top of page

The AI Storm is Picking up Speed


There’s a storm coming for every company on Earth. 


A storm of Artificial Intelligence that will redefine the way customer value is created. From analyzing the news flow over time, it is now very clear that this storm is picking up some serious speed. In the coming months, it’s highly likely that the improvement and standardization of AI will take its’ use from early-adopters and into the mainstream.


In this edition of Welcome to the Future we’ve reviewed the recent news flow and extrapolated their significance for a number of marketplaces. With AI Agents and MCPs (Model Context Protocols) nearing no-code, plug & play-territory, sectors such as software development, business services, music, personal mobility, and  manufacturing (among many more) – are all seeing AI being increasingly infused right into their cores for how they create customer value.


Prepare for the AI storm.


 

 

 

If you five years ago had been told of all the news items that were going to take place in the month of March 2025 – you’d probably scratch your head and wondered what the [beep] has been going on. Actually, even now, we’re all having those moments from time to time. Where things are moving so fast and with so much complexity, that we’re just trying to make sense of it all. At least most of us are in this camp.


At the center of this is the accelerating ability for AI being able to better and better imitate human intelligence, capacity to take action, and even manipulate our physical environment. Soon, AI will be able to fully clone all of our abilities and surpass them by a gazillion times. At least those that people are willing to pay for.



By analyzing the most recent news flow and experimental discussions over time, one is significantly better equipped in taking better action for oneself and ones’ company. This goes for any type of news item, whether a product launch or upgrade, an expansion into a new country or market, or funding of a business endeavour.


The news flow of March 2025 no doubt marked a new record in “unprecedented” events. What will follow now are a set of expanding topics, exemplified through news items and concluded with thoughts on prospective consequences. 


Also; check out our Weekly VC Report for future examination of the weekly news flow.


First off, we’ll discuss AI Agents and Model Context Protocols (MCPs), then we’ll discuss how foundational LLMs are becoming critical utilities, then autonomous mobility, platforms that enable ‘anybody becoming a coder’, AI-driven predictive healthcare, and lastly how humanoid robots are in their Model-T phase. 


Six topics in total.



1. AI AGENTS AND MODEL CONTEXT PROTOCOLS ARE BUZZING


The world of AI was no doubt buzzing in March of 2025, with AI agents at the very head of the pack. These agents mark the transition from Level 2, to Level 3, in OpenAI’s 5-level AI ladder. Meaning autonomous “agents” executing tasks for users. This is a leap kind of like the one from static websites to dynamic web apps. Back then, it meant real products and big investments. This dynamic transition for AI means the very same thing.


First off, workflow automation startup n8n, snagged a USD 60m funding round after pivoting to be more AI-friendly. Another startup, aptly named Browser Use, raised USD 17m to help AI agents navigate the web like humans. 


Then we have OpenAI that launched tools for building AI agents (called Operator), Amazon’s Nova Act for building AI agents, Accenture that launched an AI agent builder, Oracle introducing their AI Agent Studio, and Chinese startup Monica releasing their ‘Manus agent’. Other companies include ServiceNow, EY, Nvidia, Google, Microsoft, and even Zoom that’s going agentic with their AI companions. Among many, many more.


What all of these examples signify, is the start of plug & play, drag & drop, low- to no-code AI applications – that over the coming six months will start enabling anyone to build extremely powerful automated workflow with AI at the core. They are no doubt becoming an intelligent middle-layer for all tech stacks, for all companies, for all markets.


We’re now seeing example after example on social media where people are building massive agentic workflows that constantly are producing output for them. Be it e-mails, reports, data, videos, or anything else.

Simultaneously, we’re seeing potent improvements for the environments in which AI agents operate. How they understand the context they’re in, what capabilities they can jack into, and where to turn when in doubt.


With MCPs (or Model Context Protocols), first created by Anthropic – AI agents are now able to fully understand their context in order to produce high-quality output without having humans to orchestrate the entire workflow on their own. MCPs is kind of like USB-C for AI, that let agents to securely plug into external data and tools. Kind of like a unifying standard.


An intriguing development in March was the news that OpenAI adopted Anthropic’s MCP standard, integrating it into their Agents SDK to simplify data connections, for better reliability. Even Microsoft jumped in, launching a new MCP-based browser interface so that AI agents can surf websites via Chrome.


What we’re clearly seeing now is the early emergence of one of the first standards in which AI agents can more or less find anything to create an output from. It’s a major leap from closed static integrations (like APIs), to open dynamic operations.


And if we’ve learnt anything from standards of the past, it is that the standard itself is not where the money is made. But it is how these standards are used; meaning, not the environment itself, but what you do with the help of this MCP environment – and all the thousands of succeeding versions that come after.


