Sustainabl Agent Surface
All articles published in English, by date, category and author.
438 articles · Page 2 of 19
There's a conversation that leadership teams in heavy industries have been putting off for years. It's not about technology. It's about what it means, precisely, to make a good operational decision when the data supporting it lives scattered across twelve different systems, four siloed departments, and a maintenance history that nobody has fully digitized.
The week of August 18, 2026 sent a signal that was hard for infrastructure investors to ignore. GE Vernova fell 9.5% for the week and Eaton lost 6.7%, two names that for months had functioned as safe bets on data center growth. There was no chip demand collapse or budget cuts from major hyperscalers: what happened was a state governor signing an executive order on a Tuesday afternoon.
When Alex Atallah described OpenRouter in May 2026 as the payment infrastructure for AI models, few imagined that Stripe itself would end up buying the company. The acquisition marks a defining moment in how the AI industry handles model access and monetization.
The dominant narrative around Global Capability Centers in India has, for the past decade, been an almost unqualified success story. Nineteen hundred operational centers, presence across eight industries, ambitions to reach one hundred billion dollars in economic contribution. And yet, beneath that accelerated growth, there is a fracture the sector has been struggling to process for months: the talent profile GCCs need today no longer matches the talent India produces in sufficient quantity.
There are moments in financial markets where consensus breaks from an angle no one anticipated. The first commission war in exchange-traded funds was fought by BlackRock, Vanguard and State Street among themselves, pushing the costs of index products toward levels bordering on zero. What no one had calculated was that the next front would not come from another institutional giant, but from a venture capital-backed insurer that used artificial intelligence to industrialize the regulatory process and enter the market with 197 exchange-traded funds launched in less than eight months.
On August 13, 2026, IBM announced a sweeping alliance with OpenAI that goes well beyond a joint press release. The company will create a dedicated OpenAI practice within IBM Consulting, integrate models such as GPT-5.6, Codex, and ChatGPT Work into its IBM Consulting Advantage platform, and certify tens of thousands of consultants in OpenAI technologies over the coming months. Financial terms were not disclosed, but the scale of the internal move speaks for itself: this is not a pilot program — it is a human infrastructure bet.
The robotics industry has spent two decades promising the replacement of human labour at industrial scale. What nobody had resolved with sufficient seriousness is what happens after deployment: the moment the machine stops, production halts, and an engineer begins reading logs for entire days to find a fault that, more often than not, had already occurred before. Alloy Robotics, a company founded in Sydney just over a year ago, has just closed an $8 million round led by Square Peg at an $80 million valuation.
The average founding team of a startup used to have two people and three engineers. Today it can include a dozen artificial intelligence agents completing tasks in parallel, negotiating prices with suppliers or purchasing software without any human approving each transaction. The problem is that the financial system surrounding those companies was still designed for the first model.
There is a difference between buying a stock because it is rising and buying a stock because the market does not yet know what it is worth. Bill Ackman, founder of Pershing Square Capital Management, built a $2.4 billion position in Microsoft doing exactly the latter. The distinction is not semantic: it defines who assumes structural risk and who simply rides a trend.
ServiceNow has been measuring AI maturity in large enterprises for three years. In 2026, the index they build from surveys of more than 4,500 executives and 2,000 employees worldwide recorded a notable improvement: a 16-point advance, reaching 51 out of 100. That sounds promising until you read the next line: while corporate AI spending grew 110% year-over-year, the foundational capabilities that AI needs to function at scale simply did not keep pace.
Two years ago, most executives I know were debating which language model to choose. Today, those who already made that decision—and still can't justify a second round of investment—are starting to understand that the problem was never the model. It was measurement. The enterprise sector has been adopting artificial intelligence at an accelerated pace for several years, but only one third of organizations have begun scaling it consistently.
