Why Are Big Tech Firms Taking People and IP Instead of Buying Companies?

Big Tech’s AI race is not only a battle over model performance.

Who can hire more talent?

Who can absorb AI startup technology faster?

Who can pull cloud, robotics, wearables, voice, video, agents, and security technologies into its own platform?

The European Commission’s list of gatekeeper acquisitions under the Digital Markets Act, or DMA, makes this competition visible.

Under the DMA, companies designated as gatekeepers must inform the European Commission of certain acquisitions involving digital-sector services or data collection. The Commission then publishes non-confidential summaries of the information submitted by gatekeepers. These summaries show what platform giants such as Alphabet, Amazon, Apple, ByteDance, Meta, and Microsoft are buying, what assets they are taking, whom they are hiring, and which intellectual-property licenses they are securing.

On the surface, it looks like a list of merger notifications.

But when we look more closely, the picture becomes far more interesting.

Big Tech is not merely buying entire companies. It is making job offers to specific employees, hiring selected engineers, obtaining non-exclusive IP licenses, acquiring certain assets, and carving out particular businesses. Sometimes, the parties maintain that the transaction is not a “concentration” under the EU Merger Regulation, while still notifying it under the DMA.

In the AI era, Big Tech expansion is becoming smaller, faster, and more fragmented.

Google offers a representative example.

In Alphabet’s entries, Google notified a transaction involving Contextual AI in April 2026. The transaction included the hiring of certain software engineers and a License and Release Agreement involving a non-exclusive license to relevant Contextual AI intellectual property. Contextual AI is described as an enterprise software company that helps businesses create and deploy AI agents based on their own data through retrieval-augmented generation, or RAG.

The notice states that the transaction is submitted on the basis of Alphabet’s view that it does not constitute a concentration under the EU Merger Regulation, because it does not involve the acquisition of a business to which turnover can be attributed.

This structure matters.

Google does not buy the whole company.

It brings in some key engineers.

It licenses relevant IP on a non-exclusive basis.

But it does not treat the transaction as the acquisition of a revenue-generating business.

This may be a structure that exploits a gap in traditional merger-control frameworks.

In the past, competition authorities mainly examined mergers through the lens of share acquisitions, asset acquisitions, mergers, and transfers of control. But in the AI industry, a company’s value does not necessarily lie only in revenue or customer base. Key talent, model-training know-how, data pipelines, agent frameworks, evaluation tools, unpublished research outcomes, and software stacks may be even more important.

A company does not have to be bought in full for the core of its technology to move.

Google’s 2025 notice involving Windsurf shows the same pattern. Alphabet entered into a License and Release Agreement with Exafunction, known as Windsurf, which included a non-exclusive license to Exafunction’s intellectual property and the release of employment restrictions. Exafunction is described as a company providing AI-powered software development tools and AI integrated development environments. This notice also stated that Alphabet maintained its view that the agreement was not a concentration under the EU Merger Regulation and that the information was submitted for notification purposes.

This structure is especially important in the AI coding-tools market.

Coding agents and AI IDEs can become core interfaces that control developers’ workflows. Google’s securing of access to the IP and talent of an AI development-tool company such as Windsurf is not merely an addition of software functionality. It is part of a broader contest for control over the development environment of the AI era.

Google has also notified transactions across AI music, audio AI, 3D generative AI, cloud security, quantum computing, smart glasses, UI generation, and LLM evaluation tools.

For example, in December 2025, Google notified a transaction involving the acquisition of certain assets of Corpusant, including Producer.ai, an AI music generation and editing tool, and the possibility of making job offers to employees. Corpusant is a San Francisco-based AI music generation startup.

In the same month, Google notified a transaction involving Hume AI that included job offers to certain employees and a non-exclusive IP license. Hume is described as a New York-based audio AI research lab and technology company. Here, too, Google maintained that the transaction was not a concentration under the EU Merger Regulation.

In January 2026, Google also notified the acquisition of Common Sense Machines. CSM is described as a technology startup that uses generative AI models to create three-dimensional digital objects, enabling 3D content generation for design, gaming, and visualization applications.

The direction is clear.

This is not simply Google the search and advertising company.

