The last week of April 2026 told one story from two angles: massive investment and massive workforce cuts. Four of the world's largest tech companies reported earnings on the same evening, collectively pledging over $700 billion in AI infrastructure spending. Meanwhile, Meta and Microsoft confirmed plans to shed more than 20,000 employees - citing the very technology they're pouring billions into as the reason.
New AI Use Cases
Big Tech restructures - and reduces
Meta confirmed company-wide layoffs starting May 20, 2026. Around 8,000 employees (roughly 10% of headcount) will lose their positions, with another 6,000 open roles cancelled. This isn't a traditional cost-cutting exercise. Meta is reorganizing teams into AI-focused "pods" and transferring engineers into the Applied AI organization under Chief AI Officer Alexandr Wang. New role categories include "AI Builder," "AI Pod Lead," and "AI Org Lead."
Microsoft took a different route. On April 23, the company offered approximately 8,750 US employees (7% of domestic workforce) voluntary retirement packages under a "Rule of 70" formula - the first such program in the company's 51-year history. Separately, 15,000 positions were eliminated. AI and Copilot teams were explicitly exempt. Both companies cited advances in AI automation as driving the reductions, particularly in content moderation, customer support, software testing, and certain engineering roles.
Adobe closes Semrush deal and launches CX Enterprise
Adobe had a busy week. On April 28, it completed the acquisition of Semrush for $1.9 billion in cash. Semrush becomes a wholly owned subsidiary, strengthening Adobe's capabilities in SEO, Generative Engine Optimization (GEO), and Agentic Search Optimization (ASO). Adobe's data shows AI traffic to US retail sites grew 269% year-over-year as of March 2026, yet businesses have significant gaps in AI-driven brand visibility.
Alongside the acquisition, Adobe unveiled Adobe CX Enterprise at Adobe Summit - an agentic AI system for end-to-end customer experience orchestration. CX Enterprise brings together AI agents, agent skills, and Model Context Protocol (MCP) endpoints with an intelligence and governance layer for auditable workflows. The CX Enterprise Coworker component can independently assemble campaigns, create plans, and manage execution after approval. Major agency networks including Publicis, Omnicom, and WPP are already standardizing on CX Enterprise.
SAS expands Customer Intelligence 360 with AI agents
SAS announced the expansion of SAS Customer Intelligence 360 with agentic AI capabilities at SAS Innovate. A multi-agent system features specialized agents for Audiences, Journeys, Email, Search, and Recipes, all operating under a supervisory SAS 360 Agent. Each agent is context-aware, specialized, and designed to remain transparent and controllable. The Journeys Agent accepts multimodal inputs - text, images, and conversational prompts - to build customer journeys.
Infor and AWS bring agentic AI to manufacturing
Infor and AWS jointly announced expanded AI agents for manufacturing at Hannover Messe. The agents are industry-specific and cover six operational areas: profitable project management, on-time delivery, process mining, inventory flow, financial operations, and quality management. As proof of concept, boat manufacturer Xpress Boats achieved a 98% improvement in process issue diagnosis speed, 95% reduction in returns processing time, and 50% reduction in expedited shipping costs.
Databricks launches Unity AI Gateway for agent governance
Databricks integrated AI Gateway into Unity Catalog as Unity AI Gateway. The platform provides MCP governance with fine-grained permissions for agent access to external systems, end-to-end observability for LLM and MCP calls, and Coding Agent Support for tools like Codex, Cursor, and Gemini CLI. Admins get a single governance layer while developers keep using their preferred tools.
Major Vendor Updates
Big Tech earnings: $725 billion in AI investment
April 29 was one of the most significant financial evenings of the year. Alphabet, Microsoft, Meta, and Amazon all reported earnings simultaneously.
| Company | 2026 Capex Guidance | Highlight |
|---|---|---|
| Alphabet/Google | $180-190B (↑ from $175-185B) | Cloud revenue +63% YoY to $20B, backlog nearly doubled to $462B |
| Microsoft | ~$190B (Q4 alone >$40B) | Capacity constrained through year-end, $25B for higher component pricing |
| Meta | $125-145B (↑ from $115-135B) | Concurrent 8,000 layoffs, focus on Applied AI |
| Amazon | Not updated in this window | Part of the $725B combined total |
Together, the four hyperscalers plan roughly $725 billion in capital expenditure for 2026 - up 77% from last year's record $410 billion.
