Here is a 2025 AI round‑up with seven big themes, from spectacular wins to painful flops.​

1. Mega‑infrastructure & Stargate

• The U.S. “Stargate” plan put roughly $500 billion into AI infrastructure, backed by OpenAI, SoftBank, Nvidia and Oracle, signalling an unprecedented build‑out of data centers and supercomputers.​
• This cemented AI as critical national infrastructure, but also supercharged worries about energy use, concentration of power and an AI investment bubble.​

2. New frontier models

• 2025 saw waves of next‑gen foundation models: OpenAI’s GPT‑5‑series, Google DeepMind’s Gemini 2.5 “thinking” model, and Chinese lab DeepSeek’s ultra‑efficient LLMs that rattled Western rivals.​​
• DeepMind’s AlphaEvolve also broke ground by autonomously discovering validated theorems in theoretical computer science, highlighting AI as a scientific collaborator rather than just a chatbot.​

3. Platformization & AI app stores

• OpenAI’s 2025 DevDay turned ChatGPT into a platform with an Apps SDK and an ecosystem for third‑party mini‑apps, explicitly competing with Apple’s App Store and Google Play.​
• This was commercially successful in user growth and developer interest, but raised concerns over lock‑in, data access, and whether small developers can thrive under platform rules.​

4. Enterprise adoption vs grim failure rates

• Surveys suggested strong enterprise uptake of generative AI and agents, with many firms moving beyond pilots into production workflows in areas like coding assistants and customer support.​
• At the same time, analysis showed a dramatic rise in failed AI projects, with around 42% of businesses scrapping most of their AI initiatives, and some reports claiming up to 95% of pilots never made it to production.​

5. Job cuts, backfires and botched automation

• Companies publicly linked nearly 50,000+ layoffs to AI, stoking social and political backlash over “AI‑driven” job cuts.​
• Some high‑profile automation gambles flopped, such as a major bank in Australia sacking call‑center staff for a voice‑bot, only to rehire after the bot failed and call volumes spiked.​

6. Controversies, deepfakes and abuse

• 2025 saw a long list of AI controversies: political deepfakes, copyright fights (including protests over alleged “mass piracy” of authors’ works by big labs), and biased or unsafe deployments.​
• Particularly worrying were AI‑driven health‑insurance denial systems that allegedly rejected claims at much higher rates than human reviewers, prompting regulatory scrutiny.​

7. Governments race to regulate

• Governments, including the UK, advanced AI regulatory “blueprints” and sandbox‑style Growth Labs aiming to speed up beneficial AI (e.g., planning approvals, NHS waiting‑time reduction) while managing risk.​
• Globally, 2025 was a pivot from soft, voluntary frameworks toward harder rules and licensing, especially after visible harms and high‑profile AI failures.​

If you want, a follow‑up could zoom in on one area (e.g., model breakthroughs, labour impacts, or regulation) and map winners vs losers in more detail.

Sources:

1 https://etcjournal.com/2025/10/13/three-biggest-ai-stories-in-october-2025/

2 https://www.crescendo.ai/blog/ai-controversies)

3 https://mashable.com/article/biggest-ai-announcements-2025

4 https://www.raconteur.net/technology/ai-scrollytelling

5 https://techtonicshifts.blog/2025/12/07/the-ten-biggest-ai-fails-of-2025/

6 https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai

7 https://www.youtube.com/watch?v=BpAlN1KBHEQ

8 https://www.techfunnel.com/information-technology/why-ai-fails-2025-lessons/

9 https://www.gov.uk/government/news/new-blueprint-for-ai-regulation-could-speed-up-planning-approvals-slash-nhs-waiting-times-and-drive-growth-and-public-trust

10 https://www.whitecase.com/insight-our-thinking/ai-watch-global-regulatory-tracker-united-kingdom