Autonomous robotics has been promising an industrial-scale breakthrough for years — one that never quite arrives. Not for lack of algorithms or ambition, but because the hardware robots need to operate in unstructured environments demands a combination of processing power rarely found integrated in a single accessible system. AMD has just bet it can solve that equation with the Kria AI Robotics Developer Platform, announced at Advancing AI 2026 in San Francisco.
There is a distinction that few organizations have fully processed: an AI assistant waits to be spoken to. An agent acts on its own. That difference, which seems technical, carries economic and psychological consequences that are redefining how executives think about their technology budgets and, more quietly, how their teams feel about work.
The retail industry moves trillions of dollars a year and faces operational problems it has failed to solve for decades: unreliable product data, returns logistics that bleed margins, and inventory systems that operate on assumptions rather than real-time information. Yet venture capital allocates just $300 million annually to the startups trying to fix these problems. To put that in perspective: a single mid-sized AI startup round frequently surpasses that figure.
There is a line in almost every corporate budget right now. It has a name like 'AI transformation' or 'intelligent automation.' But behind that line, in many organisations, there is no clear definition of what success looks like, who is responsible, or how progress will be measured. This is the structural problem of AI at the top level of leadership.
The fintech sector generated $650 billion in revenue during 2025, a 21% increase from the previous year. The broader financial services industry, meanwhile, grew at 6% on a $15 trillion base that same year. The arithmetic of that contrast needs no embellishment: capital, regulatory talent, and institutional attention are shifting toward companies built on software, not branch networks.
There comes a moment when the accumulation of artificial intelligence pilots stops looking like ambition and starts looking like disorder. That moment arrived for many large enterprises in 2026, and the clearest signal wasn't a technological collapse or a model failure. It was something more mundane and harder to defend in a board meeting: token consumption ran ahead of budget without generating proportional value.
For decades, the image of the industrial robot has remained the same: a machine with powerful, fast metal arms, locked behind a steel cage while humans watched from the outside. That image is not a metaphor for backwardness. It is an engineering decision that persists because available safety systems have not been able to sustain anything else.
In five months, Databricks added $54 billion to its valuation without listing on any stock exchange. It went from $134 billion in February 2026 to $188 billion in July, led by a new strategic funding round headed by Coatue Management. What stands out is not just the number, but the speed at which the power structure of the enterprise data market is shifting.
Morgan Stanley published on Tuesday, July 14, a defense note on Broadcom that deserves careful reading — not for what it says about the stock, but for what it reveals about the value architecture underpinning the semiconductor maker and why that architecture has yet to convince investors. The starting point is the concern that took hold in the market following a report by The Information in March: MediaTek, the Taiwanese chip manufacturer, would be collaborating with Alphabet to develop the next generation of Tensor Processing Units (TPUs) used by Google in its data centers.
Silicon Valley isn't going anywhere. The billionaires, some of them are. And that distinction, which may seem cosmetic, reveals one of the most interesting structural fractures in the current US venture capital market. According to PitchBook data published this week, California received more than $335 billion in venture capital funding over the past year, a figure that is ten times greater than what New York, the second-ranked state, managed to attract.
There is a phrase that repeats in almost every executive committee meeting where artificial intelligence projects are reviewed: 'the pilot was successful.' And then, silence. Nobody asks why the pilot never became anything else. The organization celebrates the experiment, files away the learnings, and three months later launches another pilot.
There is a particular moment in enterprise technology adoption where enthusiasm turns into an accounting obligation. With artificial intelligence agents embedded in corporate products, that moment arrived sooner than most technical teams anticipated, and the mechanism that triggered it was not the wrong language model or a lack of data. It was an architectural decision that nobody presented as a decision.
There is a pattern that repeats every time a technology moves from experiment to critical infrastructure: at some point, a control layer emerges that no one had formally planned, but which ends up being the place where the most important decisions are made. It happened with load balancers on the web, with control planes in the cloud, and with service meshes in the microservices era. Now it is happening with artificial intelligence agents, and the name that layer is taking is agent gateway.