It is Google absorbing AI agents, AI coding, AI music, audio, 3D, cloud security, smart glasses, quantum computing, and interface-generation technologies all at once. AI is not ending with text models. It is expanding into developer tools, enterprise data, content generation, spatial computing, security, and hardware.

Amazon’s list shows a different direction.

Amazon appears especially focused on robotics, logistics, satellite communications, customer-service automation, and wearable AI. In April 2026, Amazon notified the acquisition of 3Laws Robotics. 3Laws develops safety autonomy software solutions used primarily in robotic systems. On the same day, Amazon also notified the acquisition of Globalstar. Globalstar is described as a provider of satellite communications products and services, wholesale satellite capacity, government services, and terrestrial spectrum and network solutions.

Amazon’s AI is closer to logistics and the physical world than to creative tools.

In March 2026, Amazon notified the acquisition of Fauna Robotics, which is described as developing robots for general consumers. In the same month, Amazon notified the acquisition of RIVR, a company developing hardware and software solutions used for robotic delivery of parcels and small cargo.

In July 2025, Amazon notified the acquisition of Bluush, known as Bee. Bee develops wearable devices with integrated hardware and software AI systems.

What Amazon is trying to secure is AI’s ability to move the physical world.

Conversational customer service, robotic delivery, consumer robots, wearable AI, barcode recognition, and satellite communications all align with Amazon’s existing strengths. If e-commerce, logistics, cloud, devices, and the Prime ecosystem combine with AI, Amazon can expand beyond online shopping into physical delivery and home and field automation.

Apple’s list belongs to another world.

Apple is gathering small technologies related to sensors, audio, video production, interfaces, photonics, image editing, databases, authentication, observability, and on-device AI. In June 2026, Apple notified a transaction involving Huxe that secured hiring rights for certain employees and a non-exclusive IP license. Huxe provides a mobile application for personalized audio content.

In May 2026, Apple notified a transaction involving Sonera, which included certain assets and employee hiring related to a company developing chip-based magnetic sensors. In January 2026, Q.ai was described as a company developing sensor-based technology for voice, biometric, and audio interfaces.

Apple appears to be absorbing AI less as a giant chatbot and more as a device experience and sensory interface.

PromptAI is described as developing visual intelligence for the home. Animato develops virtual avatar software for video chat and tutoring. Pixelmator, a company developing image-editing and design software for Mac, iPhone, iPad, and Vision Pro, also appears on Apple’s acquisition list.

Apple’s strategy looks less like acquiring a massive AI platform and more like collecting the components that shape product experience.

Sensors, avatars, image editing, photonics, embedded databases, permissions software, machine-learning monitoring, natural-language search, and on-device visual quality inspection may each appear small on their own. But each can connect to the iPhone, Mac, Vision Pro, Apple Watch, Siri, Apple Intelligence, productivity apps, and developer tools.

Meta is moving toward AI agents, robotics, voice, wearable memory devices, and AI-native social spaces.

In April 2026, Meta notified a transaction involving Assured Robot Intelligence. The company develops robotic-intelligence software that enables humanoid robots to autonomously learn and perform complex physical tasks.

In March 2026, Meta also notified the acquisition of certain assets of Confer. Confer operates an AI chat interface that provides access to large language models in a way that protects user conversations through end-to-end encryption. In the same month, Meta notified a transaction to acquire Moltbook from Octane. Moltbook is a platform designed for AI agents to post, comment, discuss, collaborate, and create topic-based subgroups.

This entry is especially interesting.

It suggests that Meta is interested not only in social networks for people, but also in social and collaborative spaces for AI agents. A space where AI agents, rather than human users, write posts, comment, discuss, and collaborate suggests that the subject of social media may expand from humans to agents.

Meta has also shown interest in wearable memory devices and voice AI.

In December 2025, Meta notified the acquisition of Limitless AI. Limitless is a U.S. company that provides a wearable device designed to remember users’ online and real-world conversations in order to help with productivity tasks. In August 2025, Meta notified the acquisition of WaveForms, a developer of audio-intelligence technology. In July 2025, it notified the acquisition of PlayAI, which builds human-like voice models for text-to-speech, AI characters, and conversational AI voices.