Mistral AI: Medium 3.5 and Vibe Remote Agents
Mistral AI released Mistral Medium 3.5 on April 29, a dense 128B model with a 256k context window. Available as open weights under a modified MIT license, it scores 77.6% on SWE-Bench Verified. It's Mistral's first "merged flagship model," combining instruction following, reasoning, and coding in a single set of weights. Alongside it, Mistral launched Vibe Remote Agents for asynchronous cloud coding sessions and a new Work mode in Le Chat for multi-step tasks.
OpenAI: Workspace Agents and a security incident
OpenAI announced workspace agents for ChatGPT - Codex-powered agents that automate complex workflows, run in the cloud, and help teams scale work across tools. Workspace agents are available in research preview for Business, Enterprise, Edu, and Teachers plans. They're free until May 6, with credit-based pricing starting after that date.
On the security side, OpenAI issued an urgent warning on April 29 for all macOS users to update their desktop apps before May 8. A compromised third-party package (Axios, version 1.14.1) had injected a Remote Access Trojan (RAT) into the macOS signing process on March 31. OpenAI found no evidence of compromised installations but rotated all macOS certificates and published new builds. Older versions will no longer be supported after May 8.
SoftBank plans $100 billion Roze IPO
SoftBank is preparing to spin off and take public a new AI and robotics company called "Roze," according to the Financial Times. Roze will deploy autonomous robots for data center construction and bundle existing energy, land, and infrastructure assets from SoftBank's portfolio, including ABB Robotics, acquired last year. The target valuation sits at up to $100 billion. An analyst day at a Texas data center is planned for July, with KPMG preparing financial documents. The IPO could happen as early as the second half of 2026.
DeepSeek V4 and the race for Chinese chips
DeepSeek released a preview of its V4 model on April 24 with significant upgrades in reasoning and agentic capabilities. The key detail: V4 is optimized for Huawei's Ascend 950 series. Within days, ByteDance, Tencent, and Alibaba began placing orders for Huawei's Ascend 950PR chips. Huawei plans to ship around 750,000 units of the 950PR in 2026.
Separately, DeepSeek is raising outside capital for the first time. Tencent and Alibaba are in discussions to join the round. The valuation doubled within a week from $10 billion to over $20 billion. DeepSeek is offering developers a 75% discount on V4 until May 5.
AI Governance
EU Omnibus negotiations collapse
The second political trilogue between the European Parliament, Council, and Commission ended on April 28 after roughly 12 hours without agreement. The sticking point: Annex I, specifically the conformity assessment architecture for AI systems embedded in regulated products like industrial machinery, medical devices, and in-vitro diagnostics.
The European Commission had proposed the Digital Omnibus in November 2025 to defer the high-risk compliance deadline from August 2, 2026 to December 2, 2027. Parliament and Council had aligned on new deadlines - December 2, 2027 for standalone Annex III high-risk systems and August 2, 2028 for AI embedded in regulated products.
Without formal adoption of the Omnibus before August 2, 2026, the original AI Act provisions apply as written. A further trilogue is scheduled for May 13. Companies operating in the EU or hiring EU-based candidates are in scope. The clock is ticking.
US: White House National Policy Framework
The White House published its National Policy Framework for Artificial Intelligence on March 20. The framework explicitly recommends against creating any new federal rulemaking body for AI regulation. Instead, it calls for AI to be governed through existing regulatory agencies with subject-matter expertise and industry-led standards. The most consequential section recommends federal preemption of state AI laws that "impose undue burdens," aiming for a single national standard rather than fifty different ones.
Swiss Focus
CHUV pilots Meditron in the emergency room
The Lausanne University Hospital (CHUV) begins pilot testing Meditron in May 2026, a Swiss medical large language model developed by EPFL's School of Computer and Communication Sciences. The model will support healthcare staff in clinical decision-making, starting in the emergency room. Meditron comes in two versions (7B and 70B parameters) and was trained on high-quality medical data including peer-reviewed literature and clinical practice guidelines.
ETH Zurich leads global deep tech rankings
According to the 2026 European Deep Tech Report, ETH Zurich ranks first globally for alumni-founded European deep tech startups since 2020, with 192 ventures - nearly triple Cambridge (67) and over five times MIT (35). Zurich also leads globally in robotics. In 2025, the Robotics and AI Institute (RAI) opened its second center in Zurich, following Boston.
Breakthrough Research
Loughborough: Blueprint for transparent AI
Researchers at Loughborough University published a mathematical blueprint for AI systems that reveal how they learn, remember, and make decisions, in the journal Physica D: Nonlinear Phenomena. Their prototype learned musical notes and short phrases without supervision and identified colors from visual data - all in a traceable way, without "catastrophic forgetting" or false memory formation. Professor Natalia Janson said the team rethought AI from the ground up and built a system where cognition's inner workings are fully transparent. The prototype still needs scaling for real-world use.
