The Agent Economy Is Here. Are You Paying Attention?

For a while, the market could treat AI as a fairly clean story. More compute. Better models. Bigger capex. A familiar set of winners from Nvidia to Microsoft, then data center and energy providers, now memory and photonics as scrutiny heads toward actual chokepoints and technical alpha. 

But in 2026, the AI story looks less like a software upgrade cycle and more like a real-world power struggle over agents, infrastructure, and control. China rolled out a five-year plan that leans hard into AI, open-source software, and low-altitude aviation. Anthropic ended up in open conflict with the U.S. government over military guardrails, and OpenAI lost a senior robotics and hardware leader after its Pentagon deal. And we are now watching data centers become military targets during the Iran conflict, raising hard questions about how to protect these investments going forward.

At the center of the shift from chatbot to agent is a project I flagged in my February 2 article as a potential preview of AGI. It was called Clawdbot at the time. Now it goes by OpenClaw. What has happened since makes it worth understanding properly.

OpenClaw is not best understood as a peer to Claude, GPT, or Gemini. It is an open agent layer that sits on top of those models and turns them from answer engines into action engines. It can connect to your email, calendar, messaging apps, and file systems. It’s able to run 24/7 on hardware you own, and can take autonomous action on your behalf through WhatsApp, Slack, or iMessage. 

It can even operate with fully local models, like Alibaba’s Qwen 3.5 series, where 27 billion parameter models now match what was formerly available only in massive cloud setups. We essentially have somewhat slower superintelligence available on sub-$2,000 setups like Apple’s Mac Minis, which are selling like crazy due to OpenClaw fever. At this point, it comes down to the cost of your hardware, your APIs (data), and energy consumption. 

Since February, OpenClaw has become the fastest-growing project in GitHub history, with over 247,000 stars, surpassing Linux and React. Its plugin marketplace hosts over 13,000 community-built skills. Chinese entrepreneurs are using OpenClaw-style systems to automate storefronts and run “one-person companies.” Local governments in China have moved to support the ecosystem with subsidies, office space, and compute

OpenAI recently acqui-hired Peter Steinberger, the creator of OpenClaw. On March 10, Meta bought Moltbook, the social network for AI agents tied to OpenClaw, pulling the founders into its superintelligence effort. Both halves of the experiment now sit inside two of the largest players in consumer AI.

There is an Ouroboros element worth flagging:  If open-source agents running on consumer hardware can replicate what enterprise platforms charge $20–200/month for, who captures value in the AI stack? The cloud platforms will not disappear. However, the tension (and literal secure connectivity) between edge and cloud, open-source and subscription, a Mac Mini on a desk and a hyperscaler in Virginia, will define the next chapter of AI monetization.

On the infrastructure side, the Iran conflict put a fine point on something the industry has so far danced around. If 20% of global oil supplies can be disrupted by one chokepoint, what happens when the chokepoints are data centers? The answer could lie in hardened data centers (aka: bunkers) powered by small modular reactors. China’s five-year plan calls for “hyper-scale” computing clusters backed by cheap, abundant electricity. 

Regardless, as billions if not trillions are poured into this space, it’s clear that security of these infrastructures will require more oversight and forethought than ever before. 

ROBO and THNQ were built to capture exactly these shifts. THNQ holds the AI infrastructure stack: semiconductors (Nvidia, AMD, TSMC, ASML, Infineon), cloud and compute (Alphabet, Amazon, Microsoft, Nebius, Alibaba), network security (Palo Alto Networks, CrowdStrike, Cloudflare), and data platforms (Datadog, Snowflake). ROBO captures the physical layer: actuation, sensing, logistics automation, healthcare robotics, and autonomous systems

Together, the two indices span both sides of the edge-to-cloud to physical AI spectrum this agentic shift is forcing into focus.

We will be at the Exchange conference in Las Vegas next week (March 15-18), where these themes will be front and center for the financial advisor community. Five races are running simultaneously: the agent explosion, the physical AI arms race, the governance reckoning, the hardening of sovereign infrastructure, and the future of productivity itself. The companies building across these fronts, and the indices that capture them, are not waiting for the macro to settle. Neither should investors.

For more news, information, and analysis, visit our Disruptive Technology Content Hub.

ROBO is the underlying index for the ROBO Global Robotics & Automation ETF (ROBO), the L&G ROBO Global Robotics and Automation UCITS ETF (ROBO.LN), and the Global X ROBO Global Robotics & Automation ETF (ROBO.AU). THNQ is the underlying index for the ROBO Global Artificial Intelligence ETF (THNQ) and the L&G Artificial Intelligence UCITS ETF (AIAI.LN).

VettaFi is the index provider for ROBO ETFs, THNQ ETF, and AIAI.LN, for which it receives an index licensing fee. However, ROBO ETFs, THNQ ETF, and AIAI.LN are not issued, sponsored, endorsed, or sold by VettaFi. VettaFi and its affiliates have no obligation or liability in connection with the issuance, administration, marketing, or trading of ROBO ETFs, THNQ ETF, and AIAI.LN.