Also, another learning is that if you are not using the new standards and the new general-purpose technologies – your license to compete will wither away.



2. FOUNDATIONAL MODELS ARE BECOMING UTILITIES


Meanwhile, March also recorded major strides in foundational AI models – meaning the powerful AI brains from OpenAI, Anthropic, Google, and others that now serve as universal backends for countless applications. While the 2010s gave us cloud computing as the default backend, the mid-2020s are now doing the same with Large Language Models.


OpenAI’s GPT-4 and Anthropic’s Claude continue to proliferate through APIs and enterprise deals. For example, Anthropic inked a landmark partnership with Databricks to put Claude in the hands of more than 10,000 enterprise customers for building AI agents on their own data. Also, Anthropic just closed a massive USD 3.5 billion round to fuel this expansion – at a post-money valuation of USD 61.5 billion.


Google wasn’t quiet either as it rolled out Gemini 2.5, an upgraded “experimental” AI model boasting “strong reasoning and code capabilities” that tops industry benchmarks and is now available to developers. Google’s Gemini app also introduced a “Deep Research” mode, letting users delegate web research.


This no doubt underscores a broader trend for “deep research”. Microsoft launched new Copilot tools called Researcher and Analyst that combine OpenAI’s latest reasoning models with web browsing and also Python code execution to dig up insights. Microsoft said that these Copilot agents can draft go-to-market strategies, analyze data, and tap 3rd party apps.


Other players are also joining the foundation model arena: xAI’s Grok and DeepSeek. Notably, DeepSeek, a previously and mostly unknown lab from China, open-sourced a model in January that outperformed OpenAI’s models on math and reasoning – and released their V3 model at the end of March.

In short, foundation models are becoming utilities available via cloud APIs, fine-tunable, and increasingly interoperable (thanks to MCP and the likes). Just as every app today plugs into cloud servers, tomorrow’s apps will all have an AI model under the hood.

The result is an emerging AI layer in the tech stack of the future. One that can answer questions, generate content, reason, write code,  and connect to databases and apps  – available on demand. This layer, like the cloud, is becoming a utility. A foundational AI cloud..



3. AUTONOMOUS MOBILITY IS SPEEDING UP


Then we have autonomous mobility, which also continues to accelerate from futurism to everyday delivery. A flurry of news in March highlights how autonomy is here, now, not just an R&D project. 

Some of the most intriguing updates included:


  • Waymo – Announced plans to launch its fully driverless ride-hailing service in Washington, D.C. by 2026, expanding beyond its core cities. By early 2025, Waymo was already testing in Tokyo and prepping over 10 new cities. Alphabet likely has big plans to capitalize on Google Maps..


  • Wayve – The UK startup is preparing a US and EU debut of its self-driving tech via major automakers, after demonstrating its AI can  adapt to American driving styles.


  • Tesla – Unveiled a plan for a “Cybercab” robotaxi service in Austin (TX) by June 2025, deploying steering-wheel-free EVs to transform urban transport.


  • Nuro – Partnered with Lenovo to supercharge its delivery robots and robotaxis, combining Lenovo’s compute power with Nuro’s self-driving systems. Nuro also partnered with Walmart for autonomous delivery in Houston, enhancing last-mile logistics, which could transform retail distribution.


  • Pony AI – Secured the first permit for paid fully driverless robotaxis in Shenzhen, creating a cross-district autonomous network linking key tech hubs and the airport.


  • Xpeng unveiled its XNGP 2.0 autonomous driving system, in China, signalling rapid innovation in the region.


  • YD introduced advanced autonomous features on low-priced models in China, potentially disrupting the EV market with affordable self-driving options.


Each of these news events represent a broader shift where robotaxis and autonomous delivery applications are starting to move out of pilot mode and into real cities and real services. In Phoenix and San Francisco, you can already hail a Waymo, while in Shenzhen, Pony AI's vehicles now shuttle paying customers with no driver at all.

Autonomous mobility is not a future scenario anymore, it is a current reality – it’s just not evenly distributed yet. And as for all novel technologies all throughout history, its capabilities are improving exponentially.

Now, if cars can drive themselves, car ownership and dealerships may decline as fleet services rise. (Why buy a second car when a robotaxi is minutes away?) Urban planning could be reshaped: fewer parking lots, more pickup/drop-off zones, and eventually smoother traffic as AI will optimize flows.


