AI News You Need to Know — June Edition: Capex, Inference, & Beyond

The AI trade keeps broadening across the stack, and the last few weeks have delivered a dense run of news. Below, we sort the items that matter into three layers: the capital being committed to build (capex); where AI runs and gets monetized (inference); and what it is starting to do in the real world (beyond). Each maps to companies we track in our ROBO Global Robotics & Automation Index (ROBO B) and ROBO Global Artificial Intelligence Index (THNQ B-). CapexThe spending base keeps climbing. Hyperscalers — the largest cloud operators, such as Microsoft (MSFT), Amazon (AMZN), Alphabet (GOOGL), and Meta (META) — are on track to spend roughly $725 billion this year, the large majority aimed at AI. The industrial economy is firming up beneath them, with the ISM Manufacturing index reaching 54% in May, its strongest reading since 2022. A notable shift last week was in the shape of the spending. The largest players are no longer just renting capacity, as they are now funding their own supply chains.  SoftBank (SFTBY), the Japanese technology investment conglomerate and a holding in both our ROBO robotics index and THNQ artificial intelligence index, is the clearest case. It committed up to €75 billion to build AI data centers in France, its largest European investment to date. Meanwhile, it is reportedly preparing to spin out a new company, Roze AI (still private), to deploy robots that build data centers faster. Add its pending acquisition of the robotics division of Swiss industrial group ABB (ABB), its supermajority ownership of chip-design firm Arm Holdings (ARM), and its stake in OpenAI, and SoftBank is now funding nearly every layer at once: the chips, the robots, the buildings, and the power. The image nearly writes itself, with robots built to build the homes for the AI that will run the robots.  IBM (IBM), the enterprise technology and computing company, is running a smaller version of the same playbook. It pledged more than $10 billion to quantum computing over five years. It is also building its own quantum chip foundry — a new subsidiary called Anderon, in Albany, New York, alongside the U.S. Commerce Department — rather than waiting for a supply chain to mature. A fault-tolerant quantum computer, one that can catch and correct its own errors, is the long-sought milestone. IBM is targeting it by 2029. The read for investors is that the biggest spenders are integrating vertically. That’s a tailwind for the pick-and-shovel names across the chain. Those suppliers profit regardless of which platforms ultimately win.InferenceThe marquee item here was Snowflake (SNOW), the cloud-based data and analytics platform. With AI demand plainly high, the question for investors is now who captures the economics of that usage. Snowflake makes a strong claim. Because it bills customers based on how much they use its platform, its revenue is a direct read on real AI activity. Last quarter, that activity surged. Product revenue rose 34%, the strongest sequential dollar growth in company history, and management raised its full-year forecast. The stock rose roughly 37%. Two details matter more than the headline beat. Snowflake signed a fresh $6 billion commitment to Amazon. That’s a reminder that a software platform monetizing AI still pays the cloud for the computing power underneath, so the value splits between the application layer and the infrastructure beneath it. And it agreed to acquire Natoma (private). The startup’s software governs how AI agents (programs that can take actions on their own) connect to a company’s data and tools. That is a move to own the plumbing of automated AI workflows. Within THNQ, Snowflake falls in the cloud providers and big data and analytics segments. Inference is also moving closer to the user. At the Computex trade show, chipmaker Nvidia (NVDA) and Microsoft introduced RTX Spark, a processor that brings the kind of AI computing that used to require a server rack into a thin Windows laptop. More AI will run on the device in front of you, loering latency and cost and keeping sensitive data local. Nvidia’s step into PC chips is a strategic expansion. Taiwan’s MediaTek (2454), a chip designer best known for the processors inside many smartphones, co-designed it, validating its push beyond phones and adding another name to THNQ’s semiconductors segment. The first machines ship this fall.Beyond (Applications — Digital and Physical)This is where AI stops being infrastructure and becomes a product in use, across three very different industries last week. Robinhood (HOOD), the retail brokerage, opened its platform to AI agents. Its Agentic Trading beta lets a bot — built on Anthropic’s Claude or OpenAI’s ChatGPT —  place trades on a user’s behalf inside a separate, walled-off account. Options and crypto are to follow. Handing an AI the keys to a brokerage account, even a sandboxed one, is a threshold for the firm’s 27 million customers. It’s also a sign of how quickly AI that acts on its own is moving from concept to shipped feature. The physical version runs straight through ROBO. Aptiv (APTV), a global supplier of automotive and industrial technology and a ROBO member, expanded its work with Nvidia on production-ready computing for robots and industrial systems. Mitsubishi Electric (MIELY) (the Japanese industrial and electronics group and also a ROBO member) partnered with the Chiba Institute of Technology (a Japanese research university with a respected robotics center) to develop humanoids, walking robots, and drones for factories and infrastructure. These are exactly the cyclical industrial names we have argued the market keeps underpricing as they add AI-driven revenue. In healthcare, Tempus AI (TEM), an AI-driven precision medicine company held in the THNQ index, delivered two items around the American Society of Clinical Oncology (ASCO) meeting: FDA approval expanding its xT cancer genomic test, and a study showing that its AI decision-support tool flagged lung cancer patients who were missing important genetic tests, lifting testing rates by double digits. The decision now happens at the bedside, informed by AI. Bottom LineThe opportunity set is widening across every layer of the stack at once, and the breadth is the point. It is why we hold diversified, supply-chain-aware exposure across ROBO and THNQ constituents and ecosystems, rather than a single bet on any one layer. We will keep watching the spending base, hyperscaler capex and the factory data, for any sign of a crack. For now, there isn’t one. Looking for regular updates? Subscribe here for weekly insights on robotics, AI, and healthcare technology, delivered straight to your inbox. For more news, information, and analysis, visit the Artificial Intelligence Content Hub. vettafi.com is owned by VettaFi LLC (“VettaFi”). VettaFi is the index provider for THNQ and ROBO, for which it receives an index licensing fee. However, THNQ and ROBO 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 THNQ and ROBO.