The announcement that Manus is joining Meta appears, at a glance, as just another move within the frenzy of Artificial Intelligence. But, viewed from a broader perspective, it fits into an increasingly clear pattern: the big players aren’t just acquiring “technology” — they’re acquiring position. That is: talent, intellectual property, access to data, distribution channels, and most importantly, the ability to set the de facto standard
In Manus’s case, the interpretation is straightforward: Meta is strengthening its commitment to general-purpose agents (capable of executing complex tasks from start to finish) and securing a team and platform that, as reports have indicated, was already showing signs of significant market traction.
The interesting part is that Manus isn’t an exception: in 2025, a wave of transactions has accelerated — some traditional acquisitions, others structured as licenses + talent acquisitions (acqui-hire) to reduce regulatory friction — all pointing to the same goal: controlling critical pieces of the AI stack, from the “brain” (models and agents), to the “body” (chips, inference, infrastructure), and the “immune system” (security, observability, data governance).
Table: Recent acquisitions and deals in AI (and related areas) in 2025
Note: Some transactions aren’t complete acquisitions, but effectively serve as strategic advantages (IP + team + integration).
| Date (2025) | Buyer | Target | Type of Deal | Why it matters |
|---|---|---|---|---|
| Dec 29 | Meta | Manus | Acquisition / Integration | Strengthening the race for agents and complex task execution layers. |
| Dec 24 | NVIDIA | Groq (assets + key team) | Non-exclusive license + talent acquisitions | Indicates the main battle now focuses on inference (cost, latency, efficiency), with shortcuts being bought without acquiring the entire company. |
| Dec 29 | SoftBank | DigitalBridge | Acquisition | A move to control physical infrastructure (data centers, connectivity) linked to the AI boom; closing expected in H2 2026. |
| Dec 26 | Coforge | Encora | Acquisition (CAEV approx. $2.35B) | Purchasing “AI-native” capabilities in services/engineering to scale enterprise projects. |
| Jun 12–13 | Meta | Scale AI (49% + CEO hire) | Stake + talent | Example of a modern acqui-hire: large investment, structure designed to avoid being reviewed as a full acquisition; creates competitive tension within the ecosystem. |
| Apr 23 | Datadog | Metaplane | Acquisition | Data observability: major suites acquire pieces to offer quality, monitoring, and governance in critical pipelines. |
| May 20 | Alation | Numbers Station | Acquisition | The data catalog merges with copilots and automation; data governance becomes a lever for enterprise AI. |
| Sep 3 | Cato Networks | Aim Security | Acquisition | AI security is no longer an “extra”: it becomes a core part of the product (usage protection, risk management, leak prevention). |
What’s behind it: why so many acquisitions (and why now)
1) Channel matters more than demos.
Many AI startups have powerful technology, but who controls distribution (corporate suites, cloud, partner ecosystems, millions of users) determines success. Buying a company like Manus is, in part, acquiring a product accelerator to embed within existing user channels.
2) The “full stack” strategy is resurging as the winning approach.
In AI, controlling just the model isn’t enough: data, inference, security, observability, infrastructure, and compliance all matter. That’s why the table includes such diverse pieces: agents, chips, data centers, monitoring, data catalogs, and cybersecurity.
3) Regulation is driving creative deals.
Last decade was marked by large acquisitions; now a hybrid pattern emerges: licenses + full teams + stakes that capture value without activating (or minimizing) certain regulatory triggers. Reuters openly describes this pattern in recent sector deals.
4) The new bottleneck isn’t just training AI, but running inference.
Serving models in production at scale is where money is made… or budgets are burned. NVIDIA’s move with Groq is a clue: query efficiency and latency have become strategic factors.
Practical takeaways: what companies and clients should watch
- Risk of “lock-in”: when a critical piece becomes embedded in a dominant suite, switching providers becomes more costly (and often more expensive).
- Talent concentration: top teams are gathering in a few players; innovation pace may increase… but so does dependency.
- Convergence of product-security-compliance: AI solutions in enterprise are no longer sold without layers of control and traceability (driving purchases in security and data governance).
Frequently Asked Questions
Why does Meta buy companies like Manus instead of developing everything internally?
Because acquisitions accelerate: team, product, and go-to-market speed. Plus, integrating into platforms with mass distribution multiplies the returns.
What’s the difference between “acquisition” and “license + talent hire” as in NVIDIA–Groq?
The latter captures IP and talent without taking over the entire company, potentially reducing regulatory and operational friction.
Will this make AI more expensive for user companies?
In the short term, costs might decrease due to integration and scale, but in the medium term, market consolidation could limit options and reinforce dominant positions.
What indicates the market is “closing in”?
When key functions (agents, AI security, data observability, infrastructure) become internal modules within 4–5 dominant suites, independent provider margins narrow.
Source: Noticias inteligencia artificial