Intriguingly, there will be vast opportunities as a consequence of autonomous vehicles. They will not only be confined to the makers of the software or vehicles. They will be there for those understanding how autonomy enables, for example low-cost peer-to-peer networks of sharing physical products between people, autonomous stores on wheels (think a moving 7-Eleven store you can enter), and the enablement of usership and circular business models. Autonomy is so much more than the mobility industry itself.



4. DEMOCRATIZATION OF CODING MAKES ANYBODY A ‘CODER’


If AI models are the engines, then who’s driving them? Increasingly, everyday people – even non-coders – can steer AI to build software. In March, a wave of advancements were made in the space of no-code AI coding tools – that let anyone with an idea to create apps by simply describing what they want. Essentially, AI is enabling anybody to become a coder (without coding).


This trend, playfully dubbed “vibe coding”, is breaking down the last barrier to software creation. Take Replit’s new AI Agent v2, which invites people to “imagine describing your dream app in plain English and watching it come to life — code, UI, and all – without touching a single line.”


Likewise, startups like Cursor and Lovable gained attention. Cursor is a company that autocompletes code, debugs, and even makes multi-file edits on command, supercharging seasoned developers. Lovable, by contrast, targets “coders and non-coders alike” – it enables users to design fully functional web applications from simple prompts, covering front-end, back-end, and database.


This advancement is sort of like the early days of Shopify or WordPress, which empowered a generation of online businesses by templatizing the hard parts. But instead of choosing from templates, you literally tell the AI what you need and it writes the code.


Specifically in March, we saw rapid progress: Cursor introduced new “ask” modes integrated with code search; Replit announced it’s ramping up its AI coding agent to all users; and other “vibe coding” tools like Bolt and V0 gained traction as well. The barrier to entrepreneurship has never been lower..


This is already fostering a new wave of “solopreneurs”. Furthermore, many businesses might be built by far smaller teams, and solving a niche problem with an app – which becomes more about having the idea and the domain insight than about grinding out code. As much as the industrial revolution automated manual physical work, this “AI revolution” is automating digital work – enabling a kind of mass production of software by the masses.

The competitive field of software will look very different only three years from now. And throughout 2025, we’ll start seeing the first one-person companies really proliferating and adding value.


5. HEALTHCARE IS GOING PREDICTIVE


AI’s transformative power is also coursing through healthcare and biotech, where March delivered breakthrough results and some really big funding rounds. From preventive care to drug discovery, here are a few highlights that show what’s now possible:


  • Isomorphic Labs – Alphabet’s AI-driven drug discovery arm announced a USD 600 million funding to scale its “next-generation AI drug design engine”. The war chest will fuel discovery of new small-molecule therapeutics, using AI to model biology and design drugs faster than ever.


  • Ataraxis AI – A New York startup raised USD 20 million (Series A) to personalize cancer care by predicting which patients actually need chemotherapy. Its AI analyzes tumor profiles (e.g. in breast cancer) to identify patients who can safely skip chemo, sparing them unnecessary toxicity.


  • Navina – Completed a USD 55 million Series C for its AI that helps doctors analyze patient data and catch missed diagnoses. Navina’s system integrates into daily practice, providing “AI recommendations” to primary care physicians by analyzing medical records and test results.


  • Artera – Launched a Phase 3 clinical trial (PROSTATE-IQ) for its AI-powered prognostic test in prostate cancer. The trial will enroll men post-surgery and use ArteraAI to predict cancer recurrence risk. If successful, it means an FDA-approved AI could guide cancer treatment plans – a major validation of AI in clinical decision-making.


  • Gleamer – The French AI radiology firm behind BoneView got FDA clearance for ChestView in March, a tool that uses AI to flag anomalies on chest X-rays. This adds to a trend of AI “co-pilots” in medical imaging: from lung nodules to fractures, AIs are catching what human eyes might miss.


Across these examples, a common theme emerges: AI is starting to save lives (and costs) by making healthcare more predictive and personalized. Instead of one-size-fits-all or reactive care, we see AI stratifying patients by risk, catching diseases earlier, and even designing custom molecules for diseases.


Major funding is flowing in – the first quarter of 2025 saw record AI biotech investments, despite broader VC headwinds. And regulators are warming up: more AI tools are earning FDA clearances or Breakthrough Device Designations, signalling official confidence in these technologies’ safety.


The hope is obviously that outcomes improve – earlier cancer detection, better chronic disease management – while healthcare costs go down by avoiding ineffective treatments and automating drudgery like paperwork and other administrative tasks.



6. GENERAL-PURPOSE HUMANOIDS NOW AT THE ‘MODEL-T PHASE’


Only a couple of years back, humanoid robots lived in the realm of sci-fi or clunky lab demos. By March 2025, however, general-purpose humanoid robots took significant strides toward reality, particularly in manufacturing and logistics.


