AI News – December 22-28, 2025: The Bubble Doesn't Burst, It Deflates (And Companies Buy Everything)
While we were unwrapping gifts, ServiceNow spent 8 billion and Databricks dismantled the hype of "zero revenue" companies. The last week of 2025 leaves us with
Every Monday we select and analyze the 5 most significant news stories from the world of artificial intelligence. Not just a simple summary, but a critical reading of the developments that are truly changing the industry.
Why 5 stories? Because around Christmas the background noise increases, but the important signals are few and very costly.
1. The 2025 Balance Sheet: Hype, ROI, and "Zero Revenue" Companies
As we close the year, the debate is heating up: was 2025 the year of the bubble or the revolution? ABC News defines this moment as a historic "turning point." Business subscriptions have doubled, but the return on investment (ROI) remains slow for many.
🔍 What happened: Databricks CEO Ali Ghodsi didn't mince words in an interview with Fortune, calling the valuations of AI companies with zero revenue "insane and circular." Meanwhile, StackOverflow analyzes how software will survive this saturation phase. It's not a crash, but a necessary correction: the market has stopped rewarding flashy demos and has started asking for revenue.
💡 Why it matters: We've moved from the "Wow" phase to the "Excel" phase. Companies no longer buy AI out of fear of missing out (FOMO), but only if it brings measurable efficiency. This cleans up the market: failed startups will disappear, making room for those solving real problems. It's the end of AI's innocence and the beginning of its economic adulthood.
🎯 Our take: The "bubble" won't burst with a bang, but with a whimper. Companies that built business models on hope will be swept away in Q1 2026. Those who built infrastructure and real value (see point 2) will become giants.
Also read: Predictive Economics: If AI Could Anticipate a Financial Crisis
2. Billion-Dollar Christmas Shopping: ServiceNow Buys Armis ($7.75B)
While we were buying gifts, Big Tech was buying entire ecosystems. ServiceNow has acquired Armis for nearly 8 billion dollars, going all-in on cybersecurity and AI-driven asset visibility.
🔍 What happened: This is not an isolated case. HCLSoftware bought Jasperoft ($240M) and Wobby (AI agents for data analysis). Rumors (reported by LinkedIn News) are swirling about final negotiations between Amazon and OpenAI for 10-billion-dollar investments, while Disney integrates generative AI into its global operating model.
💡 Why it matters: This is market consolidation. The big platforms (ServiceNow, Amazon, Microsoft) are buying the missing pieces to offer "end-to-end" solutions. They don't want you to use 50 different AI tools; they want you to use just one, theirs. The Armis acquisition signals that AI cybersecurity is no longer optional, but the foundation of enterprise infrastructure.
🎯 Our take: In 2026 we will see fewer "revolutionary new tools" and more "integrated features" in the platforms we already use. Innovation is being institutionalized (and monetized) by the giants.
Sources: DX Today, LinkedIn Digest
Also read: AI and the Transformation of Traditional Business Models
3. Voice is the New Keyboard: $6.6 Billion in Voice AI
Reuters reports an impressive figure: Voice AI startups raised 6.6 billion dollars in venture capital in 2025, a clear increase from the 4 billion in 2023. Text chatbots are evolving into empathetic audio assistants.
🔍 What happened: Concentrix launched "emotion-aware" conversational agents, capable of adapting their tone of voice to the user's mood. We're no longer talking about simple voice commands like "set a timer," but about complex conversations, podcasts generated in real-time, and interactive voice guides.
💡 Why it matters: The user interface is changing. Writing is slow; speaking is natural. But there's more: digital empathy is becoming a technical feature. If the AI understands from your tone of voice that you're frustrated and changes its approach, the line between human and artificial assistance becomes almost invisible.
🎯 Our take: This opens huge scenarios for customer service, but also risks of manipulation. An AI that "sounds" empathetic is much more persuasive than text on a screen. Get ready to talk to your computers much more often in 2026.
Sources: Reuters, AI Agent Store
Also read: AI for Customer Service: Chatbots That Seem Human
4. Chaotic Science: Duke University's AI Finds Order
While the business world discusses revenue, scientific research takes a quantum leap. A new AI system developed by Duke University has managed to discover "simple rules" within complex chaotic systems, from biology to climate.
🔍 What happened: The AI didn't just process data (black box), but provided interpretable explanations for the observed physical phenomena. It acts as a bridge between the complexity of big data and human understanding, finding patterns that had eluded scientists for decades.
💡 Why it matters: This is the AI we like. Not the one that generates fake images, but the one that accelerates scientific research. If we can use AI to decode climate chaos or cellular dynamics, the social value far outweighs any financial bubble.
🎯 Our take: "Explainable AI" is the necessary frontier. If AI finds a cure or a climate solution, we need to understand why it works. This study shows that AI can be a microscope for logic, not just a stochastic parrot.
Source: ScienceDaily
Also read: AI and Climate: Can Artificial Intelligence Save the Planet?
5. Google and OpenAI: The Year-End Reports
Google closed the year with a recap of 60 key announcements (including Gemini 3 Flash), while OpenAI celebrates the milestone of "one million paying enterprise customers."
🔍 What happened: The two giants are digging the moat. Google is betting everything on efficiency and native multimodality (video, audio, text together). OpenAI focuses on deep enterprise adoption, becoming the invisible infrastructure on which companies build their products.
💡 Why it matters: These numbers tell us adoption is real. One million paying companies is not "hype," it's recurring revenue. The competition in 2026 won't be about who has the most "intelligent" model in the abstract, but about who offers the most stable, secure, and integrable ecosystem.
🎯 Our take: The battle of generalist models is ending. Now the platform war begins. And in this war, whoever controls enterprise data (Google with Workspace, Microsoft/OpenAI with Office) has a lethal advantage.
Sources: Google Blog, OpenAI
Also read: Invisible Competitors: How to Identify AI Threats Before They Arrive
📊 What These Developments Really Tell Us
The Christmas week gives us a lesson in sobriety.
The euphoria of 2024 and the first half of 2025 has faded. In its place is a cold, almost cynical pragmatism. The Databricks CEO calling companies without revenue "insane" is the voice of reason that was missing.
We are witnessing a market bifurcation:
- The Infrastructure: Giants like ServiceNow, Amazon, and Google are buying and consolidating. For them, AI is an industrial asset.
- The Application: The startups that survive are vertical ones (Voice AI, Scientific AI).
There is no more room for "middleware" — those companies that only acted as an intermediary between us and ChatGPT.
2026 will be the year of truth: either AI starts generating real profits for those who adopt it (not just for those who sell it), or the financial "correction" will be brutal. But looking at Duke University and the progress in science, one thing is certain: the technology works. Now it's up to us to make the business model work.
Happy 2026 to all readers of La Bussola dell’IA. It will be an interesting year.
💬 Your opinion matters: Do you think ServiceNow's acquisition of Armis will change the cybersecurity landscape? And do you believe in empathetic voice agents or do you find them unsettling? Write to us or share on social media.
📧 Don't miss the next AI News: Subscribe to the La Bussola dell’IA newsletter to start 2026 with the right analysis.