The AIOS: An Executive Guide To Mary Meeker’s 2025 AI Trends
A Global Intelligent Operating System Is Being Built
This past weekend, Bond published its 2025 AI trends report (of Mary Meeker Internet trends fame). Mary Meeker’s report is a must-read for anyone working in or near tech/media, and it’s widely viewed as a national holiday for data nerds.
But busy executives who don’t have the time to read through nearly 400 slides of information-dense content require a way to digest the broad strokes. So, I turned to AI to help decode a report about AI and created a CustomGPT. I uploaded the report and trained ChatGPT to rely only on its data while citing sources as much as possible. It comes in handy when asking for summaries or conclusions based on the data in the report. Feel free to try it yourself. Having the report in your pocket using the ChatGPT app is especially convenient. Below is a collaboration between me and the customGPT tailored to an executive audience strapped for time:
Change is Happening Fast
AI is evolving even faster than the Internet did. Tools like ChatGPT exploded globally almost overnight. What used to take decades is now happening in months.
Massive Growth in Users, Usage & Investment
AI tools like ChatGPT hit 800 million weekly users in under 2 years.
Big tech (Apple, Google, NVIDIA, etc.) spent over $212 billion on AI-related infrastructure in 2024, a 63% increase in one year.
Costs to Train AI Models Are High — but Using Them is Getting Cheaper
Training models is expensive and rising.
However, the cost of using AI (per interaction or “token”) is dropping fast, making it more accessible for developers and companies.
Revenue Is Rising — So Are Losses
AI companies are making billions, but many are burning cash to compete, grow fast, and train bigger models.
AI Competition is Global — Especially Between the US & China
China is rapidly closing the gap with the U.S. regarding AI models, infrastructure, and robot deployment.
Open-source models (free to use and adapt) also disrupt the market.
AI Is Leaping Into the Real World
From driverless taxis to restaurant kitchens to health care scribes — AI is no longer just online; it’s showing up in physical spaces and jobs.
AI is supercharging global Internet Adoption
AI tools are helping onboard new internet users faster than ever, especially in regions outside North America.
AI is Changing How We Work — Fast
In the U.S., AI-related tech job postings are up 448%, while non-AI roles are declining.
Industries from banking to restaurants are integrating AI into daily workflows.
Technology is Compounding Like Never Before
Training data, computing power, and AI models are growing exponentially — some by 200–300% per year.
AI is accelerating innovation across nearly every sector.
The Stakes Are High — Benefits and Risks Are Both Real
Upside: AI could help solve major global problems — diseases, climate, and education.
Downside: There are real risks — misinformation, job disruption, surveillance, and safety concerns.
Below is a visualization summary I created using the custom GPT trained on the report data:
Based on the report's content, I worked with the custom GPT to surface a few actionable takeaways. While the conclusions are high-level and not specific to industry, they are nevertheless relevant and worth consideration.
Treat AI as Infrastructure — Not Just Innovation
Implication: Like the internet or electricity, AI will soon underpin every part of your business.
Executive Action:
Invest in foundational AI capabilities (data pipelines, model integration, compute access) as long-term infrastructure, not just point solutions. This includes both internal tools and customer-facing applications.
Invest Ahead of the Curve — CapEx is the New Moat
Trend: Big Tech AI CapEx hit $212B in 2024 (+63% YoY). Winners are building early and at scale.
Executive Action:
Allocate meaningful capital to AI infrastructure now — not just to stay competitive, but to build durable advantage. Partner with leading AI platforms or develop internal R&D where possible. Build versus buy needs to be an active, strategic question.
Prioritize Revenue-Generating AI Use Cases
Insight: AI isn’t just about cost-cutting anymore — it’s increasingly about productivity and revenue growth.
Executive Action:
Champion AI pilots that drive top-line impact (e.g., personalized marketing, AI-enhanced sales, product recommendation engines) over those focused only on automation. Track return on intelligence (ROI) as closely as return on investment.
Make AI Culture & Talent a CEO-Level Agenda
Trend: AI-related job postings up 448%. Non-AI IT roles down 9%.
