AI at Davos : Beyond the Model
Lessons from Davos 2026 on AI diffusion a.k.a getting the power of Artificial Intelligence to the masses.
“Miquela Sousa—known simply as Miquela—is a 22-year-old LA-based creative with effortless West Coast energy. A singer, songwriter, and model with Brazilian and Spanish roots, she blends pop, R&B, and electronic sounds, earning millions of streams on Spotify. On Instagram she has over 2.3 million followers, she shares polished fashion shoots, music teasers, and candid reflections on life and growth.”
Miquela is not just some influencer; she herself is AI-generated. Yup, she is not real. (Not in the normal human sense, but in a Terminator Arnold Schwarzenegger voice she is totally real)
This is just one of the extreme examples of influencers using AI for assistance in managing content. PR agencies are using AI for managing influencer discovery and campaign execution. Non technical founders are building and providing services using “vibe coding” - a practice where one uses a conversational AI service like Google AI Studio to explain the product they want to build, and the AI service handles the software development, testing and all things one would expect a freelance software developer to do. AI will also enable hyper-personalization at scale, power solopreneurs to compete with larger firms, and make sustainability initiatives profitable.
The advances in the economy have long been moving hand in hand with technological advances. So, it comes as no surprise that tech leaders have been attending the Davos meet since the 1990s and offering their insights. This year their topic of choice was AI, and the recurring message was that the market is no longer just experimenting with AI, but they’re now focused on execution and ROI. We wrote about this at length in this article last week: AI at Davos
At Davos, Satya Nadella drew parallels between the AI revolution and the introduction of the PC. Steve Jobs described the PC as the bicycle for the mind, one that would reform the entire workflow, driving up productivity in the process. Before we knew it faxes and inter-office memos were replaced by spreadsheets and emails sent to 100s of recipients. AI is considered the next reinvention of this computing stack - one that will redefine how applications are built & used, much like the PC did. It won’t just speed up tasks—it will reshape workflows, decision-making, and individual leverage, allowing people and small teams to operate at unprecedented scale.
Satya is likely hinting that the next step in this revolution is to get the general public aware of AI’s use in everyday lives and thus create more demand. Models like Pika, Claude, or Grok introduce AI. If it’s going to become truly embedded in everyday life, it won’t happen through novelty. It will happen quietly — through schools, hospitals, and public systems that people already rely on. Is this where the state’s role enters the picture?
Luckily, a panel from China was speaking on exactly this - the role of the state. China’s “AI+”, introduced in August 2025 as an initiative to weave AI into the fabric of every aspect of industry and commerce, was allotted a full 30 minute slot on day 3. What stood out about this session was the call to action to increase AI adoption among masses by deploying capital among those who are the farthest from access to it (as opposed to only deploying capital among cash flush companies already at the forefront of AI innovation). While the western world’s focus has been on developing the next cutting edge model, this panel was focused on diffusion and penetration of AI to make the ‘factory of the world’ more productive and efficient. The way to achieve this is through adding AI to elementary education, applying it in agriculture and welfare, and using it to drive up domestic consumption. China is introducing AI and models to children right in their school curriculum.
It’s a race against time however, as the country has set for itself a hefty target, to achieve AI penetration across 70% of key sectors by 2027 - think manufacturing, retail etc. This doesn’t simply mean putting a ChatGPT-enabled 5G smartphone in every hand. It means embedding intelligent, agentic systems into devices and infrastructure across key sectors—tools that can act autonomously, integrate into workflows, and deliver real economic value at scale.
If AI is going to diffuse beyond early adopters, affordability becomes unavoidable — not just at the application layer, but all the way down to infrastructure. This is where China’s approach felt distinct. Historically, its response to large-scale transitions has been to build first and optimize — roads, ports, power, manufacturing capacity. AI appears to be following a similar path.
Their emphasis is not on spectacular breakthroughs, but on lowering the cost of participation. On building enough power and systems so that using AI becomes ordinary rather than exceptional. Many of the efficiency gains discussed on the panel — cheaper inference, optimized deployment, research shaped by early constraints — sounded less like ambition and more like inevitability once that foundation is in place.
If AI really is meant to sit underneath everyday economic activity, then this kind of infrastructure-first thinking matters more than any single model. Infrastructure does not only entail data centers and chips but it has one more building block.
Compared to everyday sources of demand, AI consumes a lot more electricity - the building block whose availability and accessibility shapes the modern world. To meet this increased demand countries will have to look at non-traditional/ renewable sources of energy. When it comes to this China is using its geography to its advantage. Expanding its data centers towards the regions with high wind and solar energies and cooler weather which will help tap into the renewables and reduce costs of cooling respectively.
The expectation is that by 2030 the majority of the AI-related power demand will be met by green energy. Infrastructure being a forethought and not an afterthought is central to China’s ability to scale AI cheaply, sustainably and at speed.
This AI ecosystem is further enhanced thanks to a strong collaborative culture with open-source participation and competitive dynamics, pushing the costs down further. A good example is DeepSeek’s use of sparse Mixture-of-Experts models and memory-efficient inference pipelines.
Companies are encouraged to embrace open-source release strategies, bolstering domestic and international collaboration. The AI engineers, founders and funders do more with less, to optimize performance per watt, per dollar, per GPU.
One of the major contributors to this ecosystem, Tencent - best known for WeChat, PUBG and Tencent Cloud - has made a simple but consequential bet in the AI race. Rather than positioning themselves as a “Model provider” they have taken a path which might matter more in the long run, by offering a “Model agnostic - infrastructure provider”.
Tencent’s position emulates China’s strategy that the winner of the AI race will not be the one who trains the biggest model. The winner will be the economy that offers lower costs, faster diffusion, and deeper integration across the economy.
In a way Miquela’s may not remain novel and rare. AI will not be rare, centralized and reserved for those with technical fluency. AI will become ambient - embedded in classrooms, farms, public spaces, hospitals and even our homes.
Davos 2026 made one thing clear: the next phase of the AI era will not be defined by spectacle, but by execution. The AI race looks less like a sprint for the smartest model and more like a marathon of integration. The economies that understand this early will quietly, but decisively, pull ahead.
 | A guest post by
|
 | A guest post by
|
Keep them coming 👏