AI's Reality Check: Power, Payments, and People

Published10 Nov 2025

Coverage3 Nov 2025 – 10 Nov 2025

Stories tracked77

Featured5

AuthorPulse24 Desk

Last updated10 Nov 2025

This week’s pulse

This week, the abstract ambition of AI collided with tangible, physical limits. The relentless demand for compute is now creating real-world energy and land-use bottlenecks, forcing leaders to confront infrastructure realities. At the application layer, the shift from conversational to transactional AI is sparking commercial disputes and redefining user interfaces. Meanwhile, the human impact is becoming clearer, with organisations actively retraining workforces and courts delivering landmark rulings on copyright and liability.

The Physical Limit: AI's Thirst for Power Hits the Grid

What happened

The AI build-out is facing a critical power bottleneck. OpenAI's Sam Altman and Microsoft's Satya Nadella admitted they cannot accurately forecast future electricity needs, with Nadella revealing Microsoft has a surplus of GPUs it cannot use due to power constraints. This is creating real-world friction, as seen in Nevada, where a proposed 230-mile transmission line to support new data centres faces significant local opposition. The problem is so acute that one Turing Prize winner has urged governments to mandate that AI companies invest in nuclear energy.

So what

The primary constraint on AI development is shifting from securing chips to securing the power to run them. This moves the strategic focus from capital expenditure on hardware to complex, long-term investments in energy infrastructure and real estate. For builders, this introduces significant new variables and timelines, making grid access and energy policy central to infrastructure planning.

02Agentic AI

Agentic AI: From Answering Questions to Performing Tasks

What happened

AI is moving from a passive information source to an active agent that completes tasks. Google's AI Mode can now book event tickets and appointments directly within search. This shift is creating immediate commercial conflict. Amazon has issued a cease-and-desist letter to Perplexity AI, demanding it stop its browser agent from making purchases on its platform, alleging it violates its terms of service and degrades the user experience.

So what

This marks the beginning of the battle for the 'action layer'. As agents begin to transact on behalf of users, they threaten to disintermediate established e-commerce platforms and workflows. For product leaders, this forces a strategic choice: build your own agentic capabilities or risk an AI-powered intermediary capturing your customer relationship and transaction.

Copyright's New Chapter: From Litigation to Licensing

What happened

The approach to resolving AI's copyright problem is evolving. While disputes continue, with Studio Ghibli demanding OpenAI stop using its content for training, new commercial models are emerging. People Inc. signed a deal to license its content to Microsoft via a new 'pay-per-use' marketplace. In the UK, Stability AI largely won a case brought by Getty Images, with the court ruling its model does not store infringing copies.

So what

The conversation is shifting from existential legal fights to the creation of practical, commercial frameworks. The People Inc. deal provides a template for data as a licensed, metered commodity, while the Stability AI ruling offers some clarity on the legal status of trained models in the UK. For builders, this suggests the path forward may involve budgeting for data as a direct cost of goods sold, rather than relying on fair use arguments.

The Workforce Remade: From Displacement to Defined Roles

What happened

Organisations are moving from speculating about AI's impact on jobs to actively restructuring their workforces. Singapore's DBS Group will no longer hire for roles expected to be replaced by AI, instead retraining existing staff for higher-value advisory positions. Simultaneously, the need for high-quality training data is creating an 'elite gig work' niche for educated professionals hired to refine complex models, a step above low-cost data labelling.

So what

This signals a more mature, two-pronged approach to human capital. On one hand, organisations are making hard decisions about automating routine functions. On the other, new, highly-skilled roles are emerging to manage and improve AI systems. For talent leaders, this requires a shift from generic 'AI literacy' drives to building specific career paths for both displaced workers and new AI-centric roles.

The Great Repricing: Valuations Confront Scepticism

What happened

A disconnect is growing between long-term AI forecasts and short-term market sentiment. While Anthropic projects a massive jump to $70 billion in revenue by 2028, tech stocks are facing a correction. The Nasdaq fell 4% this week amid concerns over high valuations. This caution is hitting specific firms, with property portal Rightmove's shares dropping over 25% after it announced increased AI investment would slow profit growth.

So what

The market is beginning to scrutinise the path from AI investment to profitability. While long-term potential remains high, investors are now demanding clearer evidence of ROI, punishing companies that signal higher costs without immediate returns. This indicates a move from valuing AI potential to valuing demonstrated AI performance, putting pressure on leaders to articulate a clear and credible path to monetisation.

⚡ Quick picks

Faster moves.

Markets 💹: Palantir raised its 2025 revenue forecast to approximately $4.40 billion, driven by strong demand and a landmark $10 billion, 10-year contract with the US Army.

Finance 💷: Workflow AI platform Scribe achieved a $1.3 billion valuation as it launched a new product, Scribe Optimize, to help organisations identify high-impact AI applications.

Risk ⚠️: OpenAI is facing seven new lawsuits in California alleging its GPT-4o model was negligently released and contributed to suicides and severe psychological harm.

Macro 🌍: China is rolling out power subsidies of up to 50% for data centres that use domestically produced AI chips, a clear industrial policy move to accelerate technological self-reliance.

Pulse24’s view

The AI narrative is moving from the theoretical to the terrestrial. The industry's trajectory is no longer defined solely by model capability, but by tangible constraints: available megawatts, the letter of the law, workforce adaptability, and the patience of capital markets. The next phase of value creation will belong not to those with the grandest vision, but to those who master the complex, real-world logistics of execution.