Executive Action:
Retrain and reskill your workforce and recruit aggressively for AI talent (ML engineers, data scientists, AI product leads). Incentivize experimentation across teams and normalize human-AI collaboration at all levels.
Monitor Global Dynamics — Especially U.S.–China AI Competition
Geopolitical Signal: China is rapidly advancing in models, infrastructure, robotics, and open-source AI.
Executive Action:
Understand how geopolitical and regulatory changes (e.g., AI chip restrictions, data sovereignty, IP protection) could impact your global operations. Plan a scenario for AI supply chain risks. Stay close to evolving AI governance frameworks.
Optional Add-On: Create a Chief AI Officer (CAIO) Role
If AI is as foundational as it appears, it deserves centralized strategic ownership. The CAIO can drive integration across product, operations, and talent, and act as the CEO’s translator between technical potential and business value.
My POV: The AIOS Is Being Built
After skimming the entire report and “conversing” with it through this GPT, I couldn’t help but notice that the report advocates creating an entirely new “operating system.” Take the below excerpt from the final slides:
Catalyzing this growth is the global availability of easy-to-use multimodal AI tools (like ChatGPT) on pervasive mobile devices, augmented by a steep decline in inference costs and an explosion in model availability.
Both closed and open-source tools are now widely accessible and increasingly capable, enabling solo developers, startups, and enterprises alike to experiment and deploy with minimal friction. Meanwhile, large tech incumbents are weaving AI deeper into their products – rolling out copilots, assistants, and even agents that reframe how users engage with technology. Whether through embedded intelligence in SaaS or agentic workflows in consumer apps, the interface layer is being rewritten in real time.
The “AIOS”, AI Operating System, is more than software, hardware, and the ecosystems that will be built around it—it is these things in aggregate, combined with the co-intelligence of humans and machines. It’s a rewrite of the Web and a series of upgrades, patches, and custom builds that will make it through the enterprise:
1. Infrastructure
The physical and digital backbone powering AI capabilities:
Massive increases in compute capacity, fueled by GPU clusters, cloud infrastructure, and specialized AI chips (e.g., from NVIDIA), enable more sophisticated and larger-scale AI model training. AI datacenters—now described as "AI factories"—are being built out rapidly by Big Tech and sovereign nations to support rising inference demand and real-time deployments (Slide 66).
“AI is now part of infrastructure... just like the internet, just like electricity.” – Jensen Huang, NVIDIA CEO (Slide 66)¹
2. Energy
The fuel source for compute-intensive AI systems:
With AI compute needs doubling every 6–10 months, energy has become a gating factor. Powering high-performance GPUs and data centers at scale requires raw energy and increasingly efficient energy systems to manage costs and sustainability. This demand draws interest from governments and utilities, echoing the early internet's infrastructure ramp-up.
AI datacenters are "energy-in, intelligence-out" systems (Slide 66).¹
3. Human Intelligence
Human insight, oversight, and creativity integrated into the loop:
Despite AI’s rise, humans remain central to defining objectives, supplying training data, evaluating model behavior, and guiding ethical boundaries. Enterprise AI adoption (e.g., JPMorgan, Kaiser Permanente, Yum! Brands) often relies on human-AI hybrid workflows where people are augmented, not replaced, by AI (Slides 71–74).
“AI with Grok is getting very good… but it must be programmed with good, truth-seeking values.” – Elon Musk (Slide 65)¹
4. Artificial Intelligence
The cognitive engine—the reasoning, generating, and decision-making layer:
AI models are evolving rapidly across scale, multimodality, and deployment readiness. LLMs like GPT-4o and Claude 3.5 are foundational elements of the AIOS, executing tasks ranging from language understanding to autonomous decision-making. Growth in developer ecosystems (+5x at Google, +6x at NVIDIA) is accelerating capability build-out (Slides 38–39).
“AI is a compounder on internet infrastructure… enabling wicked-fast adoption.” (Slide 21)¹
My ultimate takeaway is that an AI arms race is in play because it’s a race to develop and reinvent the OS that the world will run on. It’s an OS that won’t only compute but will evaluate, reason, and take autonomous action. The architects of this fast-emerging AIOS will be the ones who influence how the world runs in the future.
Visually yours,
