to 100%. In June 2026, the European Union launched its "Technological Sovereignty Package", Chips Act 2.0 and the Cloud and AI Development Act, in a state of strategic panic, recognizing that without decisive action, strategic data, computing resources, and even government contracts would remain under foreign control. At the same time, European regulation, including the AI Act, remains heavy, bureaucratic, and slow, often hindering European players more than it restrains American or Chinese competitors.

What Remains for People

AI will not remove jobs in an orderly, layer-by-layer fashion. The first wave is already hitting work that can be converted into data: call centers, assistants, junior analysts, translators, designers, marketers, office managers, and paralegals. Not everyone will be fired at once. In many cases, where four people used to be needed, only one person working with AI will remain, or a single manager overseeing a system instead of leading a team.

It doesn't just eliminate jobs. It removes the entry point into many professions. Junior positions have always functioned as a ladder, a place where people learned, made mistakes, and accumulated experience. If this bottom rung is replaced by AI, professions won't disappear overnight, but they will stop reproducing themselves in a healthy way. What remains is a thin layer of senior specialists at the top and a void below.

This story, however, has another side. When AI can function as a technical director, assistant, accountant, analyst, designer, editor, and project manager within a single interface, small businesses get a second chance. One person can now do what previously required an entire team. Small studios, private practices, local services, independent media, consulting, education, caregiving, repair work, and craftsmanship can once again become viable parts of the economy. Not because AI is kind, but because it lowers the barrier to entry.

Large companies are using AI to cut headcount. Smaller players can use it to enter the market in the first place. The crucial question is who will control the infrastructure: the individual who uses it to strengthen themselves, or the platform that turns them into a low-cost contractor.

This is where the most important shift takes place. Once AI handles information processing and routine tasks, human connection stops being a weakness and becomes an economic asset. People will pay not only for results. They will pay for trust, taste, presence, care, and the ability to navigate situations without relying on templates.

AI can write a song. But people will still go listen to live jazz in a bar. AI can generate a training program. But people will still want a trainer who can read their body, fatigue, and fear. AI can explain a diagnosis. But a patient will still need a doctor or nurse beside them so they don't feel abandoned.

The more machines surround us, the more valuable everything becomes that proves there is a real human being in front of you, not a system, not a script, not an interface.

Gawdat makes one final, important observation. He argues that the question of whether intelligence must be biological is essentially settled. It does not have to be. However, if AI eventually reaches true superintelligence, the real danger is not that it will necessarily be cruel. The danger is that long before that point, AI will already be in the hands of humans, and humans want profit, power, control, markets, victory, and weapons. When nearly superhuman tools are given to very human motivations, we don't need machines to rebel. The owners are enough.