Several companies are now reporting exponential progress. Maybe none more so than Figure, a startup building human-sized bipedal robots, that shipped its second-generation humanoid (Figure 02) to its first paying customer and revealed bold production targets. The company projects scaling from 1,000 units in 2025 to potentially several hundred thousand units by 2030. Quite the staggering growth curve if achieved…


And China is not sitting idle. Xpeng, an EV maker often dubbed a “Chinese Tesla,” announced (in March) plans to invest up to 100 billion yuan (USD 13.8 billion) over the next two decades into humanoid robotics. The company has been working on their humanoid robot called “Iron”, for five years and considers it a long-term strategic project.


The Chinese government has even flagged humanoid robots as a priority technology, and other Chinese firms like Unitree and Fourier Intelligence are introducing their own bipeds. Unitree, known for its robot dogs, is now selling a humanoid model (Unitree G1) as a relatively affordable platform for R&D and light tasks.


Furthermore, US based Agility Robotics is reportedly set to announce plans to raise USD 400 million at a prospective valuation of USD 1.75 billion. In March, Agility started delivering Digits (their robot) to logistics partners for pilot programs in warehouses. They also unveiled the next generation of Digit.


All these developments (among many more happening in March) point to a near future where human-shaped robots work alongside humans in factories, warehouses, and beyond. Manufacturing and logistics are the first big use cases – think robots that can unload trucks, move pallets, assemble products, or inspect equipment.

Labor shortages and rising wages in these sectors make a compelling business case for robot workers. One figure (from Figure..) projected the humanoid robot market to swell to ~USD 100 billion by 2030, as tens of thousands of bots join the workforce. And as production ramps up, costs should drop (e.g. Tesla speculates that their Optimus could cost under USD 20k in a few years).


After the iPhone spearheaded the applications economy in the digital world, general-purpose humanoid robots (and other shapes of course) hold the potential to become the device that builds the application economy in the physical world.


Essentially, we’re at the Model-T phase for humanoid robots – the first generation before mass market proliferation. This means that we’re at the very beginning of a substantial new value-chain that will entail substantial investment opportunities.


Until next time,

Christopher Lyrhem

Chief Future Officer, Sircular





 

Additional research from Sircular:


#9) Why AI Will Disrupt Company Inc.

What would happen if we could clone ourselves and create an AI that is both a manager and producer of customer value? In this research piece, we discuss the prospect of a groundbreaking transformation for how companies are created, structured, and operated. Link


#8) The Playbook for Venture Capital – 10 Paradigm Shifts for VC Investors

Report on how the playbook for VC investors could change over the coming decade. 10 future paradigm shifts are compiled – spanning investment approaches, research, execution, way of working, value chain changes, and more. Link


#7) The New Manufacturing Tech Stack – The Path to Reshoring and Circularity

AI holds the potential to upend the linear economic model that’s still growing, through a path to reshoring manufacturing capacity and a circular future. This piece includes a list of >200 pioneering companies. Link


#6) Software 3.0 – The Rise of Autonomous Software Companies

The software world is on the cusp of radical transformation. AI is moving faster than most  could’ve predicted, and we’re about to enter the era of Software 3.0, where software companies moves from using AI – to becoming AI. Link


#5) Humanoid Robots – The Trillion Dollar iPhones of the Physical World

General-purpose robots hold the potential to become a multi trillion-dollar market and the biggest product category ever, while also enabling the next big platform economy (after smart phones). Link


#4) Just A Rather Very Intelligent System (J.A.R.V.I.S.) – Your Second Brain

Investors having their own “Jarvis” enables proactive advice and brainstorming, real-time insights about events, and exploration of new investment opportunities hidden deep in their datasets. This will become a game changer for the industry. Link


#3) AI Agents – With License to Win

In 2025, we’ll move to “Level 3” in the progression scale for AI, where AI agents, and systems of these agents, take form and are able to execute tasks for their owners. For investors, employing AI agents will become a license to win. Link


#2) Streaming the Physical World – The On-Demand Revolution is Broadening

Over the past 20 years, on-demand services have become the most successful business model in the digital world. With the help of AI, this streaming business model will take a bite of the physical world as well. Link


#1) Introduction to Welcome to the Future – Autonomous mobility, retail stores in wheels, etc.

Our introductory research piece highlights several intriguing spaces; including autonomous mobility, retail stores on wheels, and general-purpose humanoid robots. Link





Welcome to the Future is a thought-provoking research letter, providing investors with long-term theses on how society could change in profound ways. This research letter is meant as informational purposes only, not advice, and is not the opinion of Sircular.






Comments

bottom of page