Meta’s direction is clear.

AI is not merely a chatbot inside a messenger.

It is expanding into glasses, wearables, voice, avatars, robots, agent communities, and enterprise AI.

Meta’s social platforms could become spaces not only for human users, but also for AI agents and human-machine interaction.

Microsoft’s list is closer to enterprise AI, governance, legal, financial, and data tools.

In March 2026, Microsoft acquired Blocktool, also known as Fintool. Fintool develops financial AI research tools. In December 2025, Microsoft notified the hiring of several employees from Robin AI and stated that it was considering potential asset acquisitions. Robin is a UK company that develops AI-powered software for automating and simplifying legal processes.

In November 2025, Microsoft acquired Pi Labs. Pi Labs develops tools for evaluating and improving AI application quality and agent performance. In the same month, Microsoft acquired 42Layers, known as Osmos, which provides data-engineering tools that automate data cleaning, transformation, and integration.

This combination aligns closely with Microsoft’s enterprise AI strategy.

Financial analysis, legal automation, AI application quality evaluation, agent performance improvement, and data-engineering automation are all enterprise-workflow domains that can connect to Copilot, Azure, Microsoft 365, Dynamics, GitHub, LinkedIn, and Microsoft’s security and compliance products.

Microsoft’s most symbolic case is Inflection.

The DMA list states that in March 2024, Microsoft hired Mustafa Suleyman and 64 Inflection employees and entered into certain commercial arrangements with Inflection, including a non-exclusive IP license. The notice also explains that Microsoft submitted the information while maintaining its view that the transaction was not a concentration under the EU Merger Regulation. Inflection is described as an AI studio specializing in the development of large language models and AI chatbots, with a focus on the B2B market.

The Inflection case marked a turning point in AI-era merger control.

Microsoft did not buy the company, but it secured access to key talent and IP. It did not acquire control, but core resources of competitiveness moved. Should such a transaction be viewed as a traditional merger, a hiring arrangement, a technology license, or a new form of market concentration?

The EU DMA list reveals exactly this gray zone.

Under Article 14 of the DMA, the Commission receives notifications of digital-sector transactions by gatekeepers and publishes summaries. This does not automatically mean prohibition or approval. But it does expose Big Tech’s technology-absorption pathways to markets and regulators.

This transparency matters for three reasons.

First, it shows the map of Big Tech’s AI expansion.

Alphabet is expanding into AI agents, coding, cloud security, music, audio, 3D, quantum, smart glasses, and UI generation. Amazon is moving into robotics, logistics, satellite communications, customer service, and wearable AI. Apple is collecting sensors, interfaces, audio, video production, image editing, and on-device technology. Meta is moving toward robotic intelligence, AI-agent platforms, wearable memory, voice, and AI characters. Microsoft is absorbing financial, legal, data, AI-evaluation, agent-performance, and privacy tools.

Second, the form of corporate combination is diversifying.

Traditional 100% equity acquisitions still exist. But hiring rights, selected employee hiring, specific asset acquisitions, non-exclusive IP licenses, releases of employment restrictions, License and Release Agreements, joint ventures, and acquisitions of specific businesses appear repeatedly. This shows that Big Tech is choosing different transaction structures depending on regulatory risk, speed, and the scope of technology it needs.

Third, AI startup exit strategies are changing.

In the past, the typical startup exit was an acquisition or IPO. Now selected teams may move to Big Tech, IP licenses may be granted, companies may remain or dissolve, and products may be absorbed or discontinued. In particular, AI research talent and model-development know-how can move without the whole company being acquired.

This is the AI version of the acqui-hire.

But it is more complex than simple talent recruitment. In AI, talent alone is not enough. Training code, data-processing methods, model-evaluation systems, prompt and agent orchestration know-how, customer-deployment experience, internal tooling, and IP rights move together. That is why hiring and licensing are combined.

For regulators, this creates a difficult problem.

Startups with little revenue may not meet traditional merger-notification thresholds. But if such a startup holds technology important to future competition, Big Tech’s absorption of it may affect long-term competition. In AI markets, which are still forming, future potential may matter more than current revenue.

The DMA’s notification obligation is one response to this problem.