AI solves 42-year-old math problem
On April 28, Ernest Ryu, a senior researcher at OpenAI, discussed a notable case on an OpenAI podcast: a difficult problem about the convergence of a classic optimization algorithm had remained unsolved for 42 years. AI models cracked it. The models have reached International Mathematical Olympiad level and are beginning to provide substantial help on research-level problems.
Stanford AI Index 2026: Agent leap
The Stanford AI Index 2026 documented a dramatic advance in AI agents: success rates on real computer tasks jumped from 12% to 66%. Agents can now navigate software and systems almost as well as humans and are production-ready.
Conclusion
This week made the gap between AI investment and job displacement tangible. $725 billion flows into infrastructure while over 20,000 positions vanish at Meta and Microsoft. The companies justify both with the same technology. That's not a contradiction - it's the logic of automation becoming visible.
In the model competition, new dynamics are emerging. Mistral Medium 3.5 as an open-weights model scoring 77.6% on SWE-Bench Verified puts Europe on the board. DeepSeek V4 bets on Huawei chips, triggering a rush of orders from Chinese cloud giants. The geopolitical split in AI deepens.
The EU is stuck in a regulatory bind. The Omnibus negotiations failed, the August deadline approaches, and companies don't know which rules will apply. The next trilogue on May 13 becomes a pivotal date.
Ready to explore how Agentic AI can transform your organization? Visit us at https://www.callista.ch/agentic-ai to discover how we can guide your journey into this exciting new era of AI-powered productivity.
Sources
- Fortune - "Microsoft, Meta, and Google just announced billions more in AI spending" - April 29, 2026
- CNBC - "Google wraps up best month since 2004 as earnings push Alphabet stock up 34% in April" - April 30, 2026
- CNBC - "Meta Q1 2026 earnings report" - April 29, 2026
- Tom's Hardware - "Google, Microsoft, Meta, and Amazon capex spending to hit $725 billion in 2026" - April 30, 2026
- The Next Web - "Meta cuts 8,000 jobs and cancels 6,000 open roles as $135B AI spending reshapes the company" - April 2026
- CNBC - "20,000 job cuts at Meta, Microsoft raise concern that AI-driven labor crisis is here" - April 24, 2026
- Fortune - "Why companies like Microsoft are offering voluntary buyouts instead" - April 26, 2026
- Adobe Newsroom - "Adobe Completes Semrush Acquisition" - April 28, 2026
- Adobe Newsroom - "Adobe Redefines Customer Experience Orchestration Vision with CX Enterprise" - April 2026
- SAS Press Release - "Expanded agentic AI capabilities coming to SAS Customer Intelligence 360" - April 28, 2026
- Amazon Press Center - "Infor and AWS Bring Agentic AI to Manufacturing at Enterprise Scale" - April 2026
- Databricks Blog - "Expanding Agent Governance with Unity AI Gateway" - April 2026
- Mistral AI - "Remote agents in Vibe. Powered by Mistral Medium 3.5." - April 29, 2026
- Pulse24.ai - "Mistral AI Ships Medium 3.5 Model" - April 29, 2026
- OpenAI - "Introducing workspace agents in ChatGPT" - April 2026
- VentureBeat - "OpenAI unveils Workspace Agents, a successor to custom GPTs for enterprises" - April 2026
- OpenAI - "Our response to the Axios developer tool compromise" - April 29, 2026
- The Hacker News - "OpenAI Revokes macOS App Certificate After Malicious Axios Supply Chain Incident" - April 2026
- CNBC - "SoftBank reportedly weighs $100 billion valuation for new AI and robotics spinout" - April 30, 2026
- TechCrunch - "SoftBank is creating a robotics company that builds data centers" - April 29, 2026
- CNBC - "DeepSeek releases preview of long-awaited V4 model" - April 24, 2026
- Technology.org - "Huawei Ascend 950 Orders Surge as DeepSeek V4 Drives Chinese Cloud Giants to Buy" - April 29, 2026
- IAPP - "AI Act Omnibus: What just happened and what comes next?" - April 2026
- Modulos.ai - "EU AI Act Omnibus: The Trilogue Failed" - April 2026
- SWI swissinfo.ch - "Artificial intelligence in Switzerland: what's new in 2026" - 2026
- Loughborough University - "End of 'black box' AI? Scientists develop blueprint for transparent systems" - April 2026
- Stanford HAI - "Inside the AI Index: 12 Takeaways from the 2026 Report" - 2026