Regardless of whether a transaction is immediately prohibited, it forces gatekeepers to disclose what they are buying and whom they are hiring in the digital sector. This becomes an early-warning device for market monitoring and competition policy.

The repeated language in the list is also important.

“Without prejudice to Alphabet’s view.”

“Does not constitute a concentration.”

“Non-exclusive license.”

“Potential hiring of certain employees.”

“Certain assets.”

These phrases reveal the tension between Big Tech and regulators.

Companies say, “This is not a merger.”

The DMA says, “Tell us anyway.”

Companies say, “It is a non-exclusive license.”

Regulators ask, “Have core technology and talent moved?”

Competition policy in the AI era must look beyond the legal form of transactions and examine their substantive effects.

The important question is not only whether a company was acquired.

Did a team that could have become a future competitor disappear?

Did key research personnel move into Big Tech?

Were important IP and know-how absorbed into a particular platform?

Did a technology that could have grown independently in an open ecosystem enter a giant platform?

As a result, did the choices available to users, developers, and enterprise customers narrow?

Of course, not all acquisitions and hiring arrangements are harmful.

Startup technology may reach the market faster when combined with Big Tech infrastructure. Researchers and engineers may gain access to larger computing resources and distribution channels. Users may receive more stable products. Security, legal, compliance, and global operations that would be difficult for small companies may become possible.

The problem is cumulative effect.

One or two acquisitions may help innovation. But when dozens of AI startups, employees, IP rights, and assets move into gatekeepers, market structure changes. The space for an independent AI ecosystem may shrink, and core infrastructure and talent may concentrate in a small number of firms.

The EU DMA list makes that cumulative effect visible.

AI music moves toward Google.

AI wearables move toward Amazon and Meta.

AI coding and agent tools move toward Google and Microsoft.

Robotic intelligence moves toward Amazon and Meta.

Audio and voice AI move toward Google, Apple, and Meta.

AI evaluation and quality tools move toward Microsoft and Google.

This is not merely a list of individual transactions.

It is a map of the restructuring of the AI industry.

Wearables and agents are especially striking.

Amazon has Bee. Meta has Limitless. Google has AdHawk and HTC-related talent. Apple is gathering sensor, audio, and interface companies. This shows that AI is moving beyond smartphone apps and combining with devices worn on the body.

AI agents show the same pattern.

Contextual AI is an enterprise-data-based AI agent platform. Moltbook, acquired by Meta, is a platform where AI agents can post, comment, discuss, and collaborate. Microsoft’s Pi Labs develops tools for evaluating AI application quality and agent performance.

This tells us where the next stage of AI competition lies.

Not the model itself, but the environment in which the model works.

Agents that access enterprise data.

Coding tools that help developers.

Devices that remember reality through wearables.

Intelligence that enables robots to act in the physical world.

Quality tools that evaluate and control AI apps.

Big Tech is buying or absorbing all of these layers.

Ultimately, the DMA acquisition list converges on one question.

Where should competition be protected in the AI era?

If we look only at the model market, we may be too late.

If we look only at cloud infrastructure, it may be insufficient.

If we look only at app markets, we may be seeing only the surface.

The real competition is taking place simultaneously across talent, IP, data, agent execution environments, wearable interfaces, robot control, evaluation tools, and security infrastructure.

This is why the EU requires gatekeepers to notify digital-sector transactions.

Platform companies are gatekeepers. If gatekeepers quietly absorb future competitive technologies and talent, markets may tilt before competition even begins. The DMA list at least creates a public record of those movements.

That record is not yet a judgment.

But it is a map.

And the map points in a clear direction.

Big Tech does not necessarily buy AI companies outright.

It takes the people it needs, licenses the IP it needs, buys only the assets it needs, and carves out only the businesses it needs.

Legally, companies may argue that this is not a merger.

Industrially, however, it is a movement of core capabilities.

AI-era merger monitoring must now examine this movement.

The key question is not whether a company name disappeared.

It is where the possibility of competition went.

The key question is not only whether equity changed hands.

It is whose platform now contains the technology, talent, and execution capability.

The EU DMA gatekeeper acquisition list is a public